Artificial neural networks (ANNs) for differential diagnosis of interstitial lung disease: results of a simulation test with actual clinical cases

RATIONALE AND OBJECTIVES: To evaluate the performance of an artificial neural network (ANN) scheme with use of consecutive clinical cases and its effect on radiologists with an observer test. MATERIALS AND METHODS: Artificial neural networks were designed to distinguish among 11 interstitial lung diseases on the basis of 26 inputs (16 radiologic findings, 10 clinical parameters). Chest radiographs of 96 consecutive cases with interstitial lung disease were used. Five radiologists independently rated their radiologic findings on the 96 chest radiographs. Based on their ratings of radiologic findings and clinical parameters obtained from the hospital information system, the output values indicating the likelihood of each of the 11 interstitial lung diseases were determined. Subsequently, 30 cases were selected from these 96 cases for an observer test. Five radiologists marked their confidence levels for diagnosis of 11 possible diseases in each case without and with ANN output. The performance of ANNs and radiologists was evaluated by receiver operating characteristic analysis based on their outputs and on confidence levels, respectively. RESULTS; The average Az value (area under the receiver operating characteristic curve) indicating ANN performance for the 96 consecutive cases was 0.85 +/- 0.03. The average Az values indicating radiologists' performance without and with ANN outputs were 0.81 +/- 0.11 and 0.87 +/- 0.06, respectively. The diagnostic accuracy was improved significantly when radiologists read chest radiographs with ANN outputs (P < .05). CONCLUSION: Artificial neural networks for differential diagnosis of interstitial lung disease may be useful in clinical situations, and radiologists may be able to utilize the ANN output to their advantage in the differential diagnosis of interstitial lung disease on chest radiographs.     

Reader variance in ROC studies--generalizability to reader population at high and low performance levels

RATIONALE AND OBJECTIVES: To investigate the variability between discriminative performances of readers as a function of average performance levels during receiver operating characteristic (ROC) studies. MATERIALS AND METHODS: Four subsets of cases from previously ascertained ROC rating data by 12 observers when detecting interstitial disease and pneumothorax on posteroanterior chest films were selected for each abnormality and reanalyzed to assess changes in "reader" variance component. The subsets were selected based on a prestudy subjective assessment of the subtleness of depicted abnormality (positive cases) and the difficulty in determining its absence (negative cases). Reader variance component was estimated using a bootstrap approach for each subset and the results were used to assess a general relationship between variability and average performance level. RESULTS: The reader variance component decreased substantially (from 0.007704 to 0.000426), as expected, when the areas under the ROC curves (AUC) for detecting pneumothoraces increased from 84% to 97%. On the other hand, reader variance component increased substantially (from 0.000890 to 0.005181) when AUC for detecting interstitial disease increased from 59% to 87%. The large magnitude of and changes in the reader variance component resulted in a consistent nonmonotone relationship as a function of AUC when other related variance components were included in addition to the reader component. CONCLUSION: Among several factors affecting generalizability of ROC results to the population of readers, the reader variance component depended nonmonotonically on the average diagnostic performance and is lowest at both very high and very low levels of performance.     

Home teleconsultations with faculty subspecialists on chest images from an intensive care unit

Study objective: To determine whether teleconsultations by first-year radiology residents with faculty thoracic radiologists, using diagnostic-quality teleradiology workstations in the radiologists' homes, would add clinical value for the acute management of intensive care unit patients after regular working hours. Design and setting: First-year radiology residents recorded key findings on new computed chest radiographs from 173 cardiothoracic intensive care unit patients. After consulting with home-based thoracic radiologists on the same images via teleradiology, they recorded any revisions to their original interpretations. An interdisciplinary evaluation panel determined whether the revisions of the residents' initial readings after the teleradiology consultations would have influenced the acute clinical care of these patients. Measurements and results: In 119 of the 173 cases (69 %), differences in key findings on the chest images were observed between the first-year residents' preliminary readings and their revised readings after teleconsultation with a thoracic radiologist. The evaluation panel determined that the changes in key findings after the teleconsultations could have influenced acute patient care in 86 of the 173 cases (50 %). Conclusions: Through rapid teleradiology consultations with residents, focusing on the key findings on newly-obtained computed chest images, home-based thoracic radiologists provided information of added clinical value for the acute management of cardiothoracic intensive care unit patients in one-half of the cases studied. Diagnostic workstations in the homes of faculty subspecialists may enable first-year radiology residents on night or weekend duty to obtain clinical supervision from faculty subspecialists that approximates more closely the level of supervision that they receive during regular working hours.     

