Artificial Intelligence enhances treatment assessment in bladder cancer

A computerized artificial intelligence (AI)-based decision support system (CDSS-T) enhanced the performance of clinicians when assessing patients’ response to chemotherapy prior to radical cystectomy, according to results from a small multi-institutional observer study published in Tomography.1,2

“The AI-based decision support system has the potential to improve the diagnostic accuracy in assessing bladder cancer treatment response and result in more consistent performance among all physicians,” the authors wrote in their study conclusion.

Neoadjuvant chemotherapy is playing an evolving role in the bladder cancer treatment paradigm. Patients achieving a complete response (zero evidence of disease) to this upfront chemotherapy may be able to avoid radical cystectomy. An issue that emerges when exploring this path, however, is trying to decipher whether the remaining lesion after neoadjuvant treatment is cancer or just necrotic or scarred tissue caused by the chemotherapy.1 The aim of the study was to determine whether CDSS-T could assist in these critical patient evaluations.

“The big question was when you have such an artificial device next to you, how is it going to affect the physician?” Lubomir Hadjiyski, PhD, a professor of radiology at the University of Michigan Medical School and the senior author of the study, stated in a news release. “Is it going to help? Is it going to confuse them? Is it going to raise their performance or will they simply ignore it?”

Overall, the study evaluated the diagnostic accuracy with and without CDSS-T of 17 observers who were from different institutions, specialties, and levels of experience. There was 1 urologist, 5 oncologists, 5 abdominal radiologists, 4 diagnostic radiology residents, 1 medical student, 1 neurology fellow.

The providers were tasked with evaluating response to neoadjuvant chemotherapy in 123 bladder cancer patients who had 157 pre- and post-treatment cancer pairs. The mean age was 63 years for both males (range, 43-84 years) and females (range 37-82 years). The pre- and post-treatment cancer stages for the 157 cancer pairs were as follows: T0 (0 pre-treatment, 40 post-treatment); T1 (8, 37); T2 (76, 23); T3 (63, 38); and T4 (10, 19).

For the study, the providers were asked to give “ratings for 3 measures that assessed the level of response to chemotherapy as well as a recommendation for the next treatment to be done for each patient (radiation or surgery),” according to the authors. The providers then looked at a score that the AI system calculated and were given the option to maintain or revise their ratings. To determine the accuracy of the providers and the impact of CDSS-T, the study investigators compared the final provider ratings against tumor samples obtained during the radical cystectomies.

The results showed that, when aided by the CDSS-T, there was a significant improvement in the average performance of the 17 observers (P = .002). The benefit of the AI system in improving assessments was observed across different experience levels and specialties. The benefit was greatest among the providers with less experience, who with the help of the AI system made comparable diagnoses as the participants with greater experience.

Hadjiyski noted in the press release that despite the impressive outcomes of the study, his years of experience researching AI systems have shown him that these tools should complement the assessment of physicians, not replace it.

“One interesting thing that we figured out is that the computer makes mistakes on a different subset of cases than a radiologist would,” Hadjiyski stated. “Which means that if the tool is used correctly, it gives a chance to improve but not replace the physician’s judgment.”

References

1. Artificial intelligence helps physicians better assess the effectiveness of bladder cancer treatment. Published online April 22, 2022. Accessed April 22, 2022. https://labblog.uofmhealth.org/health-tech/artificial-intelligence-helps-physicians-better-assess-effectiveness-of-bladder-cancer

2. Sun D, Hadjiiski L, Alva A, et al. Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study. Tomography. 2022;8(2):644-656. doi: 10.3390/tomography8020054.