C I Sánchez’s research while affiliated with Radboud University Medical Centre (Radboudumc) and other places

BACKGROUND: DetecTB (Diagnostic Enhanced Tools for Extra Cases of TB), an intensified tuberculosis (TB) case-finding programme targeting prisons and high-risk communities was implemented on Palawan Island, the Philippines.OBJECTIVE: To evaluate the performance of TB detection based on computerised chest radiography (CXR) readings.DESIGN: Data from 14 094 subjects were analysed from September 2012 to June 2014. All CXRs were read by a physician and by software. Individuals with TB symptoms or CXR abnormalities according to the physician underwent Xpert® MTB/RIF testing, the remaining persons were considered TB-negative (screening reference). A subset of 200 CXRs was read by an independent human reader (radiological reference). This reader also re-read a subset of the most abnormal cases as identified using the software but read as normal by the physician (discordant cases).RESULTS: A total of 10 755 individuals were included in the analysis, 2534 of whom had a positively assessed CXR; 298 cases were Xpert-positive. Using the screening reference, the area under the receiver operating characteristic curve for software readings was 0.93 (95%CI 0.92-0.94), with a sensitivity of 0.98 (95%CI 0.97-0.99) and a specificity of 0.69 (95%CI 0.40-0.98). Based on the radiological reference, the physician performed slightly worse than the software (sensitivity, 0.82, 95%CI 0.74-0.89 and specificity, 0.87, 95%CI 0.81-0.96 vs. sensitivity, 0.83, 95%CI 0.71-0.93 and specificity, 0.87, 95%CI 0.75-0.95), although this was not statistically significant. Of the 291 discordant cases, 70% were assessed as positive, resulting in a 22% increase in TB detection when extrapolated to the full cohort.CONCLUSION: The performance of automated CXR reading is comparable to that of the attending physicians in DetecTB, and its use as a second reader could increase TB detection.

Setting: Tuberculosis (TB) screening programmes can be optimised by reducing the number of chest radiographs (CXRs) requiring interpretation by human experts. Objective: To evaluate the performance of computerised detection software in triaging CXRs in a high-throughput digital mobile TB screening programme. Design: A retrospective evaluation of the software was performed on a database of 38 961 postero-anterior CXRs from unique individuals seen between 2005 and 2010, 87 of whom were diagnosed with TB. The software generated a TB likelihood score for each CXR. This score was compared with a reference standard for notified active pulmonary TB using receiver operating characteristic (ROC) curve and localisation ROC (LROC) curve analyses. Results: On ROC curve analysis, software specificity was 55.71% (95%CI 55.21-56.20) and negative predictive value was 99.98% (95%CI 99.95-99.99), at a sensitivity of 95%. The area under the ROC curve was 0.90 (95%CI 0.86-0.93). Results of the LROC curve analysis were similar. Conclusion: The software could identify more than half of the normal images in a TB screening setting while maintaining high sensitivity, and may therefore be used for triage.

Molecular tests hold great potential for tuberculosis (TB) diagnosis, but are costly, time consuming, and HIV-infected patients are often sputum scarce. Therefore, alternative approaches are needed. We evaluated automated digital chest radiography (ACR) as a rapid and cheap pre-screen test prior to Xpert MTB/RIF (Xpert). 388 suspected TB subjects underwent chest radiography, Xpert and sputum culture testing. Radiographs were analysed by computer software (CAD4TB) and specialist readers, and abnormality scores were allocated. A triage algorithm was simulated in which subjects with a score above a threshold underwent Xpert. We computed sensitivity, specificity, cost per screened subject (CSS), cost per notified TB case (CNTBC) and throughput for different diagnostic thresholds. 18.3% of subjects had culture positive TB. For Xpert alone, sensitivity was 78.9%, specificity 98.1%, CSS $13.09 and CNTBC$90.70. In a pre-screening setting where 40% of subjects would undergo Xpert, CSS decreased to $6.72 and CNTBC to$54.34, with eight TB cases missed and throughput increased from 45 to 113 patients/day. Specialists, on average, read 57% of radiographs as abnormal, reducing CSS ($8.95) and CNTBC ($64.84). ACR pre-screening could substantially reduce costs, and increase daily throughput with few TB cases missed. These data inform public health policy in resource-constrained settings.

