AI System Helps Assess Bladder Cancer Treatment Response

An artificial intelligence (AI)-driven system assisted in evaluating bladder cancer patient responses to chemotherapy, a University of Michigan Health Lab study found.

When a patient develops bladder cancer, providers often remove the organ to prevent the cancer from spreading or regrowing. But research shows that this procedure may not be necessary depending on how the patient responds to chemotherapy.

Providers also often have trouble distinguishing between the remnants of cancer and scarred tissue.

The study, published in the journal Tomography, consisted of 14 physicians from various backgrounds who evaluated 157 bladder tumor scans before and after treatment.

They gave ratings for three separate measures that assessed the response to chemotherapy and recommendation for the next treatment. They then examined ratings provided by the AI system.  

Lower scores meant that the complete response to chemotherapy was less likely, and higher scores meant more likely.

After comparing different results, providers noticed that their evaluation improved significantly following the implementation of the AI system. The AI system also helped the fellows and the medical student who participated in the study gain experience.

Using the AI system to assist in treatment planning is possible, but it cannot act as a replacement, researchers said.

“One interesting thing that we figured out is that the computer makes mistakes on a different subset of cases than a radiologist would,” said Lubomir Hadjiyski, PhD., a professor of radiology at the University of Michigan Medical School, in the press release. “Which means that if the tool is used correctly, it gives a chance to improve but not replace the physician’s judgment.”

AI is increasingly being used in the identification and diagnosis of chronic conditions.

A study from August 2020 described how the implementation of an AI-enhanced electrocardiogram (ECG) could help detect heart failure at an advanced rate. Researchers trained the AI-enabled ECG by training it on thousands of cases involving left ventricular systolic dysfunction (LSVD). Researchers found the AI system was more proficient in identifying LSVD than standard blood tests.

Another study from March 2021 described how EHR data could be mined with the help of an AI algorithm. The algorithm uses the data to suggest optimal diagnostic approaches. Developed by a team from the USC Viterbi School of Engineering, researchers intended for the algorithm to have a similar mindset to that of a doctor.

A recent study from earlier this month also described how AI could assist in the diagnosis of heart disease in ultrasounds. Researchers started by exposing the AI model to thousands of ultrasound images, teaching it how to distinguish between healthy and unhealthy images. Researchers found that the use of the model improved their diagnosis practices.