Investigation of the radiotherapy-related changes in the eye lens using computed tomography entropy analysis

Abstract

OBJECTIVE:

The aim of this study was to apply texture analysis to investigate whether there was a change in the lens following radiotherapy.

PATIENTS AND METHOD:

Patients who received radiotherapy (RT) for head and neck cancer or brain tumor were enrolled. Computed tomography (CT) images taken in the last month before RT and the most recent images after RT were compared. Entropy values were calculated using lens attenuation values. The lens doses were obtained from the dose-volume histogram data.

RESULTS:

A total of 55 lenses were evaluated. The mean Hounsfield Unit value of the lenses was 66.14±12.16 before RT and 72.02±9.12 after RT (p = 0.007). The mean entropy value was 1.87±0.31 before RT and this reduced to 1.31±0.34 after RT (p < 0.001), respectively. As time increased, the difference in entropy also increased (p = 0.007). A correlation close to statistical significance was determined between the entropy difference and minimum, maximum and mean lens radiation dose (p = 0.052, p = 0.052, p = 0.063, respectively). The entropy difference was significantly reduced in the >4 Gy group (p = 0.046).

CONCLUSION:

Study results indicated that the entropy values in the lens were signifcantly changed after radiotherapy and the degree of the change associated with dose and time.

Keywords

Entropy lens radiotherapy texture analysis

1 Introduction

Therapeutic use of x-rays is the basis of radiotherapy and the lens is one of the most radio-sensitive tissues in the body [1]. Exposure to both ionised and non-ionised radiation causes clinical vision loss and may cause the development of cataract, which is defined as progressive opacity of the lens. It has previously been shown that for opacity to be determined, chronic exposure of 2 Gy or fractional exposure of 5 Gy was necessary. However, new studies have shown that the lens is affected even at lower doses. Re-analysis of most current epidemiological studies and data has shown that cataract could form in linear energy transfer including acute exposure at <1 Gy, and therefore, the 2012 International Commission on Radiological Protection (ICRP) recommended a threshold dose for occupational exposure of the eye lens of 50 mSv in a single year and that the 5-year average should reduce to 20mSv [2 –4]. From various reviews, it has been concluded that it is highly likely that initiation or development of various cataracts is related to exposure to ionizing radiation and although not definitive, it would be more appropriate to model radiation cataractogenesis risk on a lower threshold or non-threshold [5, 6].

The degree of retention (attenuation) of the X-ray of each structure in CT imaging determines the degree of whiteness in the image. Attenuation values of the x-rays passing through the organism are determined numerically. Each pixel corresponds to a number and these are organised as a scale. The numbers on the scale are known as Hounsfield Units (HU). To transform these numbers to an image, Gray-level intensity is used. The values of the pixels are coloured with a corresponding gray tone. Thus, a CT image formed of black, white and gray tones is obtained. In this way, the picture elements, that is the pixels, form the two-dimensional image, and each pixel has a value of co-ordinates that represent the gray-level intensity. However, as the human eye can differentiate few gray tones, many structures cannot be seen.

Advances in texture analysis techniques would provide more detailed extracted information, thus ensuring better quantification of differences in appearance which are not visible to the naked eye. Texture analysis is a vital component of computer-assisted diagnosis in medical image processing, as the classification of human tissues is difficult when based on Hounsfield units only. The technique of texture analysis evaluates the position and intensity of the signals on digital images, thus differentiating the features of pixels, and the associated grey-level intensity [7]. Textural information is initially extracted from the image, then this information is input to a decisional algorithm, which will perform the diagnostic task [8].

Texture analysis is the common name given to parameters which statistically identify characteristics such as fineness, uniformity, granulosity, regularity or irregularity of tissues from a visual perspective. Parametric heterogenity within the region of interest (ROI) is measured by entropy, as one of the texture analysis parameters [9, 10]. Entropy, as an information theory derived parameter, is suitable for the development of a system adaptive imaging method, which is based on the analysis of backscatter statistics. Entropy measures the uncertainty in a random variable. Increased entropy valus represent changes in backscatter signals, from regular to random to complex, corresponding to increased signal uncertainty [11].

