Impact of image quality,radiologists,lung segments,and Gunnar eyewear on detectability of lung nodules in chest CT

Abstract

Background

Despite the increasingly higher spatial and contrast resolution of CT, nodular lesions are prone to be missed on chest CT. Tinted lenses increase visual acuity and contrast sensitivity by filtering short wavelength light of solar and artificial origin.

Purpose

To test the impact of Gunnar eyewear, image quality (standard versus low dose CT) and nodule location on detectability of lung nodules in CT and to compare their individual influence.

Material and Methods

A pre-existing database of CT images of patients with lung nodules >5 mm, scanned with standard does image quality (150 ref mAs/120 kVp) and lower dose/quality (40 ref mAs/120 kVp), was used. Five radiologists read 60 chest CTs twice: once with Gunnar glasses and once without glasses with a 1 month break between. At both read-outs the cases were shown at lower dose or standard dose level to quantify the influence of both variables (eyewear vs. image quality) on nodule sensitivity.

Results

The sensitivity of CT for lung nodules increased significantly using Gunnar eyewear for two readers and insignificantly for two other readers. Over all, the mean sensitivity of all radiologist raised significantly from 50% to 53%, using the glasses (P value = 0.034). In contrast, sensitivity for lung nodules was not significantly affected by lowering the image quality from 150 to 40 ref mAs. The average sensitivity was 52% at low dose level, that was even 0.7% higher than at standard dose level (P value = 0.40). The strongest impact on sensitivity had the factors readers and nodule location (lung segments).

Conclusion

Sensitivity for lung nodules was significantly enhanced by Gunnar eyewear (+3%), while lower image quality (40 ref mAs) had no impact on nodule sensitivity. Not using the glasses had a bigger impact on sensitivity than lowering the image quality.

Keywords

Lung nodule image quality eyewear low dose CT inter-observer agreement

Computed tomography (CT) is the imaging modality of choice for the detection of nodular lung pathologies like lung cancer, metastases, and infectious diseases. CT screening has increased the detection rate of small nodules, including those of early peripheral lung cancer (1–4). Despite the increasingly higher spatial and contrast resolution of CT, nodular lesions are prone to be missed on chest CT (3, 5). Sensitivity of chest CT for lung nodules is in the range of 61–98% dependent on nodule characteristics and image acquisition parameters (5–7), this means that 2–39% of lung nodules are missed.

In order to increase the detectability of pulmonary nodules, different means either at the level of the image display, like the use of thin slab Maximum Intensity Projection (MIP) allowing better differentiation between nodular and non-nodular pulmonary lesions, or the use of computer-aided detection (CAD) algorithms, which detect nodular candidates as a second reader reducing the proportion of initially missed pulmonary lesions (8, 9). Even a simple double reading paradigm using a second human reader showed significant increase of detection sensitivity (10–12). The superiority of a diagnostic monitor calibrated according to the American Association of Physicists in Medicine (AAPM TG18) has been demonstrated in the framework of diagnostic radiology (13, 14). Besides keeping the ambient light conditions ideal while reading CT cases, no particular measures have been proposed to further optimize the human's optical system. Short wavelength light with high frequency (blue light) has been shown to cause visual discomfort, hazy vision, reduced contrast, and prolonged adaptation times (15). Tinted lenses have proven to increase visual acuity and contrast sensitivity by filtering short wavelength light of solar and artificial origin (16–19). This motivated us to test a specific optical eyewear system optimized for digital screen reading and image interpretation (20), these glasses can be customized for individual vision correction. The eyewear's lens technology claims to balance and optimize lighting environment for the needs of the human eye by specific coatings allowing for a shift towards a warmer color spectrum at increased contrast and a magnifying effect (Fig. 1). Additionally, the anti-reflective portion cuts glare and distracting extraneous light.
Fig. 1
Gunnar eyewear effect: the lens technology allows for a shift towards the warmer part of the color spectrum with a magnifying effect (arrows)

In order to weigh the value of an optimized optical system, which is specifically tailored for computer screen reading, the purpose of this study was to quantify its effect on radiologists' ability for detection of pulmonary nodules compared to non-assisted computer monitor reading sessions. Furthermore, the eyewear effect was compared to other variables influencing detectability of lung nodules like image quality and nodule location.

