Analysis of segmented elderly women’s lower bodies using 3D-LOOK scan data and virtual representation

Abstract

To address the clothing needs of an aging society, this study developed a scale using three-dimensional (3D) scan data that determine elderly women’s lower body shapes to improve the way garments fit the elderly. Body type elements that play an important role in garment fit were identified and five body type elements were selected for use in this study. A stepwise discriminant analysis using 176 dimensions was performed to extract parameters reflecting body shape features, resulting in 37 parameters. A scale for determining body shapes was developed using the discriminant function equation. This study differs from existing studies on body shape classification in that we determined the diverse body shape features of individuals by extracting the lower body type elements related to garment fit. This study demonstrated an organic relationship among lower body types, where a greater posterior pelvic tilt was associated with a protruding lower abdomen, flat buttocks, and an o-type frontal leg shape. The significance of this study lies in the extraction of 3D parameters that reflect the body shape features of elderly women. Such 3D parameter data can be used to create personal virtual bodies in online shopping malls in the future.

Keywords

Assessment scale elderly women body shape garment fit three-dimensional scan data virtual representation

Modern consumers have a strong desire to express their individuality. To respond to diversified consumer characteristics and needs, the clothing industry has focused on made-to-measure (MTM) production methods, where a mass production system is used to create personalized clothing.¹ MTM methods result in customized manufacturing that is made to order to suit the preferences and body types of individual customers.² In clothing production, MTM methods include the concepts of fit customization, where clothing fit is customized to individual consumer’s body types, and design customization, where customized designs and styles are provided to satisfy individual consumers’ preferences.³ MTM methods have been mainly used in the production of men's suits, which have relatively few choices in design compared to women's wear, due to a long order-to-order manufacturing time and increased production cost.^4,5 Advances in three-dimensional (3D) anthropometry technology and IT have reinvigorated efforts in the clothing research field to improve MTM production, accounting for individuals’ body types and preferences.^6
–10 In other words, MTM production, which was difficult due to realistic constraints such as time and cost, is activated due to the convergence of the clothing industry and digital technology, and it is expected that it will be able to quickly produce clothing that reflects individual tastes and fits well.^11
–13

Advanced medical technology, scientific civilization, and improved standards of life have extended the average human lifespan, leading to an increase in the absolute size of the elderly population and the relative proportion of the population that is elderly.¹⁴ The United Nations (UN) Department of Economic and Social Affairs (DESA)¹⁵ estimates that the global population will increase from 7.7 billion persons in 2019 to 8.5 billion persons in 2030, 9.7 billion persons in 2050, and peak at 10.9 billion persons in 2100. Although countries differ, population aging is a global phenomenon.^16,17 Because today’s burgeoning elderly population has greater economic power and opportunities for social and leisure activities than in the past, interest in clothes for the elderly is increasing,^18
–20 with the elderly population’s increased spending power leading to a growing “silver market.” However, elderly consumers often report high dissatisfaction with clothing purchases because sizing systems are not finely differentiated, and it is difficult to find clothing that matches their body shape.^21
–25 According to the UN State of the Future,²⁶ the future clothing industry will be driven by the silver generation and female customers. Therefore, to meet the demands of the aging society and increase elderly female consumers’ buyer satisfaction, clothing manufacturers will need to precisely ascertain elderly individuals’ body shape characteristics.²⁷

Since consumers have diverse body shapes, an investigation of body shapes is necessary to improve the way garments fit.^28,29 Previous studies in the clothing and textiles field have attempted to classify elderly females’ lower body types. Nam and Choi³⁰ classified lower body shapes into five types: “short and fat,” “average height and average weight,” “average height and lean,” “tall and obese,” and “short and slightly lean.” Cha³¹ classified lower body shapes into three types: “narrow ladder-shaped obese type,” “bell-shaped long and slim type,” and “oblong-shaped slightly obese type.” Kim³² classified lower body types into three types: “slim type with an A-shaped contour,” “short and fat type with abdominal obesity,” and “tall and obese type with an H-shaped lower body.” The results of these studies indicate that there is value in investigating overall trends in elderly women’s lower body types. However, because these studies classified body shape broadly into relatively few types, they do not provide an adequate understanding of individuals’ diverse body type characteristics. Moreover, it is difficult to intuitively identify individuals’ body types and use this information to create or modify patterns. Rather, to identify body types and manufacture clothing suited to these body types, it is necessary to gain a more detailed understanding of individuals’ body types.

In the clothing construction field, the aim of body type research is to obtain information about body parts that cause a poor fit, and to provide consumers with better fitting clothing by modifying patterns according to that information.³⁰ For example, protruding buttocks result in creases forming in the direction of the buttocks, which causes poor fit. To eliminate these creases, the back crotch length must be extended to compensate for the pattern shortage. Conversely, in individuals with flat buttocks, the excess back crotch length needs to be reduced.^33
–35 Therefore, and when studying lower body types for clothing construction, it is important to focus on body types that are relevant to the fit of skirts or pants.

Several studies have been conducted to explore the effect of lower body type factors on fit and clothing patterns. Rasband and Liechty³³ classified the lower body types that affect garment fit into 20 categories, including abdominal protrusion, pelvic tilt, buttocks protrusion, and leg curvature. Editors of Creative Publishing³⁶ classified the lower body types into five categories: waist shape, buttock shape, thigh shape, abdomen shape, and seat shape. To explain the clothing alteration method based on body types, Liechty et al.³⁴ presented detailed body types by body parts such as the neck, shoulder, chest, abdomen, and buttocks. By combining the results of these previous studies, the elements of elderly women’s lower body shapes that relate to garment fit can be classified into pelvis, abdomen, buttock, lateral leg, and frontal leg shapes.

The human body has various characteristics and appearances that depend on the skeletal frame size, extent of muscular development, posture, subcutaneous fat thickness, and fat deposit locations. Posture changes depend on the skeleton and affects the body’s external appearance because it adopts the most stable posture to maintain balance, as a whole, based on the organic relationships between individual body parts.^37,38 Pelvis position is key to determining proper and improper posture. If the pelvis is misaligned, whole-body balance is disrupted, which can be a cause of lower body obesity. Anterior pelvic tilt causes protrusion of the abdomen and buttocks, while posterior pelvic tilt causes dropping of the buttocks and protrusion of the lower abdomen even in people without much abdominal fat.^37,39 In old age, overall posture changes as cartilage becomes thinner and spinal deformation becomes more severe. In addition, bent knees in the elderly alter the overall lateral posture by affecting both the upper and lower body. These musculoskeletal conditions are major characteristics of old age.^21,40
–43 The separate parts of the body are organically connected and act together to maintain a balanced posture. Although it is important to study individual body parts, it is essential to identify the connections between the shapes of each body part and reflect this information in clothing pattern design and MTM production.

Clothing side seam lines need to be visually attractive and naturally divide the side of the body, while also maintaining consistency with the side seam line of the skirt or slacks. In clothing construction, the side seam line is the criterion that determines the lateral posture.^44
–46 Previous studies that analyzed lower body types used several methods to define the side seam line position: the side seam line as the vertical line bisecting the apex of the abdomen and the apex of the buttocks (i.e., bisecting the maximum width of the lower body);^4,47 the vertical line bisecting the length of the foot;⁴⁸ or the vertical line bisecting the thickness of the waist.^46,49,50 However, using a vertical side seam line causes problems when converting 3D body scan data into a two-dimensional (2D) pattern, because it is difficult to balance the patterns for the front and back panels. For this reason, Yoon⁴⁵ suggested defining the side seam line as the line connecting the lateral bisection point of the waist and the lateral malleolus to account for slanted lower body posture. Thus, it is important to select an appropriate definition for the side seam line, depending on the aims of the study.

With the recent commercialization of 3D human body scanners, body scan data are being used in high-tech industries, including apparel, automobiles, medicine, animation, and artifact restoration.^47,51
–53 In the apparel field, 3D scanning technology is an indispensable technology for future development of the digital fashion industry and, as such, there is active research on 3D scan data use.^47,53
–57 Although prior studies in the clothing and textiles field have used 3D body scan technology to classify body types or make clothing patterns, most only extracted 2D information from the 3D scan data, such as circumference, length, and height.^31,32,42,58 The initial human body scan data obtained from a 3D full-body scanner comprise numerous point group data with 3D coordinates.⁵¹ Because 3D scanning techniques store the positions of a massive number of surface points in a digital form, they not only provide body measurements that could previously be obtained using instruments such as a Martin anthropometer or tape measure, but they also provide diverse data corresponding to 3D shapes, such as curvature, surface area, volume, and vectors.^59,60 Accordingly, research is needed on how to extract 3D information from such numerous 3D point groups and effectively use them in the clothing studies field.^53,60,61 Cho⁶² explained some important considerations. Firstly, because the initial scan data from a 3D full-body scanner contain coordinate data based on different origins, pre-processing is required to use a single origin before it can be used in research. In addition, because the cylindrical coordinate system is an extension of the Cartesian coordinate system, converting the data to a cylindrical coordinate system could provide more useful data for identifying the surface characteristics of the human body. Cho⁶² asserted that this conversion of the coordinate system would help develop 3D clothing computer-aided design (CAD) systems that can automatically produce clothing patterns from a virtual representation of the human body.