Current studies have shown that the lens is affected even at very low doses suggesting that there could be changes in the lens structure after radiotherapy, even if they are not symptomatic. The aim of this study was to apply texture analysis, which provides more detailed information about the tissue than standard imaging methods, to investigate whether or not there was a change in the lens following radiotherapy.

2 Material and methods

2.1 Patient and radiation dose

A retrospective examination was made of patients undergoing radiotherapy in the brain or head and neck areas between February 2015 and December 2016. Those with a history of cataract or who did not have CT images before and after RT were excluded from the study. A total of 55 lenses of 29 patients were evaluated.

A retrospective comparison was made of the CT images taken in the last month before radiotherapy and the latest images after radiotherapy. A thermoplastic mask was applied to all patients. The planning tomographies were taken at 2.5 mm intervals to include the head and neck according to the diagnosis. The CT images were then transferred to the Eclipse treatment planning system (Varian Medical Systems, Version 13.0, Inc.) and contouring preparations were made. The lenses were contoured by a single radiation oncologist. Fractioned radiation therapy was applied with a 6MV photon beam. The lens dosages were obtained using the data of the dose-volume histogram.

2.2 Computed tomography examination protocol

Brain CT was applied using a 16-dedector-array CT device (Alexion, Toshiba Medical Systems, Nasu, Japan) with tube voltage of 100 to 120 kVp and tube current of 200 mAs. Slice thickness was 3 mm, reconstruction increment was 1.5 mm, and volume CT dose index was 53.10–68.50 mGy.

2.3 Image evaluation and analysis

Reconstructed axial images were uploaded to a 27-inch iMac computer (Apple Inc., Cupertino, California, United States). OsiriX V.4.9 imaging software (Pixmeo, Switzerland) was used to measure the lens HU using the ROI lining the lens contours point-to- point. Measurements were taken of both lenses. The ROI was placed so as to cover the whole lens without extending beyond the capsule and this was used in the measurement of X-ray attenuation and HU (Fig. 1). The HU values of the pixels within the ROI were calculated and transferred separately to the XLM file (eXtensible Markum Language)

Fig. 1

Axial CT scan through bilateral lens. Vertical arrow points to right lens, horizontal arrow point to globe and star points to ethmoidal air cells. The ROI (green circle) was placed covering the whole lens and without extending beyond the capsule.

Texture analysis was calculated using MATLAB version 2009b software (MATrix LABoratory, Mathworks Inc, Natick, USA) using the calculated lens HU values from the XLM file. To be able to identify the HU changes in more detail, texture parameters were derived from the histogram formed of different pixel values; these were mean, standard deviation (SD) of the histogram, and size % lower, size % upper, size % mean (% L % U and % M). The area remaining below –1 SD of the histogram indicates the L%, the area remaining above +1 SD, the U%, and the area between the M% (Fig. 2). In addition to the histogram values, the entropy values were obtained from the same data set. A record was made of the texture parameters before and after treatment. The Image Evaluation and Analysis procedures are summarised in Fig. 3.

Fig. 2

To be able to identify HU changes in more detail, a histogram formed of different pixel values was obtained. The area remaining below –1 SD of the histogram indicates the L%, the area remaining above +1 SD, the U%, and the area between the M%. SD:Standard deviation.

Fig. 3

The schematic figure demonstrates the fundamental concept of the study.

2.4 Statistical analysis

Analyses of the data were made using SPSS 22.0 software (IBM SPSS for Windows version 22, IBM Corporation, Armonk, USA). Quantitative data were stated as mean±standard deviation (SD). Conformity of the data to normal distribution was assessed with the Kolmogorov-Smirnov test. In the evaluation of independent paired groups, the Student’s t-test was used and for within-group comparisons, the Paired Samples t-test. Correlations were evaluated with Spearman’s Correlation test. Data were examined at a 95% confidence interval. A value of p < 0.05 was accepted as statistically significant.