Detection of tuberculosis (TB) on chest radiographs (CXRs) is a hard problem. Therefore, to help radiologists or even take their place when they are not available, computer-aided detection (CAD) systems are being developed. In order to reach a performance comparable to that of human experts, the pattern recognition algorithms of these systems are typically trained on large CXR databases that have been manually annotated to indicate the abnormal lung regions. However, manually outlining those regions constitutes a time-consuming process that, besides, is prone to inconsistencies and errors introduced by interobserver variability and the absence of an external reference standard. In this paper, we investigate an alternative pattern classi cation method, namely multiple-instance learning (MIL), that does not require such detailed information for a CAD system to be trained. We have applied this alternative approach to a CAD system aimed at detecting textural lesions associated with TB. Only the case (or image) condition (normal or abnormal) was provided in the training stage. We compared the resulting performance with those achieved by several variations of a conventional system trained with detailed annotations. A database of 917 CXRs was constructed for experimentation. It was divided into two roughly equal parts that were used as training and test sets. The area under the receiver operating characteristic curve was utilized as a performance measure. Our experiments show that, by applying the investigated MIL approach, comparable results as with the aforementioned conventional systems are obtained in most cases, without requiring condition information at the lesion level.

Population screening for sight threatening diseases based on fundus imaging is in place or being considered worldwide. Most existing programs are focussed on a specific disease and are based on manual reading of images, though automated image analysis based solutions are being developed. Exudates and drusen are bright lesions which indicate very different diseases, but can appear to be similar. Discriminating between them is of interest to increase screening performance. In this paper, we present a Bag of Words approach which can be used to design a system that can play the dual role of content based retrieval (of images with exudates or drusen) system and a decision support system to address the problem of bright lesion discrimination. The approach consists of a novel partitioning of an image into patches from which color, texture, edge and granulometry based features are extracted to build a dictionary. A bag of Words approach is then employed to help retrieve images matching a query image as well as derive a decision on the type of bright lesion in the given (query) image. This approach has been implemented and tested on a combination of public and local dataset of 415 images. The area under the curve for image classification is 0.90 and retrieved precision is 0.76.

We describe an interactive system for classification of normal and seven types of abnormal lung tissue in CT scans from interstitial lung disease patients, using training data from previously annotated scans and annotations by the observer in the scan under investigation. We compared seven different interactive annotation strategies using different combinations of both types of training data, in order to minimize user effort in the interactive annotation process. The lungs in all scans were divided into roughly spherical volumes of interest (VOIs). An observer labeled all VOIs in 21 thoracic CT scans. Leave-one-scan-out experiments that simulated slice-by-slice interactive annotation sessions were performed. The best results were obtained with a strategy in which the simulated user decides for each slice whether to use a classifier trained on pooled data from prior scans or a classifier trained on data from the current scan. In this approach, the labels of 88% of all VOIs were predicted correctly, meaning that only 12% of all labels needed to be changed by the simulated user.

The performance of computer-aided diagnosis (CAD) systems can be highly influenced by the training strategy. CAD systems are traditionally trained using available labeled data, extracted from a specific data distribution or from public databases. Due to the wide variability of medical data, these databases might not be representative enough when the CAD system is applied to data extracted from a different clinical setting, diminishing the performance or requiring more labeled samples in order to get better data generalization. In this work, we propose the incorporation of an active learning approach in the training phase of CAD systems for reducing the number of required training samples while maximizing the system performance. The benefit of this approach has been evaluated using a specific CAD system for Diabetic Retinopathy screening. The results show that (1) using a training set obtained from a different data source results in a considerable reduction of the CAD performance; and (2) using active learning the selected training set can be reduced from 1000 to 200 samples while maintaining an area under the Receiver Operating Characteristic curve of 0.856.

The performance of computer-aided diagnosis (CAD) systems can be highly influenced by the training strategy. CAD systems are traditionally trained using available labeled data, extracted from a specific data distribution or from public databases. Due to the wide variability of medical data, these databases might not be representative enough when the CAD system is applied to data extracted from a different clinical setting, diminishing the performance or requiring more labeled samples in order to get better data generalization. In this work, we propose the incorporation of an active learning approach in the training phase of CAD systems for reducing the number of required training samples while maximizing the system performance. The benefit of this approach has been evaluated using a specific CAD system for Diabetic Retinopathy screening. The results show that 1) using a training set obtained from a different data source results in a considerable reduction of the CAD performance; and 2) using active learning the selected training set can be reduced from 1000 to 200 samples while maintaining an area under the Receiver Operating Characteristic curve of 0.856.