Performance of a flat-panel detector in the detection of artificial erosive changes: comparison with conventional screen-film and storage-phosphor radiography

The purpose of this study was to compare a large-area, direct-readout, flat-panel detector system with a conventional screen-film system, a storage-phosphor system, and a mammography screen-film system with regard to the detection of artificial bone erosions simulating rheumatoid disease, and to assess its diagnostic performance with decreasing exposure dose. Six hundred forty regions were defined in 160 metacarpophalangeal and proximal interphalangeal joint specimens from 20 monkey paws (4 regions per joint). Artificial bone erosions were created in 320 of these 640 regions. Specimens were enclosed in containers filled with water to obtain absorption and scatter radiation conditions similar to those of a human hand. Imaging was performed using a flat-panel system, a speed class 200 screen-film system, a mammography screen-film system, and a storage-phosphor system under exactly matched conditions. Different exposure doses equivalent to speed classes of S=100, 200, 400, 800, 1600, and 3200 were used. In all images the presence or absence of a lesion was assessed by three radiologists using a five-level confidence scale. Receiver operating characteristic (ROC) analysis was performed for a total of 21,120 observations (1920 for each imaging modality and exposure level) and diagnostic performance estimated by the area under the ROC curve (A_z). The significance of differences in diagnostic performance was tested with analysis of variance. The ROC analysis showed A_z values of 0.809 (S=200), 0.768 (S=400), 0.737 (S=800), 0.710 (S=1600), and 0.685 (S=3200) for the flat-panel system, 0.770 for the speed class 200 screen-film system, 0.781 (S=200), 0.739 (S=400), 0.724 (S=800), 0.680 (S=1600) for the storage-phosphor system, and 0.798 for the mammography screen-film system. Analysis of variance showed significant differences between different combinations of imaging modalities and exposure doses ( p <0.05). The diagnostic performance of the flat-panel detector system is superior to that of a screen-film system and a storage-phosphor system for the detection of erosive lesions at clinical exposure settings (S=200). Using the flat-panel system the exposure dose can be reduced by 50% to obtain a diagnostic performance comparable to a speed class 200 screen-film system. Electronic Publication     

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.     

侵袭性肺曲霉菌病的影像学诊断

目的 探讨侵袭性肺曲霉菌病（IPA）的X线及CT表现，提高对该病的认识。方法回顾性分析经病原学或病理组织学证实的IPA6例（男4例、女2例）。基础病变包括肾移植术后5例及白血病1例，全部病例均行胸部X线及CT检查。结果4例表现为多发斑片状及片状影，其中2例伴空洞形成，1例表现为多发小结节影，1例表现为单发团块状影。影像诊断肺霉菌病3例，误诊3例。结论IPA的影像学表现无特异性，影像学检查结合临床资料对本病的诊断有一定价值。     

High-resolution computed tomography (HRCT) in the detection of “early asbestosis”

Two radiologists reviewed in masked fashion 144 standard chest radiographs, previously judged as normal by other readers, within a cohort of asymptomatic shipyard workers exposed to amosite. Among the 144 workers we selected subjects (n = 72) who fitted the following criteria: (1) documented occupational exposure to asbestos; (2) absence of any clinical symptomatology suggestive of asbestosis or of any other lung disease. Thirty-eight of them had standard chest radiogrphs that were confirmed as normal by both our radiologists, while 34 showed suspected pleural plaques or a very slight parenchymal involvement. These 72 subjects underwent high-resolution CT (HRCT): pleural plaques were shown in 33, parenchymal alterations in 7, and both pleural and parenchymal involvement in 13. HRCT findings were strictly correlated to the duration of amosite exposure and to the latency time since first exposure. Sensitivity, specificity and diagnostic accuracy values of standard chest radiographs were calculated with respect to HRCT and found to be, for pleural and parenchymal findings respectively: sensitivity 53% and 19%; specificity 72% and 94%; accuracy 60% and 72%. In conclusion, pleural and/or pulmonary involvement in asbestos-exposed workers can be shown by HRCT before the appearance of any symptomatology and abnormality on chest radiography. Furthermore the HRCT findings are correlated to the duration of exposure and latency time. time. Correspondence to; P. Boraschi     

Detectability of simulated pulmonary nodules on chest radiographs: Comparison between irradiation side sampling indirect flat-panel detector and computed radiography

Objective

To compare the detectability of simulated pulmonary nodules on chest radiographs between an irradiation side sampling indirect flat-panel detector (ISS-FPD) and computed radiography (CR).