The purpose of this study was to develop a scale to differentiate lower body types in elderly women, to provide basic data to assist personalized clothing pattern manufacturers in a MTM production system. The specific objectives of the study were:
to examine whether lower body shape elements categorized by experts based on visual evaluation accurately reflect the actual body shape characteristics of elderly women;

to examine, out of all the data that can be obtained from 3D body scans, the 3D parameters that best discriminate differences between body shape groups;

to develop a scale that can differentiate lower body types of individual elderly women.

Considering that advances in the IT industry are expected to generalize automated, customized pattern creation using 3D anthropometry and CAD systems, this study will contribute to improving the fit for elderly female consumers by providing clothing manufacturers with information to differentiate the body shapes of target consumers.

Materials and methods

Research sample

This study used body measurement data and 3D scan data collected in the 5th Size Korea Anthropometric Survey,⁶³ which collected 3D human body shape data from 5000 male and female participants aged 8–75 years. The 5th Size Korea Anthropometric Survey was conducted from 2003 to 2004. A Cyberware WB4 whole-body color scanner (Cyberware Co., USA) was used for 3D anthropometry. Researchers placed landmarks on participants’ bodies before whole-body scanning and numerical data were extracted using a whole-body measurement program developed as part of the Size Korea project. Data analysts loaded the scan data into the program and manually marked each landmark on the computer screen, after which the measurements were automatically calculated. The 3D scanning and anthropometry process was performed in compliance with the criteria in ISO 20685-1,⁶⁴ and the landmarks and measurements were collected based on the criteria in ISO 8559-1⁶⁵ and ISO 7250-1.⁶⁶

This study’s research sample comprised the 3D scan data of 91 elderly women, 60 years or older, from the 5th Size Korea Anthropometric Survey.⁶³ The participants’ general characteristics, such as age, weight, and waist size, are shown in Table 1. The study sample ranged in age from a minimum of 60 years to a maximum of 75 years, with an average age of 65.63 years. The average height, weight, and waist circumference were 152.25 cm, 56.87 kg, and 85.49 cm, respectively. The average body mass index (BMI) was 24.51, which exceeded the World Health Organization’s (WHO’s) suggested normal range, falling into the obese category. The WHO normal BMI range is 18.5–22.9, the pre-obese category is 23.0–24.9, and the obese category is a BMI of over 25.⁶⁷

Table 1.
Participants’ descriptive statistics (n = 91)

Body dimension Mean S.D. Body dimension Mean S.D.

Age (years) 65.63 4.13 Calf height (cm) 27.28 3.36

Height (cm) 152.25 4.90 Minimum leg height (cm) 5.23 0.73

Weight (kg) 56.87 7.19 Waist circumference (cm) 85.49 7.93

BMI (body mass index) 24.51 2.78 Waist circumference (omphalion) (cm) 89.97 7.45

Waist height (cm) 91.83 10.38 Hip circumference (cm) 92.10 5.11

Waist height (omphalion) (cm) 86.17 9.89 Total crotch length (cm) 70.03 4.54

Hip height (cm) 71.90 8.16 Thigh circumference (cm) 52.77 4.21

Crotch height (cm) 66.46 7.59 Knee circumference (cm) 33.69 2.66

Thigh height (cm) 76.27 8.90 Calf circumference (cm) 32.08 2.73

Knee height (cm) 38.67 4.42 Minimum leg circumference (cm) 18.70 1.99

Landmarks and dimensions

A total of 44 landmarks reflecting the lower body’s shape features based on ISO 8559-1⁶⁵and ISO 7250-1⁶⁶ were used in this study; their compositions are listed in Table 2. Figure 1 shows the landmark locations on the human body. Landmark names were coded using combinations of letters and numbers, with each letter referring to a specific body part, and each number referring to the position of the landmark in the transverse plane. The number “1” corresponds to the landmark located anteriorly and in the midline in the transverse plane; “2” corresponds to the landmark located posteriorly and in the midline; “3” and “4” correspond to the landmarks at the right- and left-hand inflection points, respectively; “5” refers to the landmark at the most posterior point right of the midline; and “6” corresponds to the most posterior point left of the midline (Figure 1).

Table 2.
Landmark symbols and names

Classification 　　Landmark Classification 　　Landmark

A (Waist area) A1 Anterior waist point E (Hip area) E1 Anterior hip point

A2 Posterior waist point E2 Posterior hip point

A3 Lateral waist point_right E3 Lateral hip point_right

A4 Lateral waist point_left E4 Lateral hip point_left

A5 Posterior-curvature waist point_right E5 Posterior-curvature hip point_right

A6 Posterior-curvature waist point_left E6 Posterior-curvature hip point_left

B (Waist (omphalion) area) B1 Anterior waist point (omphalion) F (Crotch area) F Crotch point

B2 Posterior waist point (omphalion) G (Thigh area) G1 Anterior thigh point

B3 Lateral waist point (omphalion)_right G2 Posterior thigh point

B4 Lateral waist point (omphalion)_left G3 Lateral thigh point_right

C (Abdominal area) C1 Anterior abdominal point G4 Lateral thigh point_left

C2 Posterior abdominal point H (Knee area) H1 Anterior knee point-mid patella

C3 Lateral abdominal point_right H2 Posterior knee point

C4 Lateral abdominal point_left H3 Lateral knee point_right

D (Top hip area) D1 Anterior top hip point H4 Lateral knee point_left

D2 Posterior top hip point I (Calf area) I1 Anterior calf point

D3 Lateral top hip point_right I2 Posterior calf point

D4 Lateral top hip point_left I3 Lateral calf point_right

D5 Posterior-curvature top hip point_right I4 Lateral calf point_left

D6 Posterior-curvature top hip point_left J (Minimum leg area) J1 Anterior minimum leg point

J2 Posterior minimum leg point

K K1 Lateral malleolus

K2 Sphyrion

L1 Pternion

Figure 1.
Landmark locations on the human body and reference lines.

One study objective was to use the landmarks’ 3D coordinates in the 3D body scanning data to analyze body shapes. Because the initial scan data uses coordinates based on different origins, the data must be normalized to a single origin. In this study, the origin of the lower body was defined as the intersection of the sagittal plane and the point line bisecting the waist from left to right. The origin of all scan data was converted to the origin of the lower body. Then, the 3D coordinate values of the 44 landmarks were extracted from 91 pieces of scan data. The dimensions were extracted based on the following coordinate system conversion process: firstly, the distance and angle values from the origin to each landmark were calculated by converting the Cartesian coordinate system (X, Y, Z) to a cylindrical coordinate system (R, θ, Z). The conversion formula was
$\begin{array}{l} R = \sqrt{x^{2} + y^{2}}, θ = {arctan \frac{y}{p} x > 0 and y < 0 \\ arctan (\frac{y}{x} + π) x < 0 \\ arctan (\frac{y}{x} + 2 π) x > 0 and y < 0 \end{array}$
(1)

In addition, the Cartesian coordinate system (X, Y, Z) was converted to a spherical coordinate system (ρ, θ, ф) based on Equation (2), and the distance and angle from the origin to each landmark were calculated
$ρ = \sqrt{(x^{2} + y^{2} + z^{2})}, ϕ = \arccos (\frac{z}{p})$
(2)

A total of 176 dimensions were extracted from the above coordinate system conversions and used for the analysis. IBM SPSS Statistics, version 23.0 for Windows, was used for data analysis. Rapidform 2006 software was used for the 3D scan data analysis.

Visual analysis of elderly women’s lower body shapes

In the clothing and textiles field, previous body shape studies have mostly used methods based on experts’ visual evaluation and statistical analysis of dimensions. An advantage of visual evaluation is that it is possible to differentiate morphological characteristics of individuals’ body shapes that are difficult to extract using statistical methods. Therefore, this technique is commonly used in studies that classify body parts’ lateral postures or shapes.^60,68
–71 However, visual evaluation is a subjective evaluation; therefore, the results may be inconsistent across different experts. Consequently, it is important to establish precise criteria that allow experts to make objective judgments. The rate of agreement between experts in the visual evaluation provides evidence for the degree of objectivity.⁶⁸

In this study, experts’ visual evaluation was used alongside statistical methods to classify elderly women’s lower body shapes. Firstly, visual evaluation defined the lower body shape types; then statistical methods were used to extract the parameters that best explained each type’s characteristics. In addition, statistical methods were used to develop a scale to identify new customers’ body shapes. A theoretical review was performed^{30,33
–36,70} to extract the lower body parts that cause poor clothing fit (e.g., the body parts for which different shapes require different patterns for making skirts or pants).