Material and Methods

Patients

A pre-existing database of CT images of patients with lung nodules >5 mm, scanned with standard dose image quality (150 reference mAs [ref mAs]/120 kVp) and lower dose/quality (40 ref mAs/120kVp), was used. Forty patients were retrospectively enrolled from our database, existing from a former HIPPA compliant IRB approved prospective study in 2010. Informed consent was waived due to the retrospective nature of data analysis. There were 22 men and 18 women with a median age of 55 years (range, 31–78 years) and a median body weight of 74 kg (range, 47–111 kg).

Image acquisition

All database patients were scanned in the supine position from the lung apex to the base of the chest at withheld full inspiration. No intravenous contrast was applied. Image data were obtained with a 64-row multidetector CT scanner (Somatom Sensation, Siemens, Erlangen, Germany) with a detector configuration of 32 × 0.6 mm, and flying focal spot technology, at a pitch of 2.0. Tube voltage was set to 120 kVp. Automated tube current modulation (Care Dose®, Siemens, Erlangen, Germany) with ref mAs was used to individualize radiation exposure to patient size. Instead of the routine 300 ref mAs tube charge the patients underwent a first scan with 150 ref mAs followed by a second ultra low dose scan with ref 40 mAs. Based on the international guidelines suggesting that incidentally discovered lung nodules measuring <5 mm are rather insignificant with an estimated malignancy rate of 1% (21, 22), only patients with nodules equal or larger than 5 mm were selected, resulting in 40 patients with 152 nodules. The 40 patients were mixed with 20 normal patients showing no nodules.

Image analysis

A consensus panel of two thoracic radiologists not involved in the observer study, with 9 and 12 years of experience in the interpretation of chest CT, respectively, established the standard of reference by identifying individual pathologic CT findings for each of the 18 lung segments (23). Because lung nodules may appear simultaneously in multiple lung segments, images were analyzed per lung segment and per patient. A total of 1080 lung segments were analyzed. To avoid segment classification bias the segments were altered to match prominent anatomical landmarks in the mediastinum and the lung (Fig. 2). It seemed important to quantify the effect of an optimized eyewear against the effect of image quality. Therefore, both sets of images at standard (150 ref mAs) and lower dose levels (40 ref mAs) were tested. The read-out arrangement of the cases was randomized. Every radiologist had to read the cases twice: once with and once without the specific eyewear. A break between the read-outs of 1 month to avoid recognition bias was applied. Three readers started the first read-out block with and two without the glasses. At both read-outs, cases of 40 or 150 ref mAs were shown alternatingly. There were individual randomized read-out orders of patients for each radiologist. Radiologists had to evaluate all images of a patient, find and localize the nodules according to the modified 18 anatomical lung segments. Only solid lung nodules were considered (HU > = –300), lobulated, spiculated mixed, or cavernous, but no ground-glass nodules (HU < −300). Only nodules ≥ 5 mm had to be included. The cases were presented on a PACS workstation (Picture Archiving and Communication System R11.4.1, 2009; Philips, Best, The Netherlands; Sectra, Linköping, Sweden). Ambient light conditions, which were similar to our clinical reading conditions, as well as the display window settings (window center –600 HU and width 1500 HU) were kept constant for all reading sessions and readers in order to minimize the variables and the potential bias.
Fig. 2
Modified lung segments. (a) Virtual axial plane through the most cranial point of aortic arch; (b) virtual axial plane through the most cranial point of the heart (mostly left atrium); (c) virtual axial plane through the middle of the heart (bisects the heart, cranio-caudal diameter); (d) virtual coronal plane through the ventral spine; (e) virtual coronal plane through the middle of the upper lung (halving the anterior-posterior diameter of the lungs); (f) virtual sagittal plane through the middle of the right or left lung (halving the width of each lung); (g) 45°angle plane (double oblique) between sagittal plane just right of the heart and the coronal plane right behind the heart; (h) oblique fissure; (i) horizontal fissure. Right lung segments (n = 10) and borders: 1 apical: caudal border = A; 2 posterior: cranial border = A, anterior border = E, posterior = H; 3 anterior: cranial = A, posterior = E, caudal = I; 4 lateral middle lobe: posterio-caudal = H, antero-medial = G, cranial = I; 5 medial middle lobe: postero-lateral = G, cranial = I; 6 apical lower lobe: anterior = H, caudal = B; 7 medio-basal lower lobe: lateral = F, posteror = D, cranial = B; 8 antero-lateral-basal lower lobe: medial = F, posterior = D, cranial = B; 9 latero-basal lower lobe: anterior = D, medial = F, cranial = B; 10 postero-basal lower lobe,: anterior = D, lateral = F, cranial = B. Left lung segments (n = 8) and borders: 1/2 apico-posterior: is the combined segment 1 and 2 of the right lung; 3 anterior: cranial border = A, posterior = E, caudal = B; 4 upper lingula: cranial = B, caudal = C, posterior = H; 5 lower lingula: cranial = C, posterior = H; 6 apical lower lobe: caudal = B, anterior = H; 7/8 antero-medial lower lobe: combined segment 7 and 8 of right lung; 9 latero-basal lower lobe: anterior = D, medial = F, cranial = B; 10 postero-basal lower lobe: anterior = D, lateral = F, cranial = B