Materials and methods

This study was an observer performance study. Simulated pulmonary nodules of 8 mm in diameter were superimposed on an anthropomorphic chest phantom. Chest radiographs were acquired under 2 exposure levels (4 and 3.2 mAs) with the ISS-FPD and the CR. Six thoracic radiologists evaluated all 40 images (10 patterns × 2 different exposure doses × 2 different systems) for the presence or absence of a lesion over each of 12 defined areas on a 3-megapixel monochrome liquid-crystal display. Receiver operating characteristic (ROC) curves were obtained for observation in predefined 480 areas. A jackknife method was used for statistical analysis. Differences with a P value of <0.05 were considered significant.

Results

The analysis of the observer detection of simulated pulmonary nodules showed larger areas under the ROC curve (AUC) by the ISS-FPD than by the CR. There was a statistically significant difference between the two systems at 3.2 mAs (P = 0.0330).

Conclusion

The ISS-FPD was superior to the CR for the detection of simulated pulmonary nodules at 3.2 mAs.     

Usefulness of computerized method for lung nodule detection on digital chest radiographs using similar subtraction images from different patients

Purpose

The purpose of this study is to evaluate the usefulness of a novel computerized method to select automatically the similar chest radiograph for image subtraction in the patients who have no previous chest radiographs and to assist the radiologists’ interpretation by presenting the “similar subtraction image” from different patients.

Materials and methods

Institutional review board approval was obtained, and the requirement for informed patient consent was waived. A large database of approximately 15,000 normal chest radiographs was used for searching similar images of different patients. One hundred images of candidates were selected according to two clinical parameters and similarity of the lung field in the target image. We used the correlation value of chest region in the 100 images for searching the most similar image. The similar subtraction images were obtained by subtracting the similar image selected from the target image. Thirty cases with lung nodules and 30 cases without lung nodules were used for an observer performance test. Four attending radiologists and four radiology residents participated in this observer performance test.

Results

The AUC for all radiologists increased significantly from 0.925 to 0.974 with the CAD (P = .004). When the computer output images were available, the average AUC for the residents was more improved (0.960 vs. 0.890) than for the attending radiologists (0.987 vs. 0.960).

Conclusion

The novel computerized method for lung nodule detection using similar subtraction images from different patients would be useful to detect lung nodules on digital chest radiographs, especially for less experienced readers.     

Assessing the use of digital radiography and a real-time interactive pulmonary nodule analysis system for large population lung cancer screening

Rationale and objectives

To assess the use of chest digital radiograph (DR) assisted with a real-time interactive pulmonary nodule analysis system in large population lung cancer screening.

Materials and methods

346 DR/CR patient studies with corresponding CT images were selected from 12,500 patients screened for lung cancer from year 2007 to 2009. Two expert chest radiologists established CT-confirmed Gold Standard of nodules on DR/CR images with consensus. These cases were read by eight other chest radiologists (participating radiologists) first without using a real-time interactive pulmonary nodule analysis system and then re-read using the system. Performances of participating radiologists and the computer system were analyzed.

Results

The computer system achieved similar performance on DR and CR images, with a detection rate of 76% and an average FPs of 2.0 per image. Before and after using the computer-aided detection system, the nodule detection sensitivities of the participating radiologists were 62.3% and 77.3% respectively, and the A_z values increased from 0.794 to 0.831. Statistical analysis demonstrated statically significant improvement for the participating radiologists after using the computer analysis system with a P-value 0.05.

Conclusion

The computer system could help radiologists identify more lesions, especially small ones that are more likely to be overlooked on chest DR/CR images, and could help reduce inter-observer diagnostic variations, while its FPs were easy to recognize and dismiss. It is suggested that DR/CR assisted by the real-time interactive pulmonary nodule analysis system may be an effective means to screen large populations for lung cancer.     

