Digital synthesis of lung nodules

Studies evaluating observer accuracy and visual perception of pulmonary nodules usually are based upon test films obtained from clinical practice in patients with proven pulmonary nodules. Unfortunately, such nodules do not always occur in the optimal size and location to facilitate testing. Such studies would be enhanced by the ability to place nodules of desired size and location on chest radiographs. This report describes a method of placing a computer-generated (synthesized) nodule on a digitized chest radiograph. To demonstrate the similarity of these synthesized nodules to real nodules, each digitized radiograph with a computer-generated nodule was paired with a digitized chest radiograph of a patient with a clinically proven pulmonary nodule. A total of 22 pairs of chest radiographs were then shown to 13 radiologists, who were asked to distinguish the synthesized nodule from the real nodule. With this two alternative forced-choice test, the radiologists were only able to distinguish the synthesized nodule in 51% of the cases, strongly suggesting that computer generated nodules may be used to simulate real pulmonary nodules in future tests of nodule detection.     

Insertion of virtual pulmonary nodules in CT data of the chest: development of a software tool

The purpose of this study was to develop a software tool for the insertion of virtual lung nodules into CT data. Forty software-generated nodules were inserted at random locations and sizes on 20 multi-detector row CT studies of the chest (4×1-2.5-mm slice collimation). On each scan, two virtual nodules were inserted. The size, shape, margin and attenuation could arbitrarily vary and were individually adjusted to match real lesions of each patient (real nodules: 6.5±3.1 mm; virtual nodules: 6.1±3.2 mm). Additionally, noise and a random pattern simulating local density variations were added to virtual nodules. Three blinded readers evaluated 40 real and 40 simulated nodules according to a 5-point confidence scale ranging from 1 (definitely simulated) to 5 (definitely real). A multivariate analysis of covariance was performed for statistical assessment (SPSS 11.5, Chicago, IL). Real and simulated lesions were indistinguishable for all three readers (Pillai’s trace statistic: P=0.881). However, nodule size was a statistically significant covariable regarding the differentiation of virtual compared to real nodules. Larger simulated nodules were easier to detect than smaller ones (Pillai’s trace statistic: P<0.05). The developed algorithm allowed for the synthetic generation of lung nodules that were indistinguishable from real nodules.     

Improved pulmonary nodule detection with scanning equalization radiography

The potential for improved pulmonary nodule detection with scanning equalization radiography (SER) was evaluated by means of observer performance testing during the interpretation of posteroanterior conventional radiographs and SER images of an anthropomorphic chest phantom with simulated nodules. A test set of 200 conventional and 200 SER radiographs of phantoms containing either one nodule or none was interpreted by four radiologists attempting to detect a nodule and indicate a confidence value. Their ability to detect nodules positioned over the lung was slightly improved with SER compared with conventional radiography (sensitivity, .56 vs .70); for nodules over the mediastinum or diaphragmatic areas, it was much improved (sensitivity, .29 vs .64). The results were also analyzed with receiver-operating characteristic methods, which revealed a significant improvement in lesion detect-ability over the thicker body parts with SER images. The capability of equalized chest radiographs to provide improved lesion detectability suggests that SER may set a new standard for film-based chest radiography and have a large clinical application.     

Subtle lung nodules: influence of local anatomic variations on detection

PURPOSE: To assess the influence of local anatomic noise on the detection of subtle lung nodules depicted on chest radiographs. MATERIALS AND METHODS: Six 8 x 8-cm lung regions were extracted from digital chest radiographs obtained in healthy subjects. Simulated nodules emulating the radiographic characteristics of subtle tissue-equivalent lesions 3.2-6.4 mm in diameter (equivalent to 0.1-0.4 mm in contrast-diameter product [CD]) were added to the images. On multiple renditions of each image, nodules were inserted at slightly different locations within 6 mm of the center; this process allowed different local background patterns to overlie the nodules. An observer detection study involving 15 experienced radiologists was performed. The authors performed analysis of variance and pairwise t test analyses to determine variations in nodule detectability related to nodule location and size on each image. RESULTS: Results indicated a strong correlation between nodule size and observer detection score and significant variation in nodule detectability as a function of location. Changes in nodule position caused observer score variations that were equivalent to the variation caused by an up to 185% change in nodule CD (78% average over all six images), an up to 68% change in diameter (32% average), and an up to 28% change in area under the receiver operating characteristic curve (Az) (14% average). CONCLUSION: Local anatomic variations surrounding and overlying a subtle lung nodule on a chest radiograph that are created by the projection of anatomic features in the thorax, such as ribs and pulmonary vessels, can greatly influence the detection of nodules, altering the Az by as much as 28%.     