Statistical analysis

Paired comparisons of the sensitivity for lung nodules of optical optimized and non-optimized condition using McNemar's test were carried out. The results of all five readers were pooled into a single contingency table by adding frequencies in each cell across readers. The null hypothesis of equality of sensitivity for all pairs was then tested by applying McNemar's test (24) to the pooled contingency table.

The same analysis was performed on a patient basis: the consensus panel classified a patient as positive if the patient demonstrated one or more nodule(s). Readers result was considered correct when at least one nodule was found in a positive patient. Furthermore, the nodule detectability of each lung segment was calculated for normal dose CT with the use of Gunnar eyewear.

In order to measure differences in the effect of eyewear across readers, lung segments, and tube current, we fit a logistic regression model (25) for the probability of detection p_ijkl, for i = 1,…,5 readers, j = 1,…,18 lung segments, k = 1,2 eyewear on/off, l = 1,2 if current time is 150/40 mAs. The factors in this model are readers, lung segments, eyewear (yes or no), and tube current-time (40 or 150 ref mAs) as explanatory factors as well as all possible interactions between these factors. The model was fit using maximum likelihood assuming a binomial model for detections in the R software package. The optimal model was identified using a stepwise selection procedure and non-significant interactions were dropped. The final model looked like:

α, β, γ, and δ were the standardized regression coefficients and µ the intercept: α stood for the reader effect, β for the lung segments, γ for the Gunnar effect, and δ for tube current effect. Double letters indicated the effect of the interaction between two factors: for instance, αγ represented the interaction between reader and eyewear.

Inter-rater agreement with and without eyewear was calculated using Fleiss's κ statistics (26, 27). Classification into segment with nodule or without nodule among the individual readers and for all readers combined was analyzed. Value of κ strength of agreement: <0.20, poor; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; 0.81–1.00, very good agreement (28).

