CDS tool improves reader diagnosis of bladder cancer response to treatment

A decision support tool can help physicians better diagnose bladder cancer treatment response on CT, according to an Nov. 10 study published in Academic Radiology.

Currently, there are no dependable methods for determining complete response to neoadjuvant chemotherapy, according to lead author, Kenny H. Cha, PhD, and colleagues. As a consequence, some suffer negative reactions, with little benefits. A reliable method could be used to personalize therapy plans and noninvasively select patients eligible for bladder-saving therapy, they added.

Cha et al. created a CT-based computerized decision-support system for muscle-invasive bladder cancer treatment response assessment (CDSS-T), which utilizes deep learning convolutional neural networks and radiomics to estimate which patients completely respond to neoadjuvant chemotherapy.

Researchers collected pre-  and post-chemotherapy CT scans of 123 patients with 157 muscle-invasive bladder cancer foci. Five attending abdominal radiologists, four diagnostic radiology residents, two attending oncologists and one attending urologist analyzed the likelihood of complete response (T0 disease).

Overall, physicians’ diagnoses were more accurate and less variable when aided with the CDSS-T. The average area under the curve (AUC) for CDSS-T alone was 0.80. In physicians not using CDSS-T that number was 0.74, and it was 0.77 for physicians utilizing CDSS-T—a “statistically significant” difference, the authors noted.

“To our knowledge, this is the first study to perform an observer study using a CAD system for this purpose,” Cha et al. wrote. “We observed statistically significant improvement in physicians' performance when blinded observers were provided the CDSS-T results.”

In some cases, the readers correctly identified cancers the CDSS-T did not. Cha and colleagues noted in many of those cases, the observers were unaffected and stood by their initial decisions. Comparatively, in cases wrongly classified by readers, but correctly diagnosed by decision-support, the CDSS-T “often” persuaded observers to change their finding to the correct diagnosis.

The team noted that their study could have been validated in a larger independent dataset, but believe their results can still positively impact patient care.

“Further improvement in the performance of CDSS-T is desirable, and a large-scale observer study should be conducted in an independent case set to validate the impact of the CDSS-T on clinical decision-making,” the authors concluded. “The results of this study might be useful for better selection of patients considering bladder-sparing therapy for muscle-invasive bladder cancer.”