肋骨抑制成像在胸部平片肺结节检出中的应用

目的 评价肋骨抑制成像技术在胸部平片检出肺结节中的价值.方法 回顾性分析141例胸部后前位X线片,其中95例有单发肺结节,作为研究组,46例无肺结节作为对照组.2名高年资与2名低年资放射科医师分别独立阅读所有胸部后前位X线片和经肋骨抑制成像技术处理后的胸部后前位X线片.阅读、记录结节的部位、大小并对结节存在的肯定度进行评分.观察的结果采用受试者工作特征(ROC)曲线进行分析.结果 肺结节的平均直径为(1.9±1)cm,直径范围是0.9~2.9 cm.胸部后前位X线片ROC曲线下面积(AUC)为0.844,肋骨抑制成像技术处理后的胸部后前位X线片AUC为0.873,两者有统计学差异(P<0.01).结论 肋骨抑制成像技术可以显著提高放射科医师对胸部后前位X片中肺结节的检出率.     

计算机辅助检测肺结节在数字化X线胸片肺癌筛查中的应用价值

目的 评价计算机辅助检测(CAD)肺结节系统在数字化X线胸片上肺癌筛查中的应用价值.方法 由1名放射科医师和CAD肺结节检测系统独立阅读100例连续的数字摄影(DR)X线胸片,CAD系统可以检出最长直径在5～15 mm的肺结节.由2名放射科专家(有15年胸部影像诊断经验)进行回顾性阅读,参照相应的CT图像,两人意见达成一致后标记真结节的个数和位置并保存标记结果,将标记结果作为金标准来比较放射科医师和CAD系统的肺结节检测敏感性和假阳性率.结果 放射科医师共检测到95个结节,CAD系统共检测到304个结节.在回顾性检查中2名放射科专家共标记134个真结节,其中放射科医师检测到82个(61.2%),CAD检测到105个(78.4%),CAD系统检测到而被放射科医师漏诊的结节35个,放射科医师检测到而CAD系统漏诊的结节10个.放射科医师应用CAD系统后检测到112个真结节,检测率提高到83.6%.放射科专家意见一致后认为CAD系统检出199个假阳性结节,平均每张胸片约2.0个.结论 在肺癌筛查中放射科医师和CAD系统必须联合应用才可以识别X线胸片中所有的结节.     

Development of a digital image database for chest radiographs with and without a lung nodule: receiver operating characteristic analysis of radiologists' detection of pulmonary nodules

OBJECTIVE: We developed a digital image database (www.macnet.or.jp/jsrt2/cdrom_nodules.html ) of 247 chest radiographs with and without a lung nodule. The aim of this study was to investigate the characteristics of image databases for potential use in various digital image research projects. Radiologists' detection of solitary pulmonary nodules included in the database was evaluated using a receiver operating characteristic (ROC) analysis. MATERIALS AND METHODS: One hundred and fifty-four conventional chest radiographs with a lung nodule and 93 radiographs without a nodule were selected from 14 medical centers and were digitized by a laser digitizer with a 2048 x 2048 matrix size (0.175-mm pixels) and a 12-bit gray scale. Lung nodule images were classified into five groups according to the degrees of subtlety shown. The observations of 20 participating radiologists were subjected to ROC analysis for detecting solitary pulmonary nodules. Experimental results (areas under the curve, Az) obtained from observer studies were used for characterization of five groups of lung nodules with different degrees of subtlety. RESULTS: ROC analysis showed that the database included a wide range of various nodules yielding Az values from 0.574 to 0.991 for the five categories of cases for different degrees of subtlety. CONCLUSION: This database can be useful for many purposes, including research, education, quality assurance, and other demonstrations.     

