Steven M. Montner

IMAGE FEATURE ANALYSIS OF FALSE-POSITIVE DIAGNOSES PRODUCED BY AUTOMATED DETECTION OF LUNG NODULES

RATIONALE AND OBJECTIVES To reduce the number of false-negative diagnoses by radiologists, the authors are developing a computer-aided diagnosis scheme for detection of lung nodules in digital chest images. In this study, the authors attempted to reduce the number of false-positive diagnoses obtained with a previous computer scheme by incorporating additional knowledge from experienced chest radiologists into the computer scheme. METHODS The authors applied their previous computer scheme, using less-strict criteria, to 60 clinical chest radiographs; this yielded 735 candidate nodules (23 true nodules and 712 false-positive diagnoses). These candidates were analyzed using region-growing, trend-correction, and edge-gradient techniques to determine measures by which to quantify image features of candidate nodules. RESULTS The 712 false-positive diagnoses represented various anatomic structures that were located throughout the chest image. From this analysis, we were able to decrease the number of false-positive errors from an average of 12 to approximately 5 per image without eliminating any true nodules. CONCLUSION Our results show that incorporating knowledge from experienced chest radiologists into the computer algorithm will play an important role in the development of computerized schemes for the detection of pulmonary nodules.

Characterisation of patients with interstitial pneumonia with autoimmune features.

Patients with interstitial lung disease (ILD) may have features of connective tissue disease (CTD), but lack findings diagnostic of a specific CTD. A recent European Respiratory Society/American Thoracic Society research statement proposed criteria for patients with interstitial pneumonia with autoimmune features (IPAF). We applied IPAF criteria to patients with idiopathic interstitial pneumonia and undifferentiated CTD-ILD (UCTD). We then characterised the clinical, serological and morphological features of the IPAF cohort, compared outcomes to other ILD cohorts and validated individual IPAF domains using survival as an endpoint. Of 422 patients, 144 met IPAF criteria. Mean age was 63.2 years with a slight female predominance. IPAF cohort survival was marginally better than patients with idiopathic pulmonary fibrosis, but worse than CTD-ILD. A non-usual interstitial pneumonia pattern was associated with improved survival, as was presence of the clinical domain. A modified IPAF cohort of those meeting the clinical domain and a radiographic or histological feature within the morphological domain displayed survival similar to those with CTD-ILD. IPAF is common among patients with idiopathic interstitial pneumonia and UCTD. Specific IPAF features can identify subgroups with differential survival. Further research is needed to replicate these findings and determine whether patients meeting IPAF criteria benefit from immunosuppressive therapy. IPAF is common among patients with IIP and has distinct subgroups that demonstrate differential survival http://ow.ly/Z0ShD

Potential usefulness of computerized nodule detection in screening programs for lung cancer.

RATIONALE AND OBJECTIVE To alert radiologists to possible nodule locations and subsequently to reduce the number of false-negative diagnoses, the authors are developing a computer-aided diagnostic (CAD) scheme for the detection of lung nodules in digital chest images. METHODS A computer-vision scheme was applied to photofluorographic films obtained in a mass survey for detection of asymptomatic lung cancer in Japan. Ninety-five patients with abnormal test results who had primary and metastatic lung cancers and 103 patients with normal test results were included. RESULTS The sensitivity of the computer output was comparable with that of physicians in this mass survey (62%). The computer detected approximately 40% of all nodules missed in the mass survey, but missed 17 true-positive results identified in the mass survey. The CAD scheme produced an average of 15 false-positive findings per image. CONCLUSION If the number of false-positive results can be significantly reduced, computer-vision schemes such as this may have a role in lung cancer screening programs.

COMPUTERIZED SCHEME FOR THE DETECTION OF PULMONARY NODULES : A NONLINEAR FILTERING TECHNIQUE

To aid radiologists in the detection of lung cancer, the authors are developing a computer-aided diagnosis system that locates areas suspicious for nodules in digital chest radiographs. The system involves a difference-image approach and various feature-extraction techniques. The authors describe nonlinear filters used in the difference-image approach. A morphological open operation and a ring-shaped median filter are applied in the difference-image step for signal enhancement and signal suppression, respectively. Using 60 clinical chest radiographs, the nonlinear filtering method detected approximately 63% of actual nodules with approximately 19 false-positive results per image. The locations of the false-positive detections, however, usually did not coincide with those from the linear filtering method. Thus, by using a combination of the detections from the two methods, the false-positive rate was reduced to two to three per image at a sensitivity of 60%.