Three elements related to the torso were extracted from the theoretical literature review: pelvis shape, abdomen shape, and buttock shape; and two elements were extracted related to the lower limb region: lateral leg shape and frontal leg shape.

The experts’ visual evaluation was divided into a pilot experiment and a main experiment. The specific procedure is shown in Figure 2. The experts comprised five people, each with 15 or more years of experience in the clothing studies field, and the visual evaluations were made using the frontal and lateral photo data of 91 elderly women. The objective of the pilot experiment was to set criteria for visual evaluation to classify specific types of lower body shapes. Reference lines were established by analyzing the related literature and prior studies related to the shapes of the lower body parts. Jang and Lee⁴² reported that, when analyzing aging-related knee deformities, changes in the knee shape should be examined using the lines through the mid patella and lateral malleolus as reference lines. Yoon⁴⁵ asserted that it would be appropriate to use the line between the lateral midpoint of the waist and the lateral malleolus as a reference line to account for lower body posture to develop personalized patterns using 3D body scan data. Based on previous study findings,⁴²,⁴⁵ we used three reference lines to analyze body shapes in elderly women and to aid the future manufacture of personalized patterns (Figure 1): the baseline passing through the waist depth bisection point and the lateral malleolus; the baseline passing through the waist depth bisection point, the mid patella, and the anterior minimum leg point; and the baseline passing through the crotch vertical section.

Figure 2.
Expert visual evaluation process.

In addition to these three reference lines, the experts set criteria for visual evaluation based on an overall discussion of the lower body’s outer silhouette. Table 3 shows representative silhouettes for each body type used in the experts’ visual evaluations. In the pilot experiment, the experts inspected participant photographs and analyzed the range of body shapes for lower body parts in elderly women. Next, they specified a three-stage spectrum of shapes for each of the five lower body shape elements: degree of pelvic tilt, degree of protrusion of the abdomen and buttocks, and degree of lateral and anterior bending of the knees. The experts specified three grades for each of these elements (Table 3). The pelvis shape was classified as anterior tilt, straight, and posterior tilt types, and the abdomen shape was classified based on the protrusion shape as lower bulge, upper and lower bulge, and overall bulge types. The buttock shape was classified according to the degree of protrusion as protruding, medium, and flat types. The lateral leg shape was classified based on the degree to which the knee was bent when viewed from the side as hyper-extended, straight, and bent types. The frontal leg shape was classified according to the extent of the gap between both knees when viewed from the front as o-shape, straight, and x-shape types. Except for the abdomen, the three types for each element represent the central values and changes in either direction from the center. For the abdomen, since most elderly women show a protruding abdominal shape, the three types were categorized as upper bulge, lower bulge, and upper and lower bulge. According to the literature,^{21,37,38,41,42} even lean individuals can show a lower abdominal bulge if they have an altered spinal or pelvic position; therefore, this is a common body shape observed in elderly women. Consequently, the abdomen was categorized into these three types to reflect elderly women’s specific body shape characteristics.

Table 3.
Representative silhouettes by sub-type used for experts’ visual evaluation

Lower body shape Sub-type classification Lower body shape Sub-type classification

The pelvis type The lateral leg type

(a) Anterior tilt (b) Straight (c) Posterior tilt (a) Hyper-extended (b) Straight (c) Bent

The abdomen type The frontal leg type

(a) Lower bulge (b) Upper and lower bulge (c) Overall bulge (a) O-shape (b) Straight (c) X-shape

The buttocks type

(a) Protruding (b) Medium (c) Flat

Next, we conducted the main experiment for experts’ visual evaluation. Five experts evaluated the body shapes of the anterior and lateral view photographs of 91 individuals. Participants were only included in the analysis if a consensus was achieved between at least three of the five experts. The level of agreement between the experts’ opinions was calculated using the following equation
$Agreement (%) = \frac{(a \times 5 + b \times 4 + c \times 3)}{N \times 5} \times 100$
(3)
where a is the number of participants matched by all five experts, b is the number of participants matched by four experts, c is the number of participants matched by three experts, and N is the total number of participants (N = a + b + c).

Development of a scale for elderly women’s lower body shape assessment

Body shape-related parameter extraction

Using the body shape types defined by experts’ visual evaluation as a reference, a stepwise discriminant analysis was performed to extract the parameters that best explained the characteristics of each type. Stepwise selection, which combines forward and backward elimination, was used as the parameter-selection method. This method starts with forward selection, but in each step, the previously selected parameters are reevaluated, and those that do not show a sufficient contribution to discrimination are eliminated. However, the eliminated parameters are not completely excluded from consideration, and they can be selected again in the next step. Thus, stepwise discriminant analysis is a good method for extracting parameters with high importance for classifying body types.^72,73

The discriminant analysis model included 176 dimensions to extract parameters with high predictive power for determining lower body shapes. Two methods of body shape classification are available: one using a linear discriminant function and the other using a canonical discriminant function. The parameters extracted from the stepwise discriminant analysis results were inserted as independent variables into both the linear discriminant function and canonical discriminant function. This study used standardized canonical coefficients, which are the coefficients used when calculating discriminant scores by standardizing the data to have a mean of 0 and a standard deviation of 1 before inserting into the canonical discriminant function. The coefficient values indicate the relative importance of each parameter.^72,73

Body shape classification

The accuracy rate of the linear discriminant function was calculated by obtaining the linear discriminant function by group to test the accuracy of each group identified by the linear discriminant function. Fisher’s linear discriminant method was employed, and group identification was performed using variables and constants to construct the linear discriminant function formula and by obtaining the classification coefficients. In addition, a scatter diagram of the classified groups and the participants belonging to the center of each group were selected and presented pictorially.

Participants belonging to the center of each group were those with the mean discriminant scores for that group, and the scan data for these participants were visualized. In addition, to ascertain the relative influence between lower body part shapes, the discriminant scores for each type were included in a Pearson correlation analysis, and the correlation coefficients between pairs of variables were calculated.

Development of a body shape assessment scale

The discriminant function formula was derived by obtaining the canonical discriminant coefficients, and a scale was developed to determine the five body type elements by calculating the average determination score and range for each group. To develop a scale to discriminate elderly women’s lower body shapes, the cut points between adjacent groups needed to be identified, and this required identifying the number of people and the mean discriminant score for each group. We calculated the cut points according to the following equation
$Cut Score = \frac{n_{i + 1} C_{i} + n_{i} C_{i + 1}}{n_{i} + n_{i + 1}}$
(4)
where n_i is the number of groups and C_i represents the group means.

Results

Visual analysis of elderly women’s lower body shapes

Table 4 shows the body shape classification results based on the experts’ visual evaluation of 91 participants’ photo data. The results showed that the pelvis and buttocks had a uniform distribution in each group. The results of the abdominal shape classification showed that 63.74% of participants fell into the overall bulge type. For the lateral leg classification, there were more participants with the bent type than the hyper-extended type. The results of the frontal leg classification showed that the o-shape and x-shape had similar distributions (see Table 4). These results are consistent with the previous research results indicating that, for the elderly, the subcutaneous fat moves down to the lower part of the body, increasing the circumference of the abdomen and buttocks, and that spinal shape change causes bowing of the legs, leading to shape change in the lower limbs.^{4,30,37
–39,41}

Table 4.
Experts’ visual evaluation of elderly women’s lower body shapes

Lower body component Group 1 Group 2 Group 3 Total

The pelvis typeGroup 1: anterior tiltGroup 2: straightGroup 3: posterior tilt 28(30.77%) 32(35.16%) 31(34.07%) 91(100%)

Agreement (%) = {(69 × 5 + 15 × 4 + 7 × 3)/ 91 × 5} × 100 = 93.63 (%)

The abdomen typeGroup 1: lower bulgeGroup 2: upper and lower bulgeGroup 3: overall bulge 19(20.88%) 14(15.38%) 58(63.74%) 91(100%)

Agreement (%) = {(64 × 5 + 15 × 4 + 12 × 3)/91 × 5} × 100 = 91.43 (%)

The buttocks typeGroup 1: protrudingGroup 2: mediumGroup 3: flat 26(28.57%) 30(32.97%) 35(38.46%) 91(100%)

Agreement (%) = {(71 × 5 + 12 × 4 + 8 × 3)/91 × 5} × 100 = 93.85 (%)

The lateral leg typeGroup 1: hyper-extendedGroup 2: straightGroup 3: bent 17(18.68%) 45(49.45%) 29(31.87%) 91(100%)

Agreement (%) = {(72 × 5 + 13 × 4 + 6 × 3)/91 × 5} × 100 = 94.51 (%)

The Frontal Leg typeGroup 1: o-shapeGroup 2: StraightGroup 3: x-shape 20(21.98%) 52(57.14%) 19(20.88%) 91(100%)

Agreement (%) = {(64 × 5 + 14 × 4 + 13 × 3)/91 × 5} × 100 = 91.21 (%)

Note. Average agreement level: 92.93%.