Results

Optimized optical system

Four out of five radiologists found more lung nodules with the eyewear than without. The sensitivity of CT for lung nodules increased significantly using optimized eyewear for two readers (Table 1): for reader 1 from 43% without eyewear to 53% with eyewear (P = 0.0036) and from 40% to 47% (P = 0.025) for reader 2, respectively. McNemar test indicated an increase of sensitivity for two radiologists wearing the glasses: sensitivity increased from 60% to 63% (P = 0.11) and from 60% to 62% (P = 0.34), respectively. Only the sensitivity of a single radiologist dropped from 48% to 37% wearing the glasses (Table 1). Over all, the mean sensitivity of all radiologists raised significantly from 50% to 53%, using the glasses (P value = 0.034, Table 1).
Table 1
Sensitivities for all five readers with and without Gunnar eyewear, at 150 and 40 ref mAs

Tube current time

Reader + Gunnareyewear − Gunnareyewear 150mAs 40mAs

1 0.532* 0.427* 0.468^† 0.484^†

2 0.472* 0.395* 0.463^‡ 0.403^‡

3 0.374^† 0.476^† 0.390* 0.460*

4 0.621^‡ 0.605^‡ 0.640^† 0.613^†

5 0.629^‡ 0.597^‡ 0.602^† 0.640^†

Ø reader 0.526* 0.500* 0.513^† 0.520^†

P < 0.05

^† P > 0.32

^‡ P = 0.05–0.32

mAs, milliampere-second

Per patient evaluation

Reader 1 increased the sensitivity for patients with lung nodules from 63% to 77% wearing the glasses. He detected four patients more with the glasses that would have been rated negative without the glasses. Also, the sensitivity of reader 2 rose from 60% to 69%, meaning that he would have missed three patients without the glasses. Readers 4 and 5 showed little to no change in patient detectability (approximately 90%), while the sensitivity of reader 3 dropped from 77% to 57% wearing the glasses (missed four patients with Gunnar eyewear). None of these results were significant (P between 0.063 to 1).

Image quality (ref mAs)

In contrast, sensitivity for lung nodules was not significantly affected by lowering the image quality from 150 to 40 ref mAs (Table 1). The average sensitivity was 52% at low dose and even 0.7% higher than at standard dose level (P value = 0.40). One reader found significantly more nodules at lower dose than at standard dose levels: sensitivity at lower dose level was 46% and at standard dose level 39% (P = 0.048). For one radiologist the McNemar test was indicative (not significant) for a better sensitivity at standard dose than at lower dose level, sensitivity dropped from 46% at 150 ref mAs to 40% at 40 ref mAs (P = 0.071). The remaining three readers showed insignificant change of sensitivity when lowering the dose level (P > 0.4, Table 1).

Location (lung segments)

More nodules were detected in the right lung on normal dose CT with the use of Gunnar eyewear: sensitivity for lung nodules was 55.2% in the right and 42.4% in the left lung, respectively. The difference was 12.8% with a P value of 0.12 (chi-square test). The lobes adjacent to the heart (left and right segments 4 and 5 together) demonstrated lower sensitivities of 37.5%, compared to the upper segments 1, 2, and 3 where the sensitivity rose to 54.8% (Table 2). The difference of 17.3% was not significant, but indicative (P = 0.19).
Table 2
Sensitivity (%) for lung nodules by anatomical segments

Rightlung Leftlung

R1 71 L1/2 44

R2 56

R3 73 L3 42

R4 49 L4 56

R5 38 L5 7

R6 72 L6 52

R7 47 L7/8 42

R8 59

R9 49 L9 49

R10 50 L10 47

Standard CT dose with Gunnar eyewear

Analysis of sensitivity

We found that both reader and location were highly significant factors, i.e. sensitivity varied significantly across these factors (Table 3). Although, the overall effect of Gunnar eyewear and current was not very impressive, we found that the interaction between these two was highly significant, as was the interaction effect of optimized eyewear with readers, meaning the better readers scored higher sensitivities wearing the glasses and vice versa* (Table 1).
Table 3
Analysis of deviance table for optimal logistic regression model for sensitivity

Df Deviance Residual Df Residual deviance P value (chi-square test)