Effect of heart-size parameters computed from digital chest radiographs on detection of cardiomegaly. Potential usefulness for computer-aided diagnosis.

Heart size is an important and useful diagnostic parameter in chest radiographs. However, there is a large variation in the subjective judgment of cardiac enlargement (cardiomegaly). To reduce this subjective element, the authors are developing an automated system for quantitative analysis of heart size in digital chest radiographs. Four hundred chest radiographs were reviewed initially by two radiologists and were classified into two groups: those with and those without cardiomegaly. Another radiologist reviewed 47 images which were not classified consistently in the initial review. The authors used these radiographs to construct a database for determination of cardiomegaly. Numerous parameters related to heart size were obtained in a semiautomated analysis of these radiographs, and the use of each parameter for detection of cardiomegaly was evaluated by means of receiver operating characteristic analysis. The authors also examined whether the accuracy would be improved when they applied multivariate analysis to a pair of parameters. From the analyses of the individual parameters, the automatically determined cardiothoracic ratio was found to be the single most effective measure for detecting cardiomegaly in chest radiographs. However, multivariate analysis provided results superior to those with an individual parameter.

Development of a high quality film duplication system using a laser digitizer: Comparison with computed radiography

A high quality film-duplication system was developed in order to improve the image quality of duplicated radiographs and to recover improperly exposed films. The system consists of a laser film digitizer, a laser film printer, a workstation, and a magneto-optical disk. Radiographs are digitized by the laser digitizer, processed by the computer for image enhancement, and then printed on a film by the laser printer. A nonlinear density-correction technique is employed in recovering improperly exposed radiographs using the H&D curve of the screen-film system. Using the new duplication system in our department, the average recovery rate was over 80% for chest and abdominal films rejected due to over- or underexposed. The basic imaging properties of the duplication system were compared with those of a Computed Radiography (CR) system and a conventional screen-film system. For low spatial frequencies, the MTF of the CR system is superior to that of the digital duplication system; however, for high spatial frequencies, the MTF of the duplication system is superior. The noise in the duplication system is about half of that in the CR system.

The nature and subtlety of abnormal findings in chest radiographs.

All detected abnormal findings were recorded for 1085 consecutive chest x-ray examinations. Each finding was classified by descriptive criteria and graded for subtlety. Accuracy of interpretation was evaluated in a randomized sample of 100 cases using follow-up or multiple readers. Seventy percent of standard examinations and 93% of bedside examinations revealed abnormal findings. Pulmonary infiltrates were the most commonly detected abnormality, being present in 55% of abnormal cases. Noncalcified pulmonary nodules and pneumothoraces were each present in approximately 5% of abnormal cases. The most commonly encountered subtle findings were due to intravenous catheters, pulmonary infiltrates, pneumothoraces, rib lesions, and pulmonary nodules in descending order of frequency. It is concluded that it is reasonable to use selected examples of these findings in observer tests when evaluating new imaging modalities such as digital radiography.

Neural network approach for differential diagnosis of interstitial lung diseases

A neural network approach was applied for the differential diagnosis of interstitial lung diseases. The neural network was designed for distinguishing between 9 types of interstitial lung diseases based on 20 items of clinical and radiographic information. A database for training and testing the neural network was created with 10 hypothetical cases for each of the 9 diseases. The performance of the neural network was evaluated by ROC analysis. The optimal parameters for the current neural network were determined by selecting those yielding the highest ROC curves. In this case the neural network consisted of one hidden layer including 6 units and was trained with 200 learning iterations. When the decision performances of the neural network chest radiologists and senior radiology residents were compared the neural network indicated high performance comparable to that of chest radiologists and superior to that of senior radiology residents. Our preliminary results suggested strongly that the neural network approach had potential utility in the computer-aided differential diagnosis of interstitial lung diseases. 1_© (1990) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