The lowest level of the five experts’ agreement was on the frontal leg shape, at 91.21%, and the experts’ average level of agreement was 92.93% (see Table 4). Based on the experts’ visual evaluation results, the five lower body elements selected through the literature review were shown to effectively reflect elderly women’s body shape characteristics.

Development of a scale for elderly women’s lower body shape assessment

Body shape-related parameter extraction

The stepwise discriminant analysis identified the key independent variables for determining elderly women’s lower body shapes, extracting 37 parameters. The results are shown in Table 5. Variables with a high degree of contribution for determining body shapes included 12 parameters for the pelvis, six for the abdomen, nine for the buttocks, 10 for the lateral leg, and seven for the frontal leg. For the pelvis, the eigenvalue of discriminant function 1 was 7.75, explaining 92.01% of the total distribution, and the eigenvalue of discriminant function 2 was 0.67, explaining 7.99% of the total distribution. Discriminant function 1 had the greatest explanatory power for identifying a group.

Table 5.
Canonical discriminant function results

Pelvis type
Abdomen type
Buttocks type
Lateral leg type
Frontal leg type

Items Function1 Function2 Function1 Function2 Function1 Function2 Function1 Function2 Function1 Function2

RA2 –1.40 0.85 – – –0.78 –0.67 – – – –

RA5 0.48 –0.62 – – –0.01 0.82 – – – –

RC4 –0.44 –0.31 – – – – – – – –

RD5 – – 0.56 0.81 – – – – – –

RE2 – – – – 2.19 –0.07 – – – –

RE5 – – 1.36 –0.29 – – – – – –

RE6 3.30 1.93 – – – – – – – –

RG1 –0.67 –0.31 – – – – – – – –

RH1 – – – – – – – – 1.09 –1.22

RI4 – – – – – – – – 2.82 1.92

RK1 – – – – – – – – –3.21 0.45

RK2 – – – – 0.88 0.87 – – – –

RL1 – – – – – – – – 1.63 –0.81

θA3 0.57 0.32 – – – – – – – –

θC3 – – – – – – –0.53 –0.21 – –

θG1 – – – – – – –0.71 0.01 – –

θG4 0.04 –0.88 – – – – – – 0.48 0.03

θH1 – – – – – – 3.16 –1.12 1.27 0.82

θH2 – – – – – – 1.22 –1.01 – –

θH3 –0.50 0.77 – – – – – – – –

θI1 – – – – – – –2.58 2.92 – –

θI2 – – – – – – 1.09 –0.16 – –

θK1 – – – – – – –0.82 –0.29 – –

ρE1 – – – – 0.06 –1.11 – – – –

ρE2 – – –0.56 –0.26 – – – – – –

ρE5 – – – – – – 0.91 –0.47 – –

ρG2 – – – – – – –0.71 0.00 – –

ρJ1 – – – – –0.75 0.93 – – – –

фB2 – – 0.21 –0.72 0.50 0.33 – – – –

фB4 0.93 0.28 – – – – 0.31 0.68 – –

фC1 – – –0.02 1.06 – – – – – –

фE4 –1.07 –2.00 – – – – – – – –

фE5 1.13 3.76 – – – – – – – –

фG4 – – 0.40 0.28 0.11 0.55 – – – –

фI4 –0.21 –0.63 – – – – – – – –

фJ1 – – – – 0.76 1.24 – – – –

фJ2 – – – – – – – – 0.70 0.68

Eigenvalue 7.75 0.67 1.57 0.69 3.25 1.06 4.06 0.31 3.49 0.15

% of Variance 92.01 7.99 69.45 30.55 75.48 24.52 92.91 7.09 95.90 4.10

Canonical correlation 0.94 0.63 0.78 0.64 0.87 0.72 0.90 0.49 0.88 0.36

Notes. Standardized coefficients.

Gray shade: the item that contributes the most to type determination.

Higher absolute values for the standardized canonical coefficients indicate greater discriminatory power.^72,73 The items with the greatest contribution to discriminating body shapes from lower body parts are highlighted in gray (Table 5). For discriminant function 1, the parameter with the greatest contribution to discriminating pelvis type was RE6, for abdominal type it was RE5, for buttocks type it was RE2, for lateral leg type it was θH1, and for frontal leg type it was RK1. For discriminant function 2, the parameter with the greatest contribution to discriminating pelvis type was фE5, for abdominal type it was фC1, for buttocks type it was фJ1, for lateral leg type it was θI1, and for frontal leg type it was RI4.

For type discrimination, the pelvic, abdominal, and buttock types all involve the torso, and most of the landmarks with a high contribution to discrimination are located on the torso. The lateral and frontal leg types involve the lower limbs; therefore, all the landmarks with a high contribution to discrimination are located on the lower limbs. In addition, parameters that characterize the shape of the buttocks showed a high level of contribution to the discrimination of pelvis, abdomen, and buttocks types. This shows that the shape of the buttocks has a major effect on the body type discrimination in these three areas.

Body shape classification

To test the accuracy of the linear discriminant function for the body shape types of lower body parts in elderly women, the hit ratio of the function was obtained using a classification table (Table 6). The reclassification of the pelvis type according to the linear discriminant function resulted in 27 of the 28 individuals belonging to group 1 being reclassified into group 1; 29 of the 32 individuals in group 2 were reclassified into group 2; and 30 of 31 individuals in group 3 were reclassified into group 3. This result shows an accuracy of 94%. Accuracies of 82.4%, 94.5%, 93.4%, and 91.2% were shown for the abdominal, buttocks, lateral leg, and frontal leg types, respectively. Discriminant analysis results showed that abdominal type had the lowest accuracy because a high proportion of participants in group 3 (overall bulge) were mistakenly classified into group 2 (upper and lower bulge). This shows that there is a need to develop parameters that can improve the discriminative power between the overall bulge type and the upper and lower bulge types.

Table 6.
Body shape classification results (unit: count, %)

Predicted group membership

Pelvis type
Group 1 Anterior tilt
Group 2 Straight
Group 3 Posterior tilt
Total

Group 1: anterior tilt　 27 (96.43%) 1 (3.57%) 0 (0.00%) 28 (100.00%)

Group 2: straight 2 (6.25%) 29 (90.63%) 1 (3.12%) 32 (100.00%)

Group 3: posterior tilt 0 (0.00%) 1 (3.23%) 30 (96.77%) 31 (100.00%)

Percentage accuracy of prediction: 94.5%

Predicted group membership

Abdomen type
Group 1 Lower bulge
Group 2 Upper and lower bulge
Group 3 Overall bulge
Total

18 (94.74%) 0 (0.00%) 1 (5.26%) 19 (100.00%)

Group 2: upper and lower bulge 2 (14.29%) 12 (85.71%) 0 (0.00%) 14 (100.00%)

Group 3: overall bulge 3 (5.17%) 10 (17.24%) 45 (77.59%) 58 (100.00%)

Percentage accuracy of prediction: 82.4%

Predicted group membership

Buttocks type
Group 1 Protruding
Group 2 Medium
Group 3 Flat
Total

Group 1: protruding 24 (92.31%) 1 (3.85%) 1 (3.85%) 26 (100.00%)

Group 2: medium 1 (3.33%) 29 (96.67%) 0 (0.00%) 30 (100.00%)

Group 3: flat 0 (0.00%) 2 (5.71%) 33 (94.29%) 35 (100.00%)

Percentage accuracy of prediction: 94.5%

Predicted group membership

Lateral leg type
Group 1 Hyper-extended
Group 2 Straight
Group 3 Bent
Total

Group 1: hyper-extended 17 (100.00%) 0 (0.00%) 0 (0.00%) 17 (100.00%)

Group 2: straight 2 (4.44%) 39 (86.67%) 4 (8.89%) 45 (100.00%)

Group 3: bent 0 (0.00%) 0 (0.00%) 29 (100.00%) 29 (100.00%)

Percentage accuracy of prediction: 93.4%

Predicted group membership

Frontal leg type
Group 1 O-shape
Group 2 Straight
Group 3 X-shape
Total

Group 1: o-shape 20 (100.00%) 0 (0.00%) 0 (0.00%) 20 (100.00%)

Group 2: straight 2 (3.85%) 45 (86.54%) 5 (9.62%) 52 (100.00%)

Group 3: x-shape 0 (0.00%) 1 (5.26%) 18 (94.74%) 19 (100.00%)

Note. Percentage accuracy of prediction: 91.2%.

From the results of the discriminant analysis of the five body shape elements, we selected the scatter diagrams of the separate groups and the participants belonging to the average of each group. The participants selected in each group were those with mean values for the discriminant score for each type, as shown in Table 7. The participants’ scanned data were visualized and are shown in Figures 3 –7.