Null 1237 1715.6

Readers 4 36.95 1233 1678.65 <0.0001

Slices 17 95.81 1216 1582.85 <0.0001

Gunnar eyewear 1 0.48 1215 1582.36 0.49

Current 1 0.1 1214 1582.26 0.75

Readers: Gunnar eyewear 4 6.19 1210 1576.07 0.19

Readers: Current 4 11.27 1206 1564.8 0.02

Gunnar eyewear: current 1 10.16 1205 1554.65 0.002

Readers: Gunnar Eyewear: current 4 65.47 1201 1489.18 <0.0001

Df, degree of freedom

Inter-rater agreement with and without eyewear

For all readers combined inter-rater agreement was moderate for radiologists performing without glasses with a weighted Kappa value of 0.589 ± 0.014. Agreement increased to a Kappa value of 0.614 ± 0.013 when using the eyewear.

Discussion

Although chest CT is superior in demonstrating lung nodules compared to chest radiography (28–30), there are some limitations concerning the CT sensitivity. Radiologists’ sensitivity for lung nodules detection will never reach 100% (31–37), since sensitivity is dependent on technical CT factors, lung nodule features, and perception errors. In our study we examined one technical factor, image quality (image noise dependent on tube current time), against one perception error, the human eye. Radiologists get puzzled by the increasing amount of images and the anatomical noise in the lung, against which the nodules need to be differentiated. Therefore every attempt to relax the eye could have a positive effect on sensitivity. Nodule size also affects sensitivity: nodules with a diameter >10 mm reach higher sensitivities of 80% (38), while small nodules present significantly lower sensitivities (38–41). Studies with nodule sizes, that were comparable with our mean nodule size of 8 mm showed sensitivities of 38–70% (34–37). Our results show a comparable sensitivity of 50%, without vision enhancement. Radiologist using enhancing optical eyewear system could significantly increase their sensitivity to 53%. Only one reader demonstrated a lower sensitivity using the glasses, than others, probably because of the focusing and magnifying effect of the special lenses in combination with a slightly far-sighted radiologist (uncorrected glasses were used). Inter-rater agreement changed from moderate without glasses to good with glasses, which is important to radiologists to allow for better radiological comparison. A per patient evaluation confirmed the better detectability of lung nodules with Gunnar eyewear for two readers, while two readers were not affected by the glasses and the far-sighted reader even detected fewer patients with the glasses. Having a low number of positive patients compared to the amount of nodules the impact of Gunnar eyewear on nodule detection on a patient basis was not significant.

On the other hand sensitivity was not affected by image quality comparing two acquisition dose levels of 150 and ref 40 mAs. Multiple studies have demonstrated the feasibility of lowering the radiation dose in thoracic CT without substantial loss of diagnostic information despite increased image noise (42–44). Image quality of low dose chest CT has been shown to be satisfactory for mediastinal and lung parenchyma in patients with cancer (140 kV, 60 mAs) (42), and for the evaluation of bronchiectasis in patients with cystic fibrosis, (120 kV, 70 mAs) (43). Tack et al. (44) reported that tube current can be reduced to 10 mAs in the detection of pulmonary embolism. Bankier et al. reported that a tube current of 20 mAs is sufficient for visual quantification of air trapping at expiratory chest CT (45). Our results demonstrated that a higher image noise level at 40 ref mAs was not affecting the diagnostic information, required for nodule detection. These results are consistent with CT-dose studies which defined acceptable low dose threshold tube current levels for nodule sensitivities comparable to standard dose CT (100–250 mAs): 30 mAs was found by Hetmaniak et al. (46), 43 mAs by Weng et al. (47), 5 mAs by Gergely et al. (48), and 10 mAs by Das et al. (49). However, a general accepted threshold level of tube current time for the lung nodule detection is still missing and needs to be further investigated. As mentioned, a lower tube current time than 40 mAs seems to be possible, especially regarding nodule follow-up examinations. Ko et al. (5) published a difference in sensitivity for central (80%) compared to peripheral nodule location (61%); we also found a location dependent sensitivity: lower sensitivities for left segments and segments abutting the heart. Due to motion artifacts of the heart, both segment 5 scored the lowest sensitivity. Also, in the logistic regression model location and readers were the strongest factors influencing the sensitivity. In this model the use of optimized eyewear systems did not show a significant influence on sensitivity, but isolated McNemar test demonstrated a small but significant overall increase of sensitivity of 3%, wearing the glasses. An interaction of glasses and readers could be found, what this suggests is that the effect of eyewear on sensitivity varies significantly from reader to reader: the four readers with higher overall sensitivity scored higher wearing the glasses. This is probably influenced by reader 3, with the lowest overall sensitivity, who scored lower using the eyewear system. The positive interaction of optimized eyewear and tube current indicates that the glasses worked better at higher tube currents and worse at lower tube currents.