CT Findings, Radiologic-Pathologic Correlation, and Imaging Predictors of Survival for Patients With Interstitial Pneumonia With Autoimmune Features

OBJECTIVE The objective of this study is to determine the CT findings and patterns of interstitial pneumonia with autoimmune features (IPAF) and to assess whether imaging can predict survival for patients with IPAF. MATERIALS AND METHODS The study included 136 subjects who met the criteria for IPAF and had diagnostic-quality chest CT scans obtained from 2006 to 2015; a total of 74 of these subjects had pathologic samples available for review within 1 year of chest CT examination. CT findings and the presence of an usual interstitial pneumonitis (UIP) pattern of disease were assessed, as was the UIP pattern noted on pathologic analysis. Analysis of chest CT findings associated with survival was performed using standard univariate and multivariate Cox proportional hazards methods as well as the unadjusted log-rank test. Survival data were visually presented using the Kaplan-Meier survival curve estimator. RESULTS Most subjects with IPAF (57.4%; 78/136) had a high-confidence diagnosis of a UIP pattern on CT. Substantially fewer subjects (28.7%; 39/136) had a pattern that was inconsistent with UIP noted on CT. The presence of a UIP pattern on CT was associated with smoking (p < 0.01), male sex (p < 0.01), and older age (p < 0.001). Approximately one-fourth of the subjects had a nonspecific interstitial pneumonitis pattern on CT. Of interest, nearly one-tenth of the subjects had a CT pattern that was most consistent with hypersensitivity pneumonitis rather than the customary CT patterns ascribed to lung disease resulting from connective tissue disease. Most subjects with a possible UIP pattern on CT (83.3%) had UIP diagnosed on the basis of pathologic findings. Focused multivariate analysis showed that honeycombing on CT (hazard ratio, 2.17; 95% CI, 1.05-4.47) and pulmonary artery enlargement on CT (hazard ratio, 2.08; 95% CI, 1.02-4.20) were independent predictors of survival. CONCLUSION IPAF most often presents with a UIP pattern on CT and is associated with worse survival when concomitant honeycombing or pulmonary artery enlargement is present.

Interstitial Pneumonia With Autoimmune Features: Value of Histopathology

CONTEXT - Patients with idiopathic interstitial pneumonia may display evidence of autoimmunity without meeting criteria for a defined connective tissue disease. A recent European Respiratory Society/American Thoracic Society statement proposed research criteria for interstitial pneumonia with autoimmune features (IPAF), which includes findings from the clinical, serologic, and morphologic domains. OBJECTIVES - To investigate the importance of histopathologic criteria within the morphologic domain and to report our methodology for identifying these features. DESIGN - Patients with idiopathic interstitial pneumonia at the University of Chicago who underwent surgical lung biopsy or lung transplantation were assessed for IPAF histopathologic features, using the initial pathology interpretation in the electronic records. A focused rereview of available slides by a pulmonary pathologist was then performed for patients who failed to meet IPAF criteria on initial pathology assessment. RESULTS - Of 422 patients with idiopathic interstitial pneumonia, 176 (41.7%) underwent surgical lung biopsy or lung transplant. Forty-six of those 176 patients (26.1%) met IPAF criteria by initial pathology interpretation and a positive clinical or serologic feature. Of the remaining 130 patients, 73 (56.2%) met either the clinical or serologic domains without meeting the morphologic domain, whereas 36 (27.7%) had slides available for pathology rereview. This rereview demonstrated nonspecific interstitial pneumonia in 8 of 36 patients (22.2%) and lymphoplasmacytic infiltrates in 6 of 36 patients (16.7%), resulting in an additional 7 of 36 patients (19.4%) with idiopathic interstitial pneumonia that met the IPAF criteria. In IPAF, pulmonary vasculopathy was the most prevalent finding (45 of 84; 53.6%) and predicted increased mortality (hazard ratio, 2.5; P = .04). CONCLUSIONS - Using a methodological approach to identifying IPAF pathology, we demonstrate a significant increase in the number of patients meeting IPAF criteria because of focused pathologic review and highlight the prognostic value of the IPAF pathologic findings.