The purpose of this study was to develop a computer-aided diagnosis (CAD) system for the classification of malignant and benign masses in the breast using ultrasonography based on a convolutional neural network (CNN), a state-of-the-art deep learning technique. We explored the regions for the correct classification by generating a heat map that presented the important regions used by the CNN for human malignancy/benign classification. Clinical data was obtained from a large-scale clinical trial previously conducted by the Japan Association of Breast and Thyroid Sonology. Images of 1536 breast masses (897 malignant and 639 benign) confirmed by pathological examinations were collected, with each breast mass captured from various angles using an ultrasound (US) imaging probe. We constructed an ensemble network by combining two CNN models (VGG19 and ResNet152) fine-tuned on balanced training data with augmentation and used the mass-level classification method to enable the CNN to classify a given mass using all views. For an independent test set consisting of 154 masses (77 malignant and 77 benign), our network showed outstanding classification performance with a sensitivity of 90.9% (95% confidence interval 84.5-97.3), a specificity of 87.0% (79.5-94.5), and area under the curve (AUC) of 0.951 (0.916-0.987) compared to that of the two CNN models. In addition, our study indicated that the breast masses themselves were not detected by the CNN as important regions for correct mass classification. Collectively, this CNN-based CAD system is expected to assist doctors by improving the diagnosis of breast cancer in clinical practice.

As bortezomib therapy for SLE is associated with many adverse reactions, treatment indications should be selected carefully, and protocols should aim to prevent these occurrences. Although the change in anti-dsDNA antibody titer did not support the efficacy of bortezomib as a treatment for SLE, high SRI in the treatment group suggests bortezomib may utilize mechanisms other than inhibition of anti-dsDNA antibody production.

Background/Aims: Traditional on-site monitoring of clinical trials via frequent site visits and 100% source data verification is cost-consuming, and it still cannot guarantee data quality effectively. Depending on the types and designs of clinical trials, an alternative would be combining several monitoring methods, such as risk-based monitoring and remote monitoring. However, there is insufficient evidence of its effectiveness. This research compared the effectiveness of risk-based monitoring with a remote monitoring system with that of traditional on-site monitoring. Methods: With a cloud-based remote monitoring system called beagle View®, we created a remote risk-based monitoring methodology that focused only on critical data and processes. We selected a randomized controlled trial conducted at Tohoku University Hospital and randomly sampled 11 subjects whose case report forms had already been reviewed by data managers. Critical data and processes were verified retrospectively by remote risk-based monitoring; later, all data and processes were confirmed by on-site monitoring. We compared the ability of remote risk-based monitoring to detect critical data and process errors with that of on-site monitoring with 100% source data verification, including an examination of clinical trial staff workload and potential cost savings. Results: Of the total data points (n = 5617), 19.7% (n = 1105, 95% confidence interval = 18.7–20.7) were identified as critical. The error rates of critical data detected by on-site monitoring, remote risk-based monitoring, and data review by data managers were 7.6% (n = 84, 95% CI = 6.2–9.3), 7.6% (n = 84, 95% confidence interval = 6.2–9.3), and 3.9% (n = 43, 95% confidence interval = 2.9–5.2), respectively. The total number of critical process errors detected by on-site monitoring was 14. Of these 14, 92.9% (n = 13, 95% confidence interval = 68.5–98.7) and 42.9% (n = 6, 95% confidence interval = 21.4–67.4) of critical process errors were detected by remote risk-based monitoring and data review by data managers, respectively. The mean time clinical trial staff spent dealing with remote risk-based monitoring was 9.9 ± 5.3 (mean ± SD) min per visit per subject. Our calculations show that remote risk-based monitoring saved between 9 and 41 on-site monitoring visits, corresponding to a cost of between US$13,500 and US$61,500 per trial site. Conclusion: Remote risk-based monitoring was able to detect critical data and process errors as well as on-site monitoring with 100% source data verification, saving travel time and monitoring costs. Remote risk-based monitoring offers an effective alternative to traditional on-site monitoring of clinical trials.

No abstract