There are several limitations of our study. Only two dose levels (150 and 40 mAs) were examined to compare the influence of image quality. To examine the impact of image noise on detectability of lung nodules a logistic regression test with more than two dose levels, especially with lower levels towards 5 mAs would give a more precise conclusion on this topic.

No double-blind testing was possible using yellow Gunnar glasses. The awareness of wearing glasses that may enhance the vision could have affected the readers.

The optimized eyewear seemed to work for individual radiologists but it remains unclear whether the magnifying effect or the warmer light spectrum or both are influencing the reader. Further investigations will be needed.

In conclusion, using specific eyewear systems increases not only significantly the radiologists’ sensitivity for lung nodules, but also the inter-rater agreement, however, such an eyewear effect works best on good radiologists reading high quality images.

Conflict of interest: None.

			Tube current time
1	0.532*	0.427*	0.468^†	0.484^†
2	0.472*	0.395*	0.463^‡	0.403^‡
3	0.374^†	0.476^†	0.390*	0.460*
4	0.621^‡	0.605^‡	0.640^†	0.613^†
5	0.629^‡	0.597^‡	0.602^†	0.640^†
Ø reader	0.526*	0.500*	0.513^†	0.520^†

Rightlung		Leftlung
R1	71	L1/2	44
R2	56
R3	73	L3	42
R4	49	L4	56
R5	38	L5	7
R6	72	L6	52
R7	47	L7/8	42
R8	59
R9	49	L9	49
R10	50	L10	47

	Df	Deviance	Residual Df	Residual deviance	P value (chi-square test)
Null			1237	1715.6
Readers	4	36.95	1233	1678.65	<0.0001
Slices	17	95.81	1216	1582.85	<0.0001
Gunnar eyewear	1	0.48	1215	1582.36	0.49
Current	1	0.1	1214	1582.26	0.75
Readers: Gunnar eyewear	4	6.19	1210	1576.07	0.19
Readers: Current	4	11.27	1206	1564.8	0.02
Gunnar eyewear: current	1	10.16	1205	1554.65	0.002
Readers: Gunnar Eyewear: current	4	65.47	1201	1489.18	<0.0001

References

Henschke

, Naidich

, Yankelevitz

, Early Lung Cancer Action Project: initial findings on repeat screenings. Cancer 2001;92:153–9

Henschke

, Yankelevitz

, Mirtcheva

, CT screening for lung cancer: frequency and significance of part-solid and nonsolid nodules. Am J Roentgenol 2002;178:1053–7

Lee

, Gamsu

, Czum

, Lung nodule detection on chest CT: evaluation of a computer-aided detection (CAD) system. Korean J Radiol 2005;6:89–93

National Lung Screening Trial Research Team Aberle

, Adams

, Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395–409

, Rusinek

, Naidich

, Wavelet compression of low-dose chest CT data: effect on lung nodule detection. Radiology 2003;228:70–5

Christe

, Lin

, Yen

, CT patterns of fungal pulmonary infections of the lung: comparison of standard-dose and simulated low-dose CT. Eur J Radiol 2012;81:2860–6

