UPV testing could lead to low-cost diagnosis of prostate cancer
VALENCIA (EFE). Research carried out as part of a doctoral dissertation at the Polytechnic University of Valencia (UPV) could reduce the number of unnecessary biopsies, improve the method of classifying prostate cancer and begin the process of introducing a low-cost diagnostic system in ambulatory surgery centers.
Results of the study conducted Juan Bautista Talens and coordinated by professor of the Department of Electronic Engineering UPV Jose Pelegriwas conducted in collaboration with the UPV Research Institute for Integrated Coastal Zone Management and the La Fe Hospital Research Institute and was published by BMC Medical Informatics and Decision Making.
“The intention is to develop a device that could be kept in a room and that, in a few minutes, would try to help doctors perform an analysis or better classify a disease,” explains Talens, author of the dissertation “Analysis.” , development and implementation of non-invasive tools for detecting prostate and bladder cancer using MOS technology.
Talens argues that there is a problem that many negative biopsies are done because the technique is haphazard, attempting to extract tissue from the prostate without being sure that it is actually a biopsy. taking cancerous tissue, which leads to many negative biopsies and therefore high costs.
Additionally, he adds, these are invasive tests because they must pass through the rectum to remove pieces of the prostate.
The solution was obtained as a result of research by Talens and Pelegri together with Jose Luis Ruiz And Thomas Sogorbconsists of extracting a series of signals through smell.
“We stimulate the passage of volatile compounds released from urine and pass them through a chamber with gas sensors, from which we obtain a series of data, based on which we train artificial intelligence so that it can distinguish patients with cancer from those with a benign prostate tumor.” . hyperplasia, another disease that produces similar results in the tests used to decide whether to send a patient for a biopsy, Talens points out.
He clarifies that they used an electronic nose they had, a device too large to take with them to a consultation so that they could be identified and classified.
Subsequently, they developed an entire engineering system based on software, database and electronic device, designed to be taken with you to a consultation.
Ultimately, the researcher notes, “we have created a prototype of MOOSY4 that, at a sufficiently advanced stage, can be useful in clinical consultations.”
The model, which initially managed to correctly classify 87% of patient samples, indicating significant predictive power, was subsequently improved.
“We found areas for improvement, especially in identifying prostate cancer cases with elevated prostate antigen (false negatives),” says Pelegri.
To address this limitation, an AI training strategy was proposed that gave more weight to the cancer class compared to benign hyperplasia, and after implementing this adjustment, the results indicate an improvement in the model’s ability to handle more complex cases.
This research has several directions for future development.“such as preparing a prototype to process samples from bladder cancer patients, improving an existing electronic system using an SoC (system on a chip), and continuing the development environment for future projects,” says Pelegri.