Cost-effectiveness of stereoscopic radiographs in detection of lung nodules

Detection of lung nodules was assessed using a series of radiographs which were either normal or showed only one nodule. Stereoscopic pairs were obtained in 26 cases (10 positive and 16 normal). A "singles" series comprising one radiograph from each pair was presented to each observer on three occasions. The pairs were later presented side by side and finally were viewed stereoscopically. The results indicate that a second shifted radiograph improves detection of lung nodules enough to be cost-effective. The shifted radiographs do not need to be viewed stereoscopically.     

Comparison of radiologist and CAD performance in the detection of CT-confirmed subtle pulmonary nodules on digital chest radiographs

OBJECTIVES: Detection of subtle pulmonary nodules on digital radiography is a challenging task for radiologists. The aim of this study was to evaluate the performance of a newly approved computer aided detection (CAD) system. MATERIALS AND METHODS: The sensitivity of 3 radiologists and of a CAD system for the detection of pulmonary nodules from 5 to 15 mm in size on digital chest radiography of 117 patients was compared. The reference standard was established by consensus reading of computed tomography scans by 2 experienced radiologists. Computed tomography scans and chest radiographs were performed within 4 weeks. Sixty-six pulmonary nodules from 42 patients, with a mean nodule diameter of 7.5 mm (standard deviation: 2.2 mm), were included in the statistical analysis. Seventy-five of the 117 patients did not have nodules from 5 to 15 mm of size. RESULTS: Two hundred and eighty-eight false-positive detections of the CAD system were found with an average of 2.5 false-positives per image. Sensitivity of the CAD system was 39.4% (95% confidence interval: 11.8%), when compared with 18.2% to 30.3% (95% confidence interval 9.3% to 11.1%) of the 3 radiologists. Substantial agreement for nodule detection ([kappa]N: 0.64-0.73) was found among the 3 radiologists, whereas only moderate agreement was found between the radiologists and the CAD performance ([kappa]N: 0.45-0.52). CONCLUSIONS: The CAD system's diagnostic sensitivity in detecting pulmonary nodules of 5 to 15 mm of size was superior to the 1 of radiologists. The CAD system may be used for assisting the radiologist in the detection of lung nodules on digital chest radiographs.     

Effect of high sensitivity in a computerized scheme for detecting extremely subtle solitary pulmonary nodules in chest radiographs: observer performance study

RATIONALES AND OBJECTIVES: This study investigated the effect of a high sensitivity in computer-aided diagnosis (CAD) for detecting lung nodules in chest radiographs when extremely subtle cases were presented to radiologists. MATERIAL AND METHODS: The chest radiographs used in this study consisted of 36 normal images and 54 abnormals containing solitary lung nodules, of which 25 were extremely subtle and 29 were very subtle. Receiver operating characteristic analysis for detecting lung nodules was performed without and with CAD. The levels of CAD output were simulated with a hypothetical ideal performance of 100% sensitivity, but with three or four false positives per image. Six radiologists participated in an observer study in which cases were interpreted first without and then with the use of CAD. RESULTS: The average A(z) values for radiologists without and with CAD were 0.682 and 0.808, respectively. The performance of radiologists was improved significantly when high sensitivity was used (P = .0003). However, the radiologists were not able to recognize some extremely subtle nodules (5 of 54 nodules by all radiologists), even with the correct CAD output; these nodules were then considered as non-actionable. None of 306 computer-false positives was incorrectly regarded as a nodule by all radiologists, but 63 false positives were incorrectly identified by one or more radiologists. CONCLUSION: The accuracy of radiologists in the detection of some extremely subtle solitary pulmonary nodules can be improved significantly when the sensitivity of a CAD scheme can be made to be at an extremely high level. However, all of the six radiologists failed to identify some nodules (about 10%), even with the correct output of the CAD.     

