This is a small multi-site study that validates the effectiveness of an artificial intelligence (AI)-based system to improve the clinical assessment of the response of patients treated with chemotherapy for bladder cancer, before a radical cystectomy (surgery to remove the bladder). “If you use this model intelligently, it will be of great help to you,” writes the lead author, Dr. Lubomir Hadjiyski, professor of radiology at the University of Michigan.
AI does not replace but complements human expertise
When patients develop bladder cancer, surgeons often remove the entire bladder to prevent the cancer from coming back or spreading to other organs or areas. However, in recent years, new evidence has suggested that surgery may not be necessary if a patient no longer has any signs of disease after chemotherapy. This is when AI can be of great help.
AI for post-chemotherapy analysis: it is difficult to determine if the lesion left after treatment is necrotic tissue or if the cancer persists. The researchers wondered if AI could help, without “bypassing” the doctor.
The study invited 14 physicians from different specialties (radiology, urology, oncology, medical student) to review the before and after treatment scans of 157 bladder tumors. Participants rated 3 measures of the level of response to chemotherapy and made their recommendation for the next treatment step for each patient in the study (radiotherapy or surgery).
Secondly, the doctors had access to the score calculated by the computer. Lower scores indicated a lower likelihood of complete response to chemo and vice versa for higher scores. Physicians then had the option of revising their ratings or leaving them unchanged.
Whatever their specialties and levels of experience, the participating physicians agree on the ability to improve the assessment provided by the AI system. The less experienced among them find the greatest added value to the point that they “join”, with the help of the AI model, the same diagnostic qualities as the more experienced physician participants. The tool can also help less specialized healthcare professionals or those less focused solely on clinical care.
The team, which has been working for more than 20 years on the potential of AI in the evaluation of different types of cancer and patient responses to treatments, is now convinced of the usefulness of these machine learning tools. , but as a “second opinion” in addition to that of the doctor.
“The computer also makes mistakes like a radiologist would. Used correctly, it gives a chance to improve but does not replace the doctor’s judgement.