3 Results

The study included a total of 29 patients, comprising 13 females and 16 males with a median age of 61 years (range, 38–79 years). Evaluation was made of a total of 55 lenses, as both lenses of 26 patients and the left lens only of 3 patients. Primary brain tumor was determined in 11 patients, metastatic brain tumor in 9 and head-neck cancer in 9.

The mean time from the beginning of RT to the evaluation after RT was median 3.8 months (range, 2.0–14.9 months). The radiotherapy doses were minumum 473 (30–1364) cGy, maximum 682 (51–2080) cGy, and mean 547 (44–1826) cGy. Study results also showed that the measured mean CT values of the lenses were 66.14±12.16 before RT and 72.02±9.12 after RT (p = 0.007), and the mean entropy values were 1.87±0.31 before RT and this reduced to 1.31±0.34 after RT (p < 0.001), respectively. The L% and U% values decreased, while the M% values increased (Table 1). The mean, entropy and histogram formed of different pixel values of a patient with glioblastoma are shown in Fig. 4.

Table 1
Changes occurring with radiotherapy for each of the texture analysis parameters. CI: Confidence interval, RT: Radiationtherapy, SD: Standard deviation

Parameters Before RT±SD After RT±SD Difference±SD 95% CI P value

Lower Upper

Mean 66,14±12,16 72,07±9,12 −5.931±15.541 −10,13 −1,72 0.007

Standart deviation 25,03±6,61 15,01±3,91 10.025±8.492 7,72 12,32 <0.001

Minimum 13,31±16,34 29,13±13,53 −15.818±21.999 −21,76 −9,87 <0.001

Maximum 102,78±15,93 93,93±9,14 8.855±18.392 3,88 13,82 0.001

Median 71,14±12,94 75,13±8,87 −3.990±15.989 −8,31 ,33 0.070

Entropy 1,86±,30 1,30±,33 0.555±0.487 ,42 ,68 <0.001

Size% L 19,47±2,96 16,09±2,49 3.382±4.310 2,21 4,54 <0.001

Size% U 14,90±4,04 13,32±3,46 1.584±5.241 ,16 3,00 0.029

Size% M 65,62±5,12 70,58±4,52 −4.966±6.905 −6,83 −3,09 <0.001

Parameters	Before RT±SD	After RT±SD	Difference±SD	95% CI	P value
Mean	66,14±12,16	72,07±9,12	−5.931±15.541	−10,13	−1,72	0.007
Standart deviation	25,03±6,61	15,01±3,91	10.025±8.492	7,72	12,32	<0.001
Minimum	13,31±16,34	29,13±13,53	−15.818±21.999	−21,76	−9,87	<0.001
Maximum	102,78±15,93	93,93±9,14	8.855±18.392	3,88	13,82	0.001
Median	71,14±12,94	75,13±8,87	−3.990±15.989	−8,31	,33	0.070
Entropy	1,86±,30	1,30±,33	0.555±0.487	,42	,68	<0.001
Size% L	19,47±2,96	16,09±2,49	3.382±4.310	2,21	4,54	<0.001
Size% U	14,90±4,04	13,32±3,46	1.584±5.241	,16	3,00	0.029
Size% M	65,62±5,12	70,58±4,52	−4.966±6.905	−6,83	−3,09	<0.001

Fig. 4

In a patient receiving radiotherapy because of a high-grade glial tumor in the left frontal region, the mean dose to the left lens of the patient was 14 Gy. When the values before RT (left) were compared with the values 10 months after RT (right), an increase was determined in the mean HU value (26.297, 30.970, respectively) and a decrease was observed in the entropy value (1.76, 1.05, respectively) calculated with the MATLAB program. Lower figures demonstrate that histogram analysis obtained from ROI before(left) and after(right) RT.

When the relationship between time and the entropy difference was examined, it was seen that as time increased, so the difference in entropy increased (r = 0.36, p = 0.007). At the end of the 6-month evaluation, the entropy difference in the first 6 months was 0.55±0.41 and a statistically significant difference was determined thereafter (0.79±0.31) (p = 0.03). The entropy difference was 0.54±0.40 in the first 9 months and 0.86±0.25 after 9 months and the difference was statistically significant (p = 0.002) (Fig. 5).