35例肺结核的CT误诊分析

目的　分析 35例肺结核误诊的原因 ,总结肺结核的CT表现。方法　 35例误诊病例中男 19例 ,女 16例 ,年龄 18～ 79岁。均做胸部CT平扫 ,由 2位主任医师逐例回顾分析CT片 ,并与病理对照 ,找出误诊原因。结果　 35例肺结核误诊为肺癌 2 9例 ,肺炎 4例 ,其他疾病 2例。误诊病例中呈结节及肿块病灶 14例 ,呈肺段、肺叶阴影 19例 ,肺门及纵隔肿块 2例。结论　CT表现不典型、CT检查不全面及胸部CT表现未与胸片表现综合分析是误诊的主要原因     

Lung disease in premature neonates: Impact of new treatments and technologies

Advances in perinatal medicine and neonatology have dramatically changed clinical outcomes for premature neonates and have ushered in a new era of radiological complexity. “Portable” chest radiographs continue to be the mainstay in diagnostic imaging of fragile newborns, but radiologists may be confronted with new and unexpected radiological expressions of oncefamiliar disease processes. Familiarity with the radiological impact of emerging treatments in premature neonates is essential for accurate film interpretation.     

Effect of an artificial neural network on radiologists' performance in the differential diagnosis of interstitial lung disease using chest radiographs

OBJECTIVE: We developed a new method to distinguish between various interstitial lung diseases that uses an artificial neural network. This network is based on features extracted from chest radiographs and clinical parameters. The aim of our study was to evaluate the effect of the output from the artificial neural network on radiologists' diagnostic accuracy. MATERIALS AND METHODS: The artificial neural network was designed to differentiate among 11 interstitial lung diseases using 10 clinical parameters and 16 radiologic findings. Thirty-three clinical cases (three cases for each lung disease) were selected. In the observer test, chest radiographs were viewed by eight radiologists (four attending physicians and four residents) with and without network output, which indicated the likelihood of each of the 11 possible diagnoses in each case. The radiologists' performance in distinguishing among the 11 interstitial lung diseases was evaluated by receiver operating characteristic (ROC) analysis with a continuous rating scale. RESULTS: When chest radiographs were viewed in conjunction with network output, a statistically significant improvement in diagnostic accuracy was achieved (p < .0001). The average area under the ROC curve was .826 without network output and .911 with network output. CONCLUSION: An artificial neural network can provide a useful "second opinion" to assist radiologists in the differential diagnosis of interstitial lung disease using chest radiographs.     

Comparison of scanning equalization and conventional chest radiography

A clinical comparison study of scanning equalization radiography (SER) and conventional chest radiography was performed with the latest prototype SER system. Conventional chest radiography was performed at 120 kVp with Lanex regular screens (Eastman Kodak, Rochester, NY) and Kodak Ortho-G or Ortho-C film (Eastman Kodak). The 253 volunteer patients were examined with both techniques. The chest radiographs were interpreted by four radiologists. The study group was composed of 58 normal and 195 abnormal posteroanterior and lateral chest radiographs. In 31 cases there were two major radiologic diagnoses. The number of correct interpretations increased when the SER images were examined, compared with the conventional Ortho-G (chi 2 = 4.17, P less than .05) and conventional Ortho-C (chi 2 = 16.9, P less than .001) radiographs. The overall accuracy of disease detection improved for all radiologists with the SER system. There was no disease category in which the accuracy of interpretation decreased when the SER system was used. The SER system is a clinically reliable method of improving image quality and increasing diagnostic accuracy.     

Computer-aided diagnosis for the detection and classification of lung cancers on chest radiographs ROC analysis of radiologists' performance

RATIONALE AND OBJECTIVES: The aim of the study is to investigate the effect of a computer-aided diagnostic (CAD) scheme on radiologist performance in the detection of lung cancers on chest radiographs. MATERIALS AND METHODS: We combined two independent CAD schemes for the detection and classification of lung nodules into one new CAD scheme by use of a database of 150 chest images, including 108 cases with solitary pulmonary nodules and 42 cases without nodules. For the observer study, we selected 48 chest images, including 24 lung cancers, 12 benign nodules, and 12 cases without nodules, from the database to investigate radiologist performance in the detection of lung cancers. Nine radiologists participated in a receiver operating characteristic (ROC) study in which cases were interpreted first without and then with computer output, which indicated locations of possible lung nodules, together with a five-color scale illustrating the computer-estimated likelihood of malignancy of the detected nodules. RESULTS: Performance of the CAD scheme indicated that sensitivity in detecting lung nodules was 80.6%, with 1.2 false-positive results per image, and sensitivity and specificity for classification of nodules by use of the same database for training and testing the CAD scheme were 87.7% and 66.7%, respectively. Average area under the ROC curve value for detection of lung cancers improved significantly (P = .008) from without (0.724) to with CAD (0.778). CONCLUSION: This type of CAD scheme, which includes two functions, namely detection and classification, can improve radiologist accuracy in the diagnosis of lung cancer.     