Effects of image processing on nodule detection rates in digitized chest radiographs: ROC study of observer performance

To evaluate the effects of image processing in digitized chest radiographs when high-resolution images are used, an examination was done in which the detection of pulmonary nodules in unprocessed digitized chest radiographs was compared with that in images that had undergone processing with two methods, adaptive filtration and histogram equalization. The processing techniques have been optimized in previous work to selectively enhance the retrocardiac and subdiaphragmatic areas without significant alteration of detail in the lung. Eight observers were shown 150 test radiographs (50 unprocessed, 50 processed with adaptive filtration, 50 processed with histogram equalization) containing 150 nodules. The results indicate a statistically significant (P less than .03) difference, with highest observer performance in the chest radiographs processed with adaptive filtration (median area under ROC curve = 0.78), compared with unprocessed images (median = 0.68) and chest radiographs processed with histogram equalization (median = 0.62). Performance in the lung was not significantly different. Adaptive filtration applied to selectively enhance underexposed areas of film images may improve nodule detection. Histogram equalization provided no improvement in performance.     

Improved detection of lung nodules on chest radiographs using a commercial computer-aided diagnosis system

OBJECTIVE: The aim of this study was to evaluate the usefulness of a new commercially available computer-aided diagnosis (CAD) system with an automated method of detecting nodules due to lung cancers on chest radiograph. MATERIALS AND METHODS: For patients with cancer, 45 cases with solitary lung nodules up to 25 mm in diameter (nodule size range, 8-25 mm in diameter; mean, 18 mm; median, 20 mm) were used. For healthy patients, 45 cases were selected on the basis of confirmation on chest CT. All chest radiographs were obtained with a computed radiography system. The CAD output images were produced with a newly developed CAD system, which consisted of an image server including CAD software called EpiSight/XR. Eight radiologists (four board-certified radiologists and four radiology residents) participated in observer performance studies and interpreted both the original radiographs and CAD output images using a sequential testing method. The observers' performance was evaluated with receiver operating characteristic analysis. RESULTS: The average area under the curve value increased significantly from 0.924 without to 0.986 with CAD output images. Individually, the use of CAD output images was more beneficial to radiology residents than to board-certified radiologists. CONCLUSION: This CAD system for digital chest radiographs can assist radiologists and has the potential to improve the detection of lung nodules due to lung cancer.     

Evaluation of diagnostic accuracy of CRT monitor display for personal computer in the detection of small lung nodules: with particular emphasis on comparison between JPEG and wavelet compression

PURPOSE: To compare observer performance on cathode-ray-tube(CRT) monitors for personal computers with that on conventional radiographs in the detection of small lung nodules. MATERIALS AND METHODS: Fifty-eight normal chest radiographs and 58 chest radiographs with a small lung nodule were selected. Ten radiologists examined the original conventional films on a viewbox and digitized (8 bit) uncompressed and compressed images of the same patient on a color CRT monitor with a matrix of 1,600 x 1,200, and rated the presence of lung nodules with a five-level scale of confidence. The methods of compression used in this study were the JPEG and wavelet methods, with compression ratios of 6:1 and 15:1. Results were analyzed by receiver operating characteristic methods. RESULTS: There was no significant difference between film and digitized uncompressed and compressed images obtained by the JPEG and wavelet methods with a compression ratio of 6:1. No statistically significant difference was detected between film and digitized image with wavelet compression at 15:1. However, detection was less accurate on digitized images with JPEG compression at 15:1. CONCLUSION: Digitized (8 bit) uncompressed and compressed images with a compression ratio of 6:1 are acceptable for the detection of small lung nodules. Digitized compressed images at a compression ratio of 15:1 are also acceptable when the wavelet method is used.     

Detection of subtle lung nodules: relative influence of quantum and anatomic noise on chest radiographs

PURPOSE: To assess the relative influence of quantum mottle and structured lung patterns (anatomic noise) on the detection of subtle lung nodules on chest radiographs. MATERIALS AND METHODS: Sixty 8 x 8-cm lung pattern images were extracted from digital chest radiographs in healthy individuals. Sixty quantum mottle images of the same size and quantum noise level were extracted from uniformly exposed digital radiographs. Simulated nodules with various peak contrast-diameter products (CD) that emulated subtle tissue-equivalent lung nodules were numerically superimposed at the center on three-fourths of the images. Printouts were independently viewed and scored by five experienced radiologists. The area under the receiver operating characteristic curve (Az) was estimated as a measure of the detectability of the nodules. RESULTS: At a fixed observer performance level (e.g., Az = 0.8), much smaller and lower-contrast nodules were detected on quantum mottle images (1-mm diameter, CD = 0.01 mm), compared with those on anatomic images (4.5-mm diameter, CD = 0.20 mm). The findings generally agreed with the signal-to-noise ratio calculations based on statistical observer models. CONCLUSION: The detection of subtle lung nodules on chest radiographs is limited by anatomic noise.     