Christe

, Charimo Torrente

, Lin

, Computed-tomography (CT) screening and follow up of lung nodules: effect of CT-tube current-time, nodule size and density on detectability of lung nodules and impact of tube current-time on the apparent nodule size. Am J Roentgenol 2011;197:623–30

, Arimura

, Suzuki

, Computer-aided detection of peripheral lung cancers missed at CT: ROC analyses without and with localization. Radiology 2005;237:684–90

Armato

, Li

, Giger

, Lung cancer: performance of automated lung nodule detection applied to cancers missed in a CT screening program. Radiology 2002;225:685–92

10.

Rubin

, Lyo

, Paik

, Pulmonary nodules on multi-detector row CT scans: performance comparison of radiologists and computer-aided detection. Radiology 2005;234:274–83

11.

Georgian-Smith

, Moore

, Halpern

, Blinded comparison of computer-aided detection with human second reading in screening mammography. Am J Roentgenol 2007;189:1135–41

12.

Lang

, Wortman

, Glickman

, Sensitivity of radiographic detection of lung metastases in the dog. Veterinary Radiology 1986;27:74–8

13.

Samei

, Badano

, Chakraborty

, Assessment of display performance for medical imaging systems: executive summary of AAPM TG18 report. Med Phys 2005;32:1205–25

14.

Gutierrez

, Monnin

, Valley

, A strategy to qualify the performance of radiographic monitors. Radiat Prot Dosimetry 2005;114:192–7

15.

Eperjesi

, Fowler

, Evans

. Do tinted lenses or filters improve visual performance in low vision? A review of the literature. Ophthalmic Physiol Opt 2002;22:68–77

16.

Zigman

. Vision enhancement using a short wave-length light-absorbing filter. Optom Vision Sci 1990;67:100–4

17.

Zigman

. Light filters to improve vision. Optom Vision Sci 1992;69:325–8

18.

Lynch

, Brilliant

. An evaluation of the Corning CPF550 lens. Optometric Monthly 1984;75:36–42

19.

Tupper

, Miller

. The effect of a 550 nm cut off filter on the vision of cataract patients. Ann Ophthal 1985;17:67–72

20.

Croft

, Trufasu

. Computer eyewear with spectral filtering. US Patent 2012/0307194A1, 2012. See freepatentsonline.com

21.

Henschke

, Yankelevitz

, Naidich

, CT screening for lung cancer: suspiciousness of nodules according to size on baseline scans. Radiology 2004;231:164–8

22.

MacMahon

, Austin

, Gamsu

, Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology 2005;237:395–400

23.

Sealy

, Connally

, Dalton

. Naming the bronchopulmonary segments and the development of pulmonary surgery. Ann Thorac Surg 1993;55:184–8

24.

Zar

. Biostatistical analysis. 5th edn. Upper Saddle River, NJ: Pearson Prentice-Hall, 2010

25.

Hosmer

, Lemeshow

. Applied logistic regression. 2nd edn. Hoboken, NJ: John Wiley & Sons, Inc., 2000

26.

Fleiss

. Measuring nominal scale agreement among many raters. Psychol Bull 1971;76:378–83

27.

Fleiss

. Statistical methods for rates and proportions. 2nd edn. New York, NY: John Wiley, 1981

28.

Latief

, White

, Protopapas

Search for a primary lung neoplasm in patients with brain metastasis: is the chest radiograph sufficient?

Am J Roentgenol 1997;168:1339–44

29.

Kaneko

, Eguchi

, Ohmatsu

, Peripheral lung cancer: screening and detection with low-dose spiral CT versus radiography. Radiology 1996;201:798–802

30.

Henschke

, McCauley

, Yankelevitz

, Early lung cancer action project: overall design and findings from baseline screening. Lancet 1999;354:99–105

31.

Armato

3rd , Giger

, Moran

, Computerized detection of pulmonary nodules on CT scans. Radiographics 1999;19:1303–11

32.