Table 7.
Average centroid and number of group samples for each sub-type

Classification Pelvis type
Abdomen type
Buttocks type
Lateral leg type
Frontal leg type

Centroid No. of group Centroid No. of group Centroid No. of group Centroid No. of group Centroid No. of group

Group 1 3.21 28 –2.39 19 2.14 26 –3.14 17 2.91 20

Group 2 0.59 32 0.33 14 0.61 30 –0.43 45 –0.14 52

Group 3 –3.51 31 0.70 58 –2.11 35 2.50 29 –2.68 19

Note. Canonical discriminant function 1.

Figure 3.
All-groups scatterplot and the centroid of each group of 1, 2, and 3 for the pelvis type.

Figure 4.
All-groups scatter plot and the centroid of each group of1, 2, and 3 for the abdomen type.

Figure 5.
All-groups scatter plot and the centroid of each group of 1, 2, and 3 for the buttocks type.

Figure 6.
All-groups scatter plot and the centroid of each group of 1, 2, and 3 for the lateral leg type.

Figure 7.
All-groups scatter plot and the centroid of each group of 1, 2, and 3 for the frontal leg type.

Pearson’s correlation analysis investigated the correlations between variable pairs (Table 8). There were strong positive correlations between pelvis type and buttocks type, pelvis type and abdomen type, and abdomen type and buttocks type. There were strong negative correlations between pelvis type and frontal leg type, abdomen type and frontal leg type, and buttocks type and frontal leg type (Table 8). In other words, a greater posterior pelvic tilt was associated with increased protrusion of the lower abdomen, flatter buttocks, and o-shaped frontal leg type. These findings are consistent with those of previous studies, in which a posterior pelvic tilt flattened the buttocks as the lumbar spine curved backward to maintain a balanced posture; pelvic tilt also affected the knee joints, leading to o-shaped legs.^{37,38,40
–43}

Table 8.
Correlations between lower body shape elements

Classification Pelvis type Abdomen type Buttocks type Lateral leg type Frontal leg type

Pelvis type 1

Abdomen type 0.722 1

Buttocks type 0.861 0.633 1

Lateral leg type –0.204 –0.127 –0.055 1

Frontal leg type –0.318 –0.257* –0.295 0.205 1

The correlation is significant at the 0.01 level (bilateral).

*The correlation is significant at the 0.05 level (bilateral).

Development of a body shape assessment scale

The sample sizes and mean discriminant scores in Table 7 were used to calculate the cut points. For example, the cut point between abdomen type 1 and type 2 was calculated according to the completed version of the canonical discriminant function 1 in Table 5, as shown below

Canonical discriminant function 1 for abdomen type = (0.56 × RD5) + (1.36 × RE5) + (–0.56 × ρE2) + (0.21 × фB2) + (–0.02 × фC1) + (0.4 × фG4)

When the mean measurement values for each type are substituted into the equation, the result is the mean discriminant score for each type, as shown in Table 7. The values were –2.39 for group 1, 0.33 for group 2, and 0.70 for group 3. The mean scores and sample sizes are substituted into Formula 4 to obtain the cut scores. The cut score between abdominal type 1 and abdominal type 2 was –0.82, meaning that participants with a canonical discriminant function 1 score of less than –0.82 were abdomen type 1, while those with a higher score were abdomen type 2. This process was repeated to obtain the cut scores for each type and to develop a scale that could differentiate between the types (Figure 8).

Figure 8.
Development of a body shape assessment scale.

Figure 8 shows the outcome of the scale development for identifying elderly women’s lower body shapes. From the results of calculating the range of discriminant scores that determine pelvis shape by using the discriminant function equation, a range of group 3 ≤ –1.49, –1.49 < group 2 ≤ 1.99, and 1.99 < group 1 was extracted. For the abdomen shape, a range of group 1 ≤ –0.82, –0.82 < group 2 ≤ 0.41, and 0.41 < group 3 was extracted. For the buttocks shape, a range of group 3 ≤ –0.65, –0.65 < group 2 ≤ 1.43, and 1.43 < group 1 was extracted. For the lateral leg shape, a range of group 1 ≤ –2.40, –2.40 < group 2 ≤ 1.36, and 1.36 < group 3 was extracted. For the frontal leg shape, a range of group 3 ≤ –2.00, –2.00 < group 2 ≤ 2.06, and 2.06 < group 1 was extracted.

This study found the parameters reflecting the features of elderly women’s lower body shapes, obtained the linear discriminant function, and the canonical discriminant function that used these parameters as variables, and presented the discriminant score ranges by body type, so that they can be used to determine an individual’s body shape.

Discussion

Owing to the increase in consumer-centric market demand, technological development for designing clothing products is rapidly shifting to consider consumers’ physical characteristics and preferences. We developed a scale to differentiate elderly women’s lower body shapes to prepare for a future age of personalized clothing purchases.

The 3D scanner-obtained human body data were formed as an aggregate of numerous point groups and possessed a very complicated curved surface. To use such 3D data in the apparel industry, the 3D information needs to be simplified and transformed to facilitate control by parameters.^61,62 The significance of this study lies in extracting 3D parameters that reflect elderly women’s body shape features by condensing multiple pieces of 3D data. This 3D parameter data can also help develop 3D clothing CAD systems to automatically produce clothing patterns from virtual bodies.

Of the two methods used to analyze body shapes, experts’ visual evaluation and statistical analysis, visual evaluation was used. Kim⁶⁰ used expert visual evaluation to classify lateral body shapes in adult males and reported an overall agreement across experts of 66.75%. Choi and Nam⁵³ used expert visual evaluation to classify lateral body shapes in adult females and reported an overall agreement of 87.6%. When Kim⁷⁰ used expert visual evaluation in adult males, the overall agreement for abdominal type was 87.49%, and for lateral body shape was 83.22%. In the present study, the overall agreement between experts was 92.93%, which was higher than that reported in previous studies. We believe that our study showed a higher level of agreement because the observations were divided by body part rather than based on the whole-body level. In addition, elderly women may show relatively clearer body shape characteristics compared to other age groups.

Seong et al.⁴⁶ classified the abdominal types of women in their 20s into lower bulge, flat, and bulge types. In the present study, abdominal types were divided into lower bulge type, upper and lower bulge type, and overall bulge type, reflecting the body shape characteristics of elderly women. According to the literature,^21,40
–43 aging is associated with changes in the lateral posture as spinal deformities become more severe, where, even in lean persons, posterior pelvic tilt causes protrusion of the lower abdomen. Therefore, in this study, the flat type was not selected as an abdominal type. Kim⁷⁰ identified four abdominal types in adult males: bulge, slanting, round, and straight. This demonstrates that the abdominal bulging shape has different characteristics, depending on sex.

This study divided elderly women’s lower body shape into five parts, and a three-grade scale was constructed for each part to represent its shape change or shape characteristics. This method is limited in terms of differentiating individuals’ diverse and complex body shapes. Therefore, future studies will need to further subdivide the body shapes for lower body parts. In addition, elucidating the organic relationships between these body shapes for different parts would provide more practical data for clothing companies. This study developed a scale to differentiate individuals’ body shapes. For future personalized clothing production, follow-up studies are needed to connect the results of body shape assessment with clothing pattern making. Furthermore, because this study limited its scope to the lower body, a body shape analysis of the upper half of the body is needed. In addition, this type of body shape analysis needs to be conducted in subsequent research that compares body shapes across different racial groups.

Conclusions

Body shape analysis on body parts related to clothing construction needs to be studied to assist the apparel industry. However, previous studies have classified diverse human body shapes by condensing them into a few types, which is inadequate for understanding individuals’ diverse body types and does not identify shape types of specific body parts. This study differs from existing studies on body shape classification in that it determined the diverse body shape features of individuals by extracting the lower body type elements related to clothes construction.

To maintain a balanced posture, each part of the body is organically interconnected. The pelvis plays an especially important role in maintaining overall balance. This study focused on the pelvis, investigating the organic relationships between pelvic tilt and components of lower body shape. Anterior pelvic tilt was found to result in a shape with a protruding abdomen and buttocks, while the posterior pelvic tilt was associated with o-shaped legs and flat buttocks. The organic relationship among lower body types is expected to help quickly identify the consumer's body shape.