Observer-independent nodule-detectability index for low-dose lung cancer screening CT: a pilot study

For the wide dissemination of lung cancer screening by low-dose computed tomography (CT), it is important to determine the optimal conditions for scan and image reconstruction based on objective standards of evaluation. Our aim in this study was to propose a quantitative index of nodule detectability without an observer test. It was essential to determine the apparent size and density of nodules visible on CT images for developing the nodule-detectability index based on a statistical observer-independent method. Therefore, we introduced a computer simulation technique for CT images based on the spatial resolution of the system to evaluate the size and density accurately. By use of scan/reconstruction parameter settings as employed for low-dose CT screening, a detectability index was obtained for target nodules (ideal spheres) of various sizes and with varying contrast (ΔCT) between nodule density and background density. The index was compared with the qualitative results of observer tests of nodule detectability. As the target nodule diameter or ΔCT was increased, the index value increased, implying improved nodule visibility. According to the index, the detection limits for nodules with ΔCTs of 70, 100, or 150 Hounsfield units were approximately 6, 5, and 4 mm in diameter, respectively. Index values were well correlated with nodule detectability as assessed by four observers. The proposed index was effective for quantifying nodule detectability, and its validity was confirmed by an observer test. This index has potential use in the determination of optimal scan/reconstruction parameters for lung cancer screening by low-dose CT without observer test.     

Neural network based detection of pulmonary nodules on chest radiographs

PURPOSE: We investigated the capabilities of an artificial neural network-based Computer-Aided Diagnosis (CAD) system in improving early detection of pulmonary nodules on chest radiographs. MATERIAL AND METHODS: We used a data-set of 145 digitized chest films. Two different radiologists read the radiographs to detect the sites of possible nodules. The system uses two neural networks trained on a training-set of 100 radiographs selected from the data-set. The first network is used to focus attention on the sites of potential nodules while the second calculates the likeliness of nodule presence in ROIs. The clinical test was performed on 45 more radiographs from the training-set, but different from those in the data-set, which were positive for both benign and malignant nodules. These latter plain films showed 65 nodular lesions which differed by shape and acquisition technique. RESULTS: Sensitivity was 89% in all radiographs while specificity, evaluated by ROI, and accuracy, were 98%. CONCLUSIONS: There are potential limitations in nodule detection on plain radiographs. Some of them are operator-dependent, such as nonsystematic investigation, lesion underestimation, and poor reading, and some are technique-dependent, such as X-ray beam/tube, low voltage, patient positioning, focus-film distance and development process. CADs may contribute to improving detection of pulmonary nodules because the false-negative rate is decreased and sensitivity consequently increased. The high sensitivity and specificity rates of neural networks encourage further trials on wider data-sets to help the radiologist in the early detection of pulmonary nodules.     

VolumeRAD容积成像在肺结节诊断中的应用

目的探讨VolumeRAD容积成像检测肺结节的价值。方法39例，男21例，女18例，平均年龄（53．8±9．7）岁。用CT进行随访的肺结节患者进行了额外的PA胸部平片检查和VolumeRAD容积成像。三名胸部x线诊断医师在PACS工作站上独立对胸部平片和VolumeRAD图像进行分析，结节依据清晰度被分为“清晰可见”，“不确定”，和“不可见”，依据大小分为：A组（直径〈5mm），B组（直径5～10mm），C组（直径〉10mm）。胸部CT作为参考标准，结节“清晰可见”为真阳性，计算胸部平片和VolumeRAD成像对肺结节的诊断灵敏度。结果CT上共显示结节247个，胸部平片和VolumeRAD成像对肺结节诊断的灵敏度分别为（69．90±1．30）％和（24．02±1．17）Y00（P〈0．05），在各组间和总数间差别均有统计学意义。结论与PA胸片比较，VolumeRAD容积成像具有提高已知肺小结节检测性能的潜力。     