Brown

, McNitt-Gray

, Goldin

, Patientspecific models for lung nodule detection and surveillance in CT images. IEEE Trans Med Imaging 2001;20:1242–50

33.

, Betke

. Chest CT: automated nodule detection and assessment of change over time—preliminary experience. Radiology 2001;218:267–73

34.

Armato

3rd , Giger

, MacMahon

. Automated detection of lung nodules in CT scans: preliminary results. Med Phys 2001;28:1552–61

35.

Lee

, Hara

, Fujita

, Automated detection of pulmonary nodules in helical CT images based on an improved template-matching technique. IEEE Trans Med Imaging 2001;20:595–604

36.

Gurcan

, Sahiner

, Petrick

, Lung nodule detection on thoracic computed tomography images: preliminary evaluation of a computer-aided diagnosis system. Med Phys 2002;29:2552–8

37.

Wormanns

, Fiebich

, Saidi

, Automatic detection of pulmonary nodules at spiral CT: clinical application of a computer-aided diagnosis system. Eur Radiol 2002;12:1052–7

38.

Armato

3rd , Li

, Giger

, Lung cancer: performance of automated lung nodule detection applied to cancers missed in a CT screening program. Radiology 2002;225:685–92

39.

Ambrogi

, Paci

, Pompeo

, Transxiphoid video-assisted pulmonary metastasectomy: relevance of helical computed tomography occult lesions. Ann Thorac Surg 2000;70:1847–52

40.

Diederich

, Semik

, Lentschig

, Helical CT of pulmonary nodules in patients with extrathoracic malignancy: CT-surgical correlation. Am J Roentgenol 1999;172:353–60

41.

Fischbach

, Knollmann

, Griesshaber

, Detection of pulmonary nodules by multislice computed tomography: improved detection rate with reduced slice thickness. Eur Radiol 2003;13:2378–83

42.

Yamada

, Ono

, Tsuboi

, Low-dose CT of the thorax in cancer follow-up. Eur J Radiol 2004;51:169–74

43.

, Lee

, Kim

Multidetector CT of bronchiectasis: effect of radiation dose on image quality

. Am J Roentgenol 2003;181:501–5

44.

Tack

, De Maertelaer

, Petit

, Multi-detector row CT pulmonary angiography: comparison of standard-dose and simulated low-dose techniques. Radiology 2005;236:318–25

45.

Bankier

, Schaefer-Prokop

, De Maertelaer

, Air trapping: comparison of standard-dose and simulated low-dose thin-section CT techniques. Radiology 2007;242:898–906

46.

Hetmaniak

, Bard

, Albuisson

, Pulmonary nodules: dosimetric and clinical studies at low dose multidetector CT (in French). J Radiol 2003;84:399–404

47.

Weng

, Wu

, Pan

, The feasibility of low-dose CT for pulmonary metastasis in patients with primary gynecologic malignancy. Clin Imaging 2004;28:408–14

48.

Gergely

, Neumann

, Reiger

, Lung nodule detection with ultra-low-dose CT in routine follow-up of cancer patients (in German). Rofo 2005;177:1077–83

49.

Das

, Mühlenbruch

, Heinen

, Performance evaluation of a computer-aided detection algorithm for solid pulmonary nodules in low-dose and standard-dose MDCT chest examinations and its influence on radiologists. Br J Radiol 2008;81:841–7

			Tube current time
Reader	+ Gunnareyewear	− Gunnareyewear	150mAs	40mAs
1	0.532*	0.427*	0.468^†	0.484^†
2	0.472*	0.395*	0.463^‡	0.403^‡
3	0.374^†	0.476^†	0.390*	0.460*
4	0.621^‡	0.605^‡	0.640^†	0.613^†
5	0.629^‡	0.597^‡	0.602^†	0.640^†
Ø reader	0.526*	0.500*	0.513^†	0.520^†