Body dimension	Mean	S.D.	Body dimension	Mean	S.D.
Age (years)	65.63	4.13	Calf height (cm)	27.28	3.36
Height (cm)	152.25	4.90	Minimum leg height (cm)	5.23	0.73
Weight (kg)	56.87	7.19	Waist circumference (cm)	85.49	7.93
BMI (body mass index)	24.51	2.78	Waist circumference (omphalion) (cm)	89.97	7.45
Waist height (cm)	91.83	10.38	Hip circumference (cm)	92.10	5.11
Waist height (omphalion) (cm)	86.17	9.89	Total crotch length (cm)	70.03	4.54
Hip height (cm)	71.90	8.16	Thigh circumference (cm)	52.77	4.21
Crotch height (cm)	66.46	7.59	Knee circumference (cm)	33.69	2.66
Thigh height (cm)	76.27	8.90	Calf circumference (cm)	32.08	2.73
Knee height (cm)	38.67	4.42	Minimum leg circumference (cm)	18.70	1.99

Classification		Landmark	Classification		Landmark
A (Waist area)	A1	Anterior waist point	E (Hip area)	E1	Anterior hip point
A2	Posterior waist point	E2	Posterior hip point
A3	Lateral waist point_right	E3	Lateral hip point_right
A4	Lateral waist point_left	E4	Lateral hip point_left
A5	Posterior-curvature waist point_right	E5	Posterior-curvature hip point_right
A6	Posterior-curvature waist point_left	E6	Posterior-curvature hip point_left
B (Waist (omphalion) area)	B1	Anterior waist point (omphalion)	F (Crotch area)	F	Crotch point
B2	Posterior waist point (omphalion)	G (Thigh area)	G1	Anterior thigh point
B3	Lateral waist point (omphalion)_right	G2	Posterior thigh point
B4	Lateral waist point (omphalion)_left	G3	Lateral thigh point_right
C (Abdominal area)	C1	Anterior abdominal point	G4	Lateral thigh point_left
C2	Posterior abdominal point	H (Knee area)	H1	Anterior knee point-mid patella
C3	Lateral abdominal point_right	H2	Posterior knee point
C4	Lateral abdominal point_left	H3	Lateral knee point_right
D (Top hip area)	D1	Anterior top hip point	H4	Lateral knee point_left
D2	Posterior top hip point	I (Calf area)	I1	Anterior calf point
D3	Lateral top hip point_right	I2	Posterior calf point
D4	Lateral top hip point_left	I3	Lateral calf point_right
D5	Posterior-curvature top hip point_right	I4	Lateral calf point_left
D6	Posterior-curvature top hip point_left	J (Minimum leg area)	J1	Anterior minimum leg point
			J2	Posterior minimum leg point
			K	K1	Lateral malleolus
			K2	Sphyrion
				L1	Pternion

Lower body component	Group 1	Group 2	Group 3	Total
The pelvis typeGroup 1: anterior tiltGroup 2: straightGroup 3: posterior tilt	28(30.77%)	32(35.16%)	31(34.07%)	91(100%)
Agreement (%) = {(69 × 5 + 15 × 4 + 7 × 3)/ 91 × 5} × 100 = 93.63 (%)
The abdomen typeGroup 1: lower bulgeGroup 2: upper and lower bulgeGroup 3: overall bulge	19(20.88%)	14(15.38%)	58(63.74%)	91(100%)
Agreement (%) = {(64 × 5 + 15 × 4 + 12 × 3)/91 × 5} × 100 = 91.43 (%)
The buttocks typeGroup 1: protrudingGroup 2: mediumGroup 3: flat	26(28.57%)	30(32.97%)	35(38.46%)	91(100%)
Agreement (%) = {(71 × 5 + 12 × 4 + 8 × 3)/91 × 5} × 100 = 93.85 (%)
The lateral leg typeGroup 1: hyper-extendedGroup 2: straightGroup 3: bent	17(18.68%)	45(49.45%)	29(31.87%)	91(100%)
Agreement (%) = {(72 × 5 + 13 × 4 + 6 × 3)/91 × 5} × 100 = 94.51 (%)
The Frontal Leg typeGroup 1: o-shapeGroup 2: StraightGroup 3: x-shape	20(21.98%)	52(57.14%)	19(20.88%)	91(100%)
Agreement (%) = {(64 × 5 + 14 × 4 + 13 × 3)/91 × 5} × 100 = 91.21 (%)

	Pelvis type	Abdomen type	Buttocks type	Lateral leg type	Frontal leg type
RA2	–1.40	0.85	–	–	–0.78	–0.67	–	–	–	–
RA5	0.48	–0.62	–	–	–0.01	0.82	–	–	–	–
RC4	–0.44	–0.31	–	–	–	–	–	–	–	–
RD5	–	–	0.56	0.81	–	–	–	–	–	–
RE2	–	–	–	–	2.19	–0.07	–	–	–	–
RE5	–	–	1.36	–0.29	–	–	–	–	–	–
RE6	3.30	1.93	–	–	–	–	–	–	–	–
RG1	–0.67	–0.31	–	–	–	–	–	–	–	–
RH1	–	–	–	–	–	–	–	–	1.09	–1.22
RI4	–	–	–	–	–	–	–	–	2.82	1.92
RK1	–	–	–	–	–	–	–	–	–3.21	0.45
RK2	–	–	–	–	0.88	0.87	–	–	–	–
RL1	–	–	–	–	–	–	–	–	1.63	–0.81
θA3	0.57	0.32	–	–	–	–	–	–	–	–
θC3	–	–	–	–	–	–	–0.53	–0.21	–	–
θG1	–	–	–	–	–	–	–0.71	0.01	–	–
θG4	0.04	–0.88	–	–	–	–	–	–	0.48	0.03
θH1	–	–	–	–	–	–	3.16	–1.12	1.27	0.82
θH2	–	–	–	–	–	–	1.22	–1.01	–	–
θH3	–0.50	0.77	–	–	–	–	–	–	–	–
θI1	–	–	–	–	–	–	–2.58	2.92	–	–
θI2	–	–	–	–	–	–	1.09	–0.16	–	–
θK1	–	–	–	–	–	–	–0.82	–0.29	–	–
ρE1	–	–	–	–	0.06	–1.11	–	–	–	–
ρE2	–	–	–0.56	–0.26	–	–	–	–	–	–
ρE5	–	–	–	–	–	–	0.91	–0.47	–	–
ρG2	–	–	–	–	–	–	–0.71	0.00	–	–
ρJ1	–	–	–	–	–0.75	0.93	–	–	–	–
фB2	–	–	0.21	–0.72	0.50	0.33	–	–	–	–
фB4	0.93	0.28	–	–	–	–	0.31	0.68	–	–
фC1	–	–	–0.02	1.06	–	–	–	–	–	–
фE4	–1.07	–2.00	–	–	–	–	–	–	–	–
фE5	1.13	3.76	–	–	–	–	–	–	–	–
фG4	–	–	0.40	0.28	0.11	0.55	–	–	–	–
фI4	–0.21	–0.63	–	–	–	–	–	–	–	–
фJ1	–	–	–	–	0.76	1.24	–	–	–	–
фJ2	–	–	–	–	–	–	–	–	0.70	0.68
Eigenvalue	7.75	0.67	1.57	0.69	3.25	1.06	4.06	0.31	3.49	0.15
% of Variance	92.01	7.99	69.45	30.55	75.48	24.52	92.91	7.09	95.90	4.10
Canonical correlation	0.94	0.63	0.78	0.64	0.87	0.72	0.90	0.49	0.88	0.36

Classification	Pelvis type	Abdomen type	Buttocks type	Lateral leg type	Frontal leg type
Group 1	3.21	28	–2.39	19	2.14	26	–3.14	17	2.91	20
Group 2	0.59	32	0.33	14	0.61	30	–0.43	45	–0.14	52
Group 3	–3.51	31	0.70	58	–2.11	35	2.50	29	–2.68	19

Classification	Pelvis type	Abdomen type	Buttocks type	Lateral leg type	Frontal leg type
Pelvis type	1
Abdomen type	0.722**	1
Buttocks type	0.861**	0.633**	1
Lateral leg type	–0.204	–0.127	–0.055	1
Frontal leg type	–0.318**	–0.257*	–0.295**	0.205	1

Footnotes

Acknowledgment

This paper was reconstructed using some data from Dr Park’s (2009) doctoral dissertation.

Declaration of conflicting interests

The authors have no conflicts of interest to declare.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Sunmi Park

References

Chun

Lim

HS.

A study on the use of MTM CAD program for mass customization of men's suit. Res J Costume Culture 2003; 11: 647–656.

Oliver

Bickle

Shim

Profile of male made-to-measure customers: body characteristics and purchase selection. Cloth Text Res J 1993; 11: 59–62.

Burns LD, Mullet KK and Bryant NO. The business of fashion: Designing, manufacturing, and marketing. London: Bloomsbury Publishing, 2016.

Park

Nam

YJ.

A study on the basic slacks pattern by the automatic drafting for the order. J Korean Soc Cloth Text 2001; 25: 91–102.

Choi JY and Song HK. Investigation of production condition of ready-made and MTM men’s suits-focusing on brands of department store. Fashion Text Res J 2016; 18: 746–754.