Pulmonary nodule detection using chest CT images

Purpose: 
Automated methods for the detection of pulmonary nodules and nodule volume calculation on CT are described. Material and Methods: 
Gray-level threshold methods were used to segment the thorax from the background and then the lung parenchyma from the thoracic wall and mediastinum. A deformable model was applied to segment the lung boundaries, and the segmentation results were compared with the thresholding method. The lesions that had high gray values were extracted from the segmented lung parenchyma. The selected lesions included nodules, blood vessels and partial volume effects. The discriminating features such as size, solid shape, average, standard deviation and correlation coefficient of selected lesions were used to distinguish true nodules from pseudolesions. With texture features of true nodules, the contour-following method, which tracks the segmented lung boundaries, was applied to detect juxtapleural nodules that were contiguous to the pleural surface. Volume and circularity calculations were performed for each identified nodule. The identified nodules were sorted in descending order of volume. These methods were applied to 827 image slices of 24 cases. Results: 
Computer-aided diagnosis gave a nodule detection sensitivity of 96% and no false-positive findings. Conclusion: 
The computer-aided diagnosis scheme was useful for pulmonary nodule detection and gave characteristics of detected nodules.     

Evaluation of automated lung nodule detection on low-dose computed tomography scans from a lung cancer screening program(1)

RATIONALE AND OBJECTIVES: The purpose of this study was to evaluate the performance of a fully automated lung nodule detection method in a large database of low-dose computed tomography (CT) scans from a lung cancer screening program. Because nodules demonstrate a spectrum of radiologic appearances, the performance of the automated method was evaluated on the basis of nodule malignancy status, size, subtlety, and radiographic opacity. MATERIALS AND METHODS: A database of 393 thick-section (10 mm) low-dose CT scans was collected. Automated lung nodule detection proceeds in two phases: gray-level thresholding for the initial identification of nodule candidates, followed by the application of a rule-based classifier and linear discriminant analysis to distinguish between candidates that correspond to actual lung nodules and candidates that correspond to non-nodules. Free-response receiver operating characteristic analysis was used to evaluate the performance of the method based on a jackknife training/testing approach. RESULTS: An overall nodule detection sensitivity of 70% (330 of 470) was attained with an average of 1.6 false-positive detections per section. At the same false-positive rate, 83% (57 of 69) of the malignant lung nodules in the database were detected. When the method was trained specifically for malignant nodules, a sensitivity of 80% (55 of 69) was attained with 0.85 false-positives per section. CONCLUSION: We have evaluated an automated lung nodule detection method with a large number of low-dose CT scans from a lung cancer screening program. An overall sensitivity of 80% for malignant nodules was achieved with 0.85 false-positive detections per section. Such a computerized lung nodule detection method is expected to become an important part of CT-based lung cancer screening programs.     

Computer-Aided Detection of Malignant Lung Nodules on Chest Radiographs: Effect on Observers' Performance

Objective

To evaluate the effect of computer-aided detection (CAD) system on observer performance in the detection of malignant lung nodules on chest radiograph.

Materials and Methods

Two hundred chest radiographs (100 normal and 100 abnormal with malignant solitary lung nodules) were evaluated. With CT and histological confirmation serving as a reference, the mean nodule size was 15.4 mm (range, 7-20 mm). Five chest radiologists and five radiology residents independently interpreted both the original radiographs and CAD output images using the sequential testing method. The performances of the observers for the detection of malignant nodules with and without CAD were compared using the jackknife free-response receiver operating characteristic analysis.

Results

Fifty-nine nodules were detected by the CAD system with a false positive rate of 1.9 nodules per case. The detection of malignant lung nodules significantly increased from 0.90 to 0.92 for a group of observers, excluding one first-year resident (p = 0.04). When lowering the confidence score was not allowed, the average figure of merit also increased from 0.90 to 0.91 (p = 0.04) for all observers after a CAD review. On average, the sensitivities with and without CAD were 87% and 84%, respectively; the false positive rates per case with and without CAD were 0.19 and 0.17, respectively. The number of additional malignancies detected following true positive CAD marks ranged from zero to seven for the various observers.

Conclusion

The CAD system may help improve observer performance in detecting malignant lung nodules on chest radiographs and contribute to a decrease in missed lung cancer.