Artificial intelligence in prostate cancer

Artificial intelligence (A.I.) has recently become a buzzword in so many aspects of our lives, but it has been used to some degree in health care for a while. One area of health care where A.I. has made significant strides is the diagnosis and treatment of prostate cancer.

"We are just at the tip of the iceberg of utilizing A.I. for prostate cancer," says Dr. David D. Yang, a radiation oncologist with Harvard-affiliated Brigham and Women's Hospital and Dana-Farber Cancer Institute. "So far, it has been shown to help improve the care for men with prostate cancer in limited, yet effective ways."

A.I. refers to systems or computers capable of mimicking human thought processes to solve problems. With prostate cancer, A.I. is mainly used via software programs that apply complex computational algorithms to improve precision and efficiency in diagnosis, treatment, and management. Here's a closer look at three ways A.I. is being used in this area.

Examining biopsies

Men suspected of having prostate cancer based on high or rising blood levels of prostate-specific antigen (PSA) may have a biopsy to look for cancer. Typically, tissue samples are taken from many locations in the prostate and then examined under a microscope by a pathologist. The pathologist must determine whether the sample contains cancer cells and, if so, assign a Gleason score to indicate how aggressive the cancer cells appear to be. This can be challenging and time-consuming, particularly where there is only a small amount of cancer. The assignment of the Gleason score is also partially subjective and can vary between pathologists.

A.I. may help speed up this process and increase the accuracy of a cancer diagnosis. Paige Prostate is the first FDA-approved A.I.-based software for this use. Like many other A.I. models, this one "learned" how to detect the presence of prostate cancer by training on many prostate samples where pathologists had indicated whether there is cancer present.

The software analyzes a digital copy of the prostate biopsy slides. Then it identifies areas with the highest likelihood of harboring cancer, so the pathologist can review them. "This can help the pathologist by flagging areas they should look at with more detail and decreases the time needed to find those areas during their initial review," says Dr. Yang. "It's like putting a highlighter to the samples so some parts stand out." This A.I. model can also provide a preliminary Gleason score for the pathologist to review and may help reduce some of the subjectivity of Gleason scoring.

Working with MRIs

A.I. is also starting to be used alongside MRI to more accurately identify prostate cancer. MRI is used as a diagnostic tool to identify areas within the prostate that might harbor cancer. These areas can then be directly biopsied.

The integration of A.I. and MRI works like this: After an MRI scan is completed, the resulting images are run through A.I. software that measures the volume of the prostate gland and identifies suspicious lesions. The software assesses visible features, such as brightness and shadows, as well as those invisible to the human eye.

When finished, the software creates a "map" on the image showing where cancer might lie. It provides a Prostate Imaging Reporting and Data System (PI-RADS) score to indicate how likely a suspicious area is a clinically significant cancer. PI-RADS scores range from 1 (most likely not cancer) to 5 (very suspicious). The radiologist confirms the A.I.'s interpretation before it becomes available to other clinicians and patients.

"The advantage of this type of A.I. is that it aids the radiologist in more accurately and quickly identifying areas of possible cancer," says Dr. Yang. "In addition to guiding prostate biopsies, this level of detailed information can help determine the cancer grade and potential treatment options. A man with low-risk cancer could go on active surveillance, while a man with intermediate or high-risk prostate cancer may benefit from radiation therapy or surgery."

Androgen deprivation therapy

Men with intermediate-risk prostate cancer who choose radiation therapy for the treatment of localized tumors also may benefit from A.I. Typically, these men are advised to add a short course of androgen deprivation therapy (ADT) to reduce their testosterone levels.

"However, we do not know which patients with intermediate-risk prostate cancer would most likely benefit from ADT and which patients can be safely spared its side effects," says Dr. Yang.

An A.I. tool called ArteraAI Prostate analyzes digitized prostate pathology slides to help identify which patients are most likely to be aided by ADT.

"More research is still needed to ensure this A.I. model's predictions are accurate, but preliminarily, it is an exciting potential tool that will help us make more personalized treatment recommendations," says Dr. Yang.

Even with A.I.-assisted medical information in hand, doctors must still interpret the results and make their recommendations to their patients. "It is still up to the physician to decide whether X or Y is the right diagnosis and best treatment option," says Dr. Yang. "A.I. will continue to be a valuable tool, but it won't replace the human element of guiding you through the process of diagnosis, treatment, and management of your prostate cancer."

Image: © pcess609/Getty Images