Fig. 5

Changes in mean and values after radiation therapy. Entropy difference according to the interval of nine months. The entropy difference was determined to have increased according to RT dose.

A correlation close to statistical significance was determined between the entropy difference and minimum, maximum and mean lens radiation dose (p = 0.052, p = 0.052, p = 0.063, respectively). The mean lens doses were classified as≤4 Gy and >4 Gy [12]. In the≤4 Gy group the entropy difference was 0.45±0.36 and in the >4 Gy group, 0.69±0.40. The entropy difference was more reduced in the >4 Gy group compared to the≤4 Gy group and the difference between the pre and post-RT entropy values was determined to have statistically significantly increased (p = 0.046) (Fig. 4).

4 Discussion

It is well known that exposure to ionised radiation causes opacification of the lens and the formation of cataract at later stages. The latent period and severity of the effects vary according to age and gender. Induction as a result of ionizing radiation has been shown to be related to dose, dose rate and fractionation [3]. Although there have been recent studies that have shown the importance of the stochastic effect, a systematic review of current eye epidemiology data has shown that the probable risk of cataract is most likely increased according to the level of exposure [13]. What is less clear is the value of the threshold for detectable opacities or vision-impairing cataracts and it is not currently possible to make a specific quantitative estimate of lens effect thresholds. In the current study, whether or not there were radiologically structural changes was examined with texture analysis in lenses exposed to different degrees of radiation. The study was applied in a relatively short period and took measurements from the same patients. Therefore, other factors that play a role in radiation cataractogenesis, such as age, genetic factors, gender and hormonal modulation, which could have had a confounding effect on the results, were eliminated [14]. As a result of the study, a correlation close to statistical significance was determined between the entropy difference and lens radiation dose. However, this reduction in entropy was statistically significantly different between the <4 Gy group and the > 4 Gy group. Therefore, because of the change in lens entropy, the lens dose must be kept as low as possible when planning radiotherapy. In addition, care must be taken that the lens dose is <4 Gy.

For many years, cataracts have been evaluated and graded using a slitlamp, which provides a magnified view. However, as this method is dependent on the experience of the practitioner, there can be problems with manual grading. There are also studies which have evaluated cataracts with tomography. For example, Chow et al. have suggested new methods of automatic detection, based on texture and intensity analysis, in cases of cortical and posterior subcapsular cataract. It is thought that a new method such as this could overcome the problems of existing methods and improve performance, particularly in respect of ROI detection, lens mask generation and opacity detection [15]. In the examination of patients with acute traumatic cataracts, unsuspected lens injury, or opacification of the anterior chamber, CT could be useful [16].

As the result of studies of a system of computerised analysis of digital images of posterior capsule opacification developed by Aslam et al, the system was reported to be evidence-based, objective and clinically useful. Substantial evidence has been provided of its validity and reliability [17]. As in the above-mentioned studies, the current study was based on CT images. However, the previous studies were conducted on the presence of cataracts, unlike the current study which examined the relationship between radiation dose-texture alterations, and this is therefore, the differentiating feature of this study. However, all these studies with digital imaging can be useful in respect of prediction and diagnosis of cataract development.

To date there has not been full clarification of the potential mechanisms behind radiation cataractogenesis. However, one of the mechanisms with a key role in the development of radiation cataractogenesis appears to be genomic damage of the lens epithelial cells (LECs) [13]. Reasons for lens opacification include excessive LEC proliferation, incomplete LEC differentiation to lens fiber cells, meridional row disorganization, oxidative stress and various post-translational modifications of crystallins [14]. It is not known which processes in cataractogenesis and lens biology correspond to the changes in the lens detected with texture analysis on the tomography images in this study. These changes, which develop in the early period may remain stable in the long-term, may disappear or may result in cataract, which impairs vision. In the follow-up of this retrospective study, eye examinations were not performed on the patients as there were no complaints related to vision. Therefore, there is a need for experimental or prospective studies on this subject.

