Alexandru Stancu

WIDEN: A tool for medical image management in multicenter clinical trials.

Background It has been proposed that in clinical trials in which the therapeutic strategy is driven by functional imaging, central review of the images should be done in real time. Purpose We report our experience with a new tool for image exchange and review, called Web-Based Imaging Diagnosis by Expert Network (WIDEN), which we implemented for the HD0607 prospective multicenter Italian clinical trial in which Hodgkin lymphoma treatment was adapted based on results of an interim positron emission tomography (PET) scan performed after the first two cycles of chemotherapy. Methods We used WIDEN for general management of the clinical trial, site imaging qualification, image exchange, workflow control, blinded independent central review, inter-observer variability assessment, consensus creation, audit, and statistical analysis. Results As of February 2013, the interim PET was available for 512 patients; upon central review, 103 of the scans were judged to be positive and 409 to be negative. The median scan uploading and downloading times were 1 min, 25 s and 1 min, 55 s, respectively; the average and median times for diagnosis exchange were 47 h, 53 min and 37 h, 43 min, respectively. The binary concordance between pairs of reviewers (Cohen’s kappa) ranged from 0.72 to 0.85. The 5-point scale concordance among all reviewers (Krippendorf’s alpha) was 0.77. Conclusions WIDEN proved to be an effective tool for medical imaging exchange and online review. Data security, simplicity, feasibility, and prompt scan review were demonstrated. Central reviews were completed promptly.

On-demand lung CT analysis with the M5L-CAD via the WIDEN front-end web interface and an OpenNebula-based cloud back-end

The development of algorithms for the analysis of medical images has been progressively growing over the past two decades. The most common approach is the deployment of standalone workstations, equipped with provider-dependent Graphic User Interfaces (GUI) from which the algorithm execution is triggered interactively. There are, however, several drawbacks: among them, the GUI development cost, the GUI learning curve for the users, the high fixed cost of the software licenses, the difficulty in upgrading the software release. For a few years, the hypothesis of using Grid Services has been explored by several research groups. It turned out that there were other drawbacks: the high costs and security risks of integrating computing resources of medical centers into a Grid Computing Infrastructure. The emerging of Cloud computing, accessible via secure Web protocols, solves most - if not all - the problems. In the specific case of lung Computer Assisted Detection, a further important reason favors the SaaS (Software as a Service) approach: it was demonstrated by several works that combining CAD algorithms improves the overall performance. The system we present is composed by three main building blocks: WIDEN (Web-based Image and Diagnosis Exchange Network) handles the workflow, the image upload and the CAD result notification; the OpenNebula-based cloud IaaS (Infrastructure as a Service) batch farm allocates virtual computing and storage resources; the M5L CAD provides the nodule detection functionality. Our proposed implementation securely handles sensitive patient data, since images are transferred with the HTTPS protocol and the underlying virtual batch farm is isolated. Moreover it is efficient since it dynamically backend.

A multi-thread WEB-based CADe system for nodule detection on chest multislice CT scans

Purpose To compare performance of radiologists and computer-assisted detection (CAD) program for detection of lung nodules (LN) on hiresolution CT images of the chest. Methods Hi-resolution CT images of the chest of 10 patients with suspected or known metastatic lung disease were analyzed (6 men, 4 women, aged from 24 to 73 years). Chest CT was performed on SOMATOM Definition AS scanner with capability of simultaneous acquisition of 40 slices. Images were reconstructed with slice thickness of 1.5 mm using hi-resolution kernel (B60). Two radiologists with chest CT reading experience of 5 and 14 years were asked to measure and record all visible LN looking through all slices of each patient. They worked independently of each other and were blinded of CAD results at the time of the first CT reading. Results of this independent reading were used to calculate detection sensitivity for each radiologist. At the second step the same images were read by two radiologists in consensus in order to determine true LN. LungCAD program which is the part of scanner software was used in completely automatic mode to detect LN at the final step. In order to calculate detection sensitivity lung lesions were considered to be true LN if one of the two criteria was satisfied: (1) LN detected by one radiologist should be confirmed by the other radiologist (consensus of two radiologists); (2) lung lesions which were detected by CAD program only were have to be confirmed by two radiologists in consensus. Results 144 true LN were detected in 10 patients (5–26 LN per patient). 88 true LN were detected by first radiologist, 92 by the second and 122 by CAD program (detection sensitivity 61, 64 and 85 % respectively). Jointly two radiologists detected 111 LN what was less compared to CAD program alone. 33 (23 %) LN were detected only by CAD program and 22 (15 %) by radiologists only. 20 LN detected by CAD program were not confirmed by radiologists and were considered to be false positives (0–6 per patient, 2 per patient in average). Reasons for false positive detections were peripheral vascular structures (50 %), pleural plaques (30 %), big vessels nearby mediastinum (20 %) and calcified hilar lymph nodes (20 %). LN below 3 mm size were difficult to detect both by humans and program. The biggest advantage was shown by CAD program in detection of lesions from 4 to 10 mm in size (Table 1). Conclusion CAD program has identified more true LN than each one or both of the two radiologists jointly. Number of false positives was on average of 2 per CT examination. Given the high sensitivity and relatively low time consumed CAD program may be recommended to improve detection of LN on hi-resolution CT images of the lungs.