Zhong

Three-dimensional technology for apparel mass customization: Part III: visualization of three-dimensional garments. J Text Inst 2003; 94: 92–102.

Apeagyei

Otieno

Usability of pattern customising technology in the achievement and testing of fit for mass customisation. J Fash Mark Manage 2007; 11: 349–365.

Senanayake

Little

TJ.

Mass customization: points and extent of apparel customization. J Fash Mark Manage 2010; 14: 282–299.

McKinney

Gill

Dorie

, et al. Body-to-pattern relationships in women’s trouser drafting methods: Implications for apparel mass customization. Cloth Text Res J 2017; 35: 16–32.

10.

Lang C, Xia S and Liu C. Style and fit customization: a web content mining approach to evaluate online mass customization experiences. J Fash Mark Manage 2021; 25: 224–241.

11.

Kim S. A study on the types and features of digital customizing in contemporary fashion. Art Des 2020; 23: 201–222.

12.

Woo S and Kwon S. A study on personalized product demand manufactured by smart factory. Manag Inform Syst Rev 2019; 38: 23–41.

13.

Yuan Y and Huh J. Customized CAD modeling and design of production process for one-person one-clothing mass production system. Electronics 2018; 7: 270–293.

14.

Population Reference Bureau. 2020 World Population Data Sheet. Population Reference Bureau 2020, https://www.prb.org/worldpopdata (2020, accessed 10 March 2021).

15.

DESA UN. World Population Prospects 2019. United Nations. Department of Economic and Social Affairs.

16.

Park

YR.

EU’s policy paradigm for active ageing towards super-aged society. J Contemp Eur Stud 2013; 31; 135–158.

17.

Scherbov

Sanderson

Andruchowitz

Aging Demographic Data Sheet 2020, http://pure.iiasa.ac.at/id/eprint/16279/1/AgingDemDataSheet2020.pdf (2020, accessed 10 March 2021).

18.

Jackson

O'Neal

GS.

Dress and appearance responses to perceptions of aging. Cloth Text Res J 1994; 12: 8–15.

19.

Salusso

Borkowski

Reich

, et al. An alternative approach to sizing apparel for women 55 and older. Cloth Text Res J 2006; 24: 96–111.

20.

Chae

JM.

Clothing purchase behavior of active senior-differences according to gender and shopping orientation types. Fash Text Res J 2020; 22: 25–37.

21.

Goldsberry

Shim

Reich

Women 55 years and older: Part I: current body measurements as contrasted to the PS 42-70 data. Cloth Text Res J 1996; 14: 108–120.

22.

Yeo

Kwon

YS.

A research on the actual condition of the elderly males and females' clothing life. J Korean Soc Cloth Text 2005; 29: 177–188.

23.

Kim

Choi

An analysis of body shapes in aged abdominal obese women for apparel pattern design. J Korean Soc Cloth Text 2006; 29: 177–178.

24.

Lee

Kim

A study on the clothing behaviors and wearing sensation of women's ready-to-wear. Korean J Hum Ecol 2007; 16: 123–135.

25.

Jeon K Lee

M. A study on life style and clothing purchasing behaviors of female consumers-focused on comparison with adolescence, middle age, and senescence. Res J Costume Culture 2008; 16: 444–460.

26.

Park

Glenn

Gordon

, et al. Future UN report 2040. 1st ed. Paju: Kyobobooks, 2014, p.340.

27.

Schofield

Ashdown

Hethorn

, et al. Improving pant fit for women 55 and older through an exploration of two pant shapes. Cloth Text Res J 2006; 24: 147–160.

28.

Fan

Yu W Hunter

Clothing appearance and fit: science and technology. Cambridge: Woodhead Publishing, 2004.

29.

Bougourd J. Sizing systems, fit models and target markets. In: ASHDOWN, S.P.(Ed.). Cambridge: Woodhead Publishing, 2007

30.

Nam

Choi

IS.

Classification and characteristics of the lower body type of elderly women. Res J Costume Culture 1999; 7: 154–164.

31.

Cha

SJ.

Classification of the lower body shape of the elderly women using 3D data-focused on 60's women. Korean Soc Basic Des Art 2017; 18: 389–402.

32.

Kim KA. Development of slacks patterns for lower body types of 60’s women using 3D body measurement data. Unpublished Doctoral Dissertation, Sungkyunkwan University, Seoul, 2019.

33.

Rasband

Liechty

Fabulous fit: speed fitting and alterations. 2nd ed. New York: Fairchild Books, 2005.

34.

Liechty

Rasband

Pottberg-Steineckert

Fitting and pattern alteration: a multi-method approach. 2nd ed. New York: Fairchild Books, 2010.

35.

Palmer P and Alto M. Pants for real people: fit and sew for anybody. Washington: Palmer/Pletsch Incorporated, 2012.

36.

Editors of Creative Publishing. The perfect fit: the classic guide to altering pattern. Beverly: Quarry Books, 2005.

37.

Kendall FP. Muskeln: funktionen und tests. Elsevier, Stuttgart: Urban & Fischer, 2009.

38.

Stoudt

HW.

The anthropometry of the elderly. Hum Factors 1981; 23: 29–37.

39.

Rosencranz

ML.

Clothing concepts: a social-psychological approach. London: Macmillan, 1972.

40.

Perry P (ed.). The Vogue sewing book of fitting, adjustments, and alterations. New York: Butterick Fashion Marketing Company, 1972.

41.

Lee

JY.

Analysis of deformed lateral body of elderly women (2): focused on classification of deformed lateral body type. Korean Soc Fash Des 2009; 9: 59–74.

42.

Jang

Lee

JR.

Analysis on the lower body shape of the varus-typed elderly women. Fash Text Res J 2017; 19: 569–578.

43.

Lee

Kim

Nam

, et al. Biotypology clothing. Seoul: Kyobobooks, 2002.

44.

Park

Kim

HK.

A study on the basic slacks pattern for middle-aged women. J Korean Home Econo Assoc 1997; 35: 79–94.

45.

Yoon

MK.

A study on the pattern design for mass customization according to the styles of women's 3D body scan data. Unpublished Doctoral Dissertation, Seoul National University, Seoul, 2008.

46.

Seong

Choi

Kim

, et al. Characteristics and classification of lower body of unmarried adult female aged twenties. J Korean Soc Cloth Text 1997; 21: 727–739.

47.

Song

Ashdown

SP.

Categorization of lower body shapes for adult females based on multiple view analysis. Text Res J 2011; 81: 914–931.

48.

Chung

Lee

. Somatotype classification and discrimination in the lower torso and legs of adult females. J Korean Soc Cloth Text 1998; 22: 241–249.

49.

Cho

Kim

HK.

Analysis of lower body on Korean young women and the effect of the lower body type on the shape of flare skirt (II). J Korean Home Econo Assoc 1992; 30: 1–13.

50.

Classification of lower frontal and lateral body shapes for the junior-high schoolgirls and analysis of size distribution according to the body types. Fash Text Res J 2003; 5: 173–180.

51.

Istook

Hwang

3D body scanning systems with application to the apparel industry. J Fash Mark Manage 2001; 5: 120–132.

52.

Azouz ZB, Shu C, Lepage R, et al. Extracting main modes of human body shape variation from 3D anthropometric data. In: fifth international conference on 3-D Digital imaging and modeling (ed Stephanie K), Ottawa, ON, Canada, 13-16 June 2005, pp.335–342. Los Alamitos, Calif.: IEEE Computer Society.

53.

Choi

Nam

YJ.

Classification of upper lateral body shapes for the apparel industry. Hum Factors Ergon Manuf Serv Ind 2010; 20: 378–390.

54.

Simmons

Istook

Devarajan

Female figure identification technique (FFIT) for apparel Part II: development of shape sorting software. J Text App Tech Manage 2004; 4: 1–15.

55.

Connell

Ulrich

Brannon

, et al. Body shape assessment scale: Instrument development for analyzing female figures. Cloth Text Res J 2006; 24: 80–95.

56.

Istook

CL.

Three-dimensional body scanning to improve fit. Advances in apparel production. Cambridge: Woodhead Publishing, 2008.

57.

Hong

Jung

Park

Comparison of three‐dimensional Korean male anthropometric data with modeling data generated by digital human models. Hum Factors Ergon Manuf Serv Ind 2014; 24: 671–684.

58.

Lee

Hong

Application of 3D human data for the engineering design of clothing. Fash Inform Tec 2004; 1: 28–37.

59.

Han

HS.

A study on the automatic setting of 3D body landmarks and measuring paths for apparel. Unpublished Doctoral Dissertation, Seoul National University, Seoul, 2007.

60.

Kim

Development of classifying method for upper lateral somatotype of men in their 20s by using space vector. Unpublished Doctoral Dissertation, Seoul National University, Seoul, 2009.

61.

Cho

Okada

Park

, et al. An interactive body model for individual pattern making. Int J Cloth Sci Tech 2005; 17: 91–99.