The treatment for cataracts formed as a result of radiation is the removal of the lens and replacement with an artificial lens. However, in recent years there have been studies with promising results for the prevention of the development of cataracts. Therefore, the early identification of individuals at high risk of cataract development is important in respect of the radiation dose received by the lens.

A successful technique for lens regeneration, which is dependent on endogenous stem cells has been demonstrated by Lin et al. [18]. The reversal of protein aggregation in cataracts has been shown to be achieved with Lanosterol in experimental animal models [19]. Similarly, Makley et al., demonstrated that pharmacological chaperoning of a-crystallin partially restores transparency in cataract models [20]. It is known that as a result of extensive acetylation of crystallins in the lens, there is a partial structural alteration and enhancement of stability, in addition to improved α-crystallin chaperone-like activity [21]. Another study has also reported that a Bowman-Birk inhibitor and antioxidant formulation could have protective benefits for astronauts against space radiation-induced cataracts during or after long-term manned space missions [22]. When the efficacy of these agents has been proven on humans, they could be used in the near future for the prevention of radiation cataracts.

From the results of this study, some changes were observed associated with dose and time-dependent radiation in the lens tissue. As time passed, so the entropy difference in the lens increased. This difference could predict symptomatic cataract. Following radiotherapy, patients are already routinely followed up radiologically at regular intervals. The integration of texture analysis with conventional imaging techniques could enable more detailed information to be obtained about the response to radiotherapy of both the tumor and normal tissue. To the best of our knowledge, this is the first study to have studied radiation-related lens alterations. With proof of the reliability of this method in future studies, the development of cataracts associated with radiation could be determined in the early stage with routine imaging, and thus prevention would be possible with protective agents.

In conclusion, in this study, the lenses of cancer patients exposed to ionised radiation were examined with texture analysis, which is a new method in radiological imaging. Following radiation, it was observed that the Hounsfield Unit values of the lenses increased and entropy values decreased. These alterations were seen to be directly proptional to time and dose. With the advent of the clinical use of agents preventing the formation of cataracts and texture analysis, it may be possible to have early detection of cataracts which impair vision and thus, preventative steps could be taken.

Conflict of interest

No conflict of interest was declared by the authors.

References

Poppe

, Experimental investigations on cataract formation following whole-body roentgen irradiation, Acta radiologica 47(2) (1957), 138–148.

Bouffler

, Ainsbury

, Gilvin

and Harrison

, Radiation-induced cataracts: The Health Protection Agency’s response to the ICRP statement on tissue reactions and recommendation on the dose limit for the eye lens, Journal of Radiological Protection 32(4) (2012), 479.

Ainsbury

, Bouffler

, Dörr

et al. Radiation cataractogenesis: A review of recent studies, Radiation Research 172(1) (2009), 1–9.

Clement

C.H.

, ICPR statement on tissue reactions and early and late effects of radiation in normal tissues and organs threshold doses for tissue reactions in a radiation protection context. ICRP Publication 118 (2012), 1–2.

Electric Power Research Institute. Epidemiology and mechanistic effects of radiation on the lens of the eye: review and scientific appraisal of the literature. Palo Alto, CA: EPRI; 2014. Publication Number 3002003162.

Shore

R.E.

, Radiation and cataract risk: Impact of recent epidemiologic studies on ICRP judgments, Mutation Research/Reviews in Mutation Research 770 (2016), 231–237.

Castellano

, Bonilha

, Li

and Cendes

, Texture analysis of medical images, Clinical Radiology 59(12) (2004), 1061–1069.

Tourassi

G.D.

, Journey toward computer-aided diagnosis: Role of image texture analysis, Radiology 213(2) (1999), 317–320.

Ganeshan

, Panayiotou

, Burnand

, Dizdarevic

and Miles

, Tumour heterogeneity in non-small cell lung carcinoma assessed by CT texture analysis: A potential marker of survival, European Radiology 22(4) (2012), 796–802.

10.