The diagnostic performance of a PC-based teleradiology link.

AIM: Evaluation of the diagnostic performance of a personal computer based teleradiology link. MATERIALS AND METHODS: Two experienced radiologists assessed 100 cases, all based on chest and skeletal films using teleradiology for 50. These assessments were compared with the consensus of a panel of three independent radiologists. RESULTS: Diagnostic performance of teleradiology and conventional film was similar (sensitivity 88 vs. 90%; specificity 96 vs. 90%; accuracy 91 vs. 90%; not significant). However, the quality of teleradiology images was rated poorer, and the confidence in diagnosis was lower with teleradiology. ROC curve analysis, taking into account diagnostic confidence, showed significantly poorer performance for teleradiology at all thresholds when chest X-rays only were considered. There was no significant difference for skeletal images, although the two smooth curves crossed, suggesting teleradiology might be better when the specificity is high. CONCLUSION: These findings suggest that when this type of teleradiology system is used, the value of rapid reporting must be balanced against poorer image quality, particularly for chest X-rays.     

Detectability of lung nodules using flat panel detector with dual energy subtraction by two shot method: evaluation by ROC method

The aim of this study was to evaluate the effectiveness of dual-exposure dual energy subtraction technique in flat-panel chest radiography for lung nodules detection. Chest radiographs were acquired in 100 patients (57 men and 43 women; mean age, 60.2 years; range, 18-89 years) using a flat-panel digital chest system. These images were evaluated by seven radiologists. A continuous rating scale of 0-100 was used to represent each observer's confidence level regarding the presence or absence of lung nodules. Observer performance for detection of lung nodules with subtraction images was tested by using receiver operating characteristic (ROC) analysis of individual and averaged reader data. The average area under the ROC curve (Az value) significantly increased with subtraction images (Az=0.79 in standard radiographs versus Az=0.84 with subtraction images, p<0.05). In conclusion, the two-exposure dual-energy subtraction chest radiography significantly would improve detection of lung nodules.     

Improved radiology film library operations.

An interdisciplinary task force at the authors' institution developed a design for a new film library that placed heavy emphasis on service to its customers. The task force established locations and staffing for "satellite" inpatient film libraries organized by clinical specialties. Inpatient films were not permitted to be taken from the radiology department but were available for 24-hour viewing. Film locations were recorded in the radiology information system. Reorganization led to important improvements in film library operations and in the clinical staff's opinion of film library service. Objective measures of performance, such as the fraction of requested films available for conferences, showed significant improvement (P less than .001). Closer working relationships developed between film librarians, radiologists, and their clinical colleagues. Film library personnel were recruited more easily and stayed on the job longer. Several hundred thousand film-tracking transactions per year were recorded. The authors conclude that decentralization of certain film library activities improved many critical aspects of performance.     

Comparison of 2048-line digital display formats and conventional radiographs: an ROC study

Observer performance tests were conducted to compare the effects on diagnostic accuracy of digital hard copy and video display formats versus conventional radiographic film. Digital images were obtained by digitizing conventional chest radiographs to a 2048 x 2048 matrix with a laser film scanner. Three digital display formats were used: laser-printed digital film, a 2048-line video monitor without user interaction, and a 2048-line video monitor with user interaction. Thirty-one posteroanterior chest radiographs, determined by consensus of four thoracic radiologists to contain septal lines (n = 11), parenchymal nodules (n = 7), nodules and septal lines (n = 7), or neither abnormality (n = 6), were used for the study. Images were interpreted by four radiologists in four separate viewing sessions. Diagnostic accuracy was determined by receiver-operating characteristic analysis for each observer with each viewing technique. No statistical differences in diagnostic accuracy, determined by the area under the receiver-operating-characteristic curve, were found between the analog film, the digital film, and the two video digital display formats. This preliminary study suggests that 2048-line digital displays may be an acceptable alternative to the traditional lightbox viewing method for the perception of these two abnormalities commonly seen on chest radiographs.