62.

Cho

CS.

Recognition and partition of human body using 3 dimensional vector data.

Collect Pap Hanshin 1999; 16: 455–474.

63.

Korea Research Institute of Standards and Science. The National Size Survey of Korea. 5th Size Korea Anthropometric Survey. Report, Korean Agency for Technology and Standards, Ministry of Commerce, Industry and Energy, Korea.

64.

ISO 20685-1:2018. 3-D scanning methodologies for internationally compatible anthropometric databases — Part 1: evaluation protocol for body dimensions extracted from 3-D body scans.

65.

ISO 8559-1: 2017. Size designation of clothes. Part 1: anthropometric definitions for body measurement.

66.

ISO 7250-1:2017. Basic human body measurements for technological design — Part 1: body measurement definitions and landmarks.

67.

Akram

Astrup

Atinmo

, et al. Obesity: preventing and managing the global epidemic. World Health Organization - Technical Report Series. 1 December 2000.

68.

Nam

YJ.

A study on classification of somatotype based on the lateral view of women's upper body. Unpublished Doctoral Dissertation, Seoul National University, Seoul, 1991.

69.

Lee

Nam

Comparison of somatotypes from various classification methods-between 18 and 24 years old Korean women. Fash Text Res J 2004; 6: 221–228.

70.

Kim

BA.

A Study of classification method for abdomen and upper lateral somatotype of Korea adult men. Unpublished Master's Thesis, Seoul National University, Seoul, 2015.

71.

Lee

JY.

An analysis of abdominal obesity and development abdominal type estimation model for the elderly women. Unpublished Doctoral Dissertation, Seoul National University, Seoul, 2017.

72.

Kim

Jeon

SAS discrimination and classification analysis. Seoul: Free Academy, 1997.

73.

Lee

Jung

Seo

Research methodology in clothing & textiles. 3rd ed. Seoul: Gyomunsa, 2019.

Lower body shape	Sub-type classification			Lower body shape	Sub-type classification
The pelvis type				The lateral leg type
The pelvis type	(a) Anterior tilt	(b) Straight	(c) Posterior tilt	The lateral leg type	(a) Hyper-extended	(b) Straight	(c) Bent
The abdomen type				The frontal leg type
The abdomen type	(a) Lower bulge	(b) Upper and lower bulge	(c) Overall bulge	The frontal leg type	(a) O-shape	(b) Straight	(c) X-shape
The buttocks type
The buttocks type	(a) Protruding	(b) Medium	(c) Flat

	Pelvis type		Abdomen type		Buttocks type		Lateral leg type		Frontal leg type
Items	Function1	Function2	Function1	Function2	Function1	Function2	Function1	Function2	Function1	Function2
RA2	–1.40	0.85	–	–	–0.78	–0.67	–	–	–	–
RA5	0.48	–0.62	–	–	–0.01	0.82	–	–	–	–
RC4	–0.44	–0.31	–	–	–	–	–	–	–	–
RD5	–	–	0.56	0.81	–	–	–	–	–	–
RE2	–	–	–	–	2.19	–0.07	–	–	–	–
RE5	–	–	1.36	–0.29	–	–	–	–	–	–
RE6	3.30	1.93	–	–	–	–	–	–	–	–
RG1	–0.67	–0.31	–	–	–	–	–	–	–	–
RH1	–	–	–	–	–	–	–	–	1.09	–1.22
RI4	–	–	–	–	–	–	–	–	2.82	1.92
RK1	–	–	–	–	–	–	–	–	–3.21	0.45
RK2	–	–	–	–	0.88	0.87	–	–	–	–
RL1	–	–	–	–	–	–	–	–	1.63	–0.81
θA3	0.57	0.32	–	–	–	–	–	–	–	–
θC3	–	–	–	–	–	–	–0.53	–0.21	–	–
θG1	–	–	–	–	–	–	–0.71	0.01	–	–
θG4	0.04	–0.88	–	–	–	–	–	–	0.48	0.03
θH1	–	–	–	–	–	–	3.16	–1.12	1.27	0.82
θH2	–	–	–	–	–	–	1.22	–1.01	–	–
θH3	–0.50	0.77	–	–	–	–	–	–	–	–
θI1	–	–	–	–	–	–	–2.58	2.92	–	–
θI2	–	–	–	–	–	–	1.09	–0.16	–	–
θK1	–	–	–	–	–	–	–0.82	–0.29	–	–
ρE1	–	–	–	–	0.06	–1.11	–	–	–	–
ρE2	–	–	–0.56	–0.26	–	–	–	–	–	–
ρE5	–	–	–	–	–	–	0.91	–0.47	–	–
ρG2	–	–	–	–	–	–	–0.71	0.00	–	–
ρJ1	–	–	–	–	–0.75	0.93	–	–	–	–
фB2	–	–	0.21	–0.72	0.50	0.33	–	–	–	–
фB4	0.93	0.28	–	–	–	–	0.31	0.68	–	–
фC1	–	–	–0.02	1.06	–	–	–	–	–	–
фE4	–1.07	–2.00	–	–	–	–	–	–	–	–
фE5	1.13	3.76	–	–	–	–	–	–	–	–
фG4	–	–	0.40	0.28	0.11	0.55	–	–	–	–
фI4	–0.21	–0.63	–	–	–	–	–	–	–	–
фJ1	–	–	–	–	0.76	1.24	–	–	–	–
фJ2	–	–	–	–	–	–	–	–	0.70	0.68
Eigenvalue	7.75	0.67	1.57	0.69	3.25	1.06	4.06	0.31	3.49	0.15
% of Variance	92.01	7.99	69.45	30.55	75.48	24.52	92.91	7.09	95.90	4.10
Canonical correlation	0.94	0.63	0.78	0.64	0.87	0.72	0.90	0.49	0.88	0.36

	Predicted group membership
Pelvis type	Group 1 Anterior tilt	Group 2 Straight	Group 3 Posterior tilt	Total
Group 1: anterior tilt	27 (96.43%)	1 (3.57%)	0 (0.00%)	28 (100.00%)
Group 2: straight	2 (6.25%)	29 (90.63%)	1 (3.12%)	32 (100.00%)
Group 3: posterior tilt	0 (0.00%)	1 (3.23%)	30 (96.77%)	31 (100.00%)
Percentage accuracy of prediction: 94.5%
	Predicted group membership
Abdomen type	Group 1 Lower bulge	Group 2 Upper and lower bulge	Group 3 Overall bulge	Total
	18 (94.74%)	0 (0.00%)	1 (5.26%)	19 (100.00%)
Group 2: upper and lower bulge	2 (14.29%)	12 (85.71%)	0 (0.00%)	14 (100.00%)
Group 3: overall bulge	3 (5.17%)	10 (17.24%)	45 (77.59%)	58 (100.00%)
Percentage accuracy of prediction: 82.4%
	Predicted group membership
Buttocks type	Group 1 Protruding	Group 2 Medium	Group 3 Flat	Total
Group 1: protruding	24 (92.31%)	1 (3.85%)	1 (3.85%)	26 (100.00%)
Group 2: medium	1 (3.33%)	29 (96.67%)	0 (0.00%)	30 (100.00%)
Group 3: flat	0 (0.00%)	2 (5.71%)	33 (94.29%)	35 (100.00%)
Percentage accuracy of prediction: 94.5%
	Predicted group membership
Lateral leg type	Group 1 Hyper-extended	Group 2 Straight	Group 3 Bent	Total
Group 1: hyper-extended	17 (100.00%)	0 (0.00%)	0 (0.00%)	17 (100.00%)
Group 2: straight	2 (4.44%)	39 (86.67%)	4 (8.89%)	45 (100.00%)
Group 3: bent	0 (0.00%)	0 (0.00%)	29 (100.00%)	29 (100.00%)
Percentage accuracy of prediction: 93.4%
	Predicted group membership
Frontal leg type	Group 1 O-shape	Group 2 Straight	Group 3 X-shape	Total
Group 1: o-shape	20 (100.00%)	0 (0.00%)	0 (0.00%)	20 (100.00%)
Group 2: straight	2 (3.85%)	45 (86.54%)	5 (9.62%)	52 (100.00%)
Group 3: x-shape	0 (0.00%)	1 (5.26%)	18 (94.74%)	19 (100.00%)

Classification	Pelvis type		Abdomen type		Buttocks type		Lateral leg type		Frontal leg type
Classification	Centroid	No. of group	Centroid	No. of group	Centroid	No. of group	Centroid	No. of group	Centroid	No. of group
Group 1	3.21	28	–2.39	19	2.14	26	–3.14	17	2.91	20
Group 2	0.59	32	0.33	14	0.61	30	–0.43	45	–0.14	52
Group 3	–3.51	31	0.70	58	–2.11	35	2.50	29	–2.68	19