Ganeshan

, Skogen

, Pressney

, Coutroubis

and Miles

, Tumour heterogeneity in oesophageal cancer assessed by CT texture analysis: Preliminary evidence of an association with tumour metabolism, stage, and survival, Clinical Radiology 67(2) (2012), 157–164.

11.

Zhou

, Huang

C.-C.

, Shung

K.K.

et al. Entropic imaging of cataract lens: An in vitro study, PloS One 9(4) (2014), e96195.

12.

Gunderson

L.L.

, Clinical radiation oncology, 2015: Elsevier Health Sciences.

13.

Dauer

L.T.

, Hamada

and Blakely

E.A.

, National Council on Radiation Protection and Measurements Commentary Number 26: Impact of Revised Guidance on Radiation Protection for the Lens of the Eye, Journal of the American College of Radiology 14(7) (2017), 980–982.

14.

Gupta

V.B.

, Rajagopala

and Ravishankar

, Etiopathogenesis of cataract: An appraisal, Indian Journal of Ophthalmology 62(2) (2014), 103.

15.

Chow

Y.C.

, Gao

, Li

, Lim

J.H.

, Sun

and Wong.

T.Y.

, Automatic detection of cortical and PSC cataracts using texture and intensity analysis on retro-illumination lens images. in Engineering in Medicine and Biology Society, EMBC, 2011 Annual International Conference of the IEEE. 2011.

16.

Boorstein

, Titelbaum

, Patel

, Wong

and Grossman

, CT diagnosis of unsuspected traumatic cataracts in patients with complicated eye injuries: Significance of attenuation value of the lens, American Journal of Roentgenology 164(1) (1995), 181–184.

17.

Aslam

T.M.

, Patton

and Graham

, A freely accessible, evidence based, objective system of analysis of posterior capsular opacification; evidence for its validity and reliability, BMC Ophthalmol 5 (2005), 9.

18.

Lin

, Ouyang

, Zhu

et al. Lens regeneration using endogenous stem cells with gain of visual function, Nature 531(7594) (2016), 323–328.

19.

Zhao

, Chen

X.-J.

, Zhu

et al. Lanosterol reverses protein aggregation in cataracts, Nature 523(7562) (2015), 607–611.

20.

Makley

L.N.

, McMenimen

K.A.

, DeVree

B.T.

et al. Pharmacological chaperone for α-crystallin partially restores transparency in cataract models, Science 350(6261) (2015), 674–677.

21.

Moafian

, Khoshaman

, Oryan

, Kurganov

B.I.

and Yousefi

, Protective Effects of Acetylation on the Pathological Reactions of the Lens Crystallins with Homocysteine Thiolactone, PloS one 11(10) (2016), e0164139.

22.

Davis

J.G.

, Wan

X.S.

, Ware

J.H.

and Kennedy

A.R.

, Dietary supplements reduce the cataractogenic potential of proton and HZE-particle radiation in mice, Radiation Research 173(3) (2010), 353–361.

Parameters	Before RT±SD	After RT±SD	Difference±SD	95% CI		P value
				Lower	Upper
Mean	66,14±12,16	72,07±9,12	−5.931±15.541	−10,13	−1,72	0.007
Standart deviation	25,03±6,61	15,01±3,91	10.025±8.492	7,72	12,32	<0.001
Minimum	13,31±16,34	29,13±13,53	−15.818±21.999	−21,76	−9,87	<0.001
Maximum	102,78±15,93	93,93±9,14	8.855±18.392	3,88	13,82	0.001
Median	71,14±12,94	75,13±8,87	−3.990±15.989	−8,31	,33	0.070
Entropy	1,86±,30	1,30±,33	0.555±0.487	,42	,68	<0.001
Size% L	19,47±2,96	16,09±2,49	3.382±4.310	2,21	4,54	<0.001
Size% U	14,90±4,04	13,32±3,46	1.584±5.241	,16	3,00	0.029
Size% M	65,62±5,12	70,58±4,52	−4.966±6.905	−6,83	−3,09	<0.001