To paraphrase a famous poem, “How can AI serve prostate cancer? Let me count the ways.” In some ways, clinical researchers and software engineers have established a solid foundation for applying Artificial Intelligence (AI) to the detection, diagnosis, treatment and follow-up of prostate cancer (PCa). On the other hand, given the oceanic potential, we’ve only scratched the surface.
I want to briefly describe some of the ways that AI, including Machine Learning and Deep Learning, will serve to benefit PCa patients:
- Improved accuracy and efficiency in interpreting multiparametric MRI (mpMRI) of the prostate. In some cases, where mpMRI is unavailable or simply not feasible, AI can help biparametric MRI (bpMRI, or two rather than three parameters) is more available than mpMRI, AI can help compensate for the absence of dynamic contrast enhancement, a third parameter that reveals tumor blood flow.
- Better diagnostic performance on the part of radiologists, including reduced variability between more experienced and less experienced readers. Some research has shown that diagnostic algorithms based in Deep Learning perform at a level close to that of expert radiologists.
- Decrease the number of steps needed to achieve an optimum level of diagnosis.
- Greater accuracy in identifying patients who can safely avoid the need for a biopsy.
- Greater accuracy in detecting clinically significant disease, and evaluating its aggression level, prior to biopsy.
- Advantages in planning treatment, thanks to excellence in risk stratification and PCa staging.
Of course, with time and technologic evolution, I expect that more discoveries await. For now, the beginnings are not just promising, they are fulfilling expectations at the experimental level. Some centers, such as ours have already implemented aspects of AI in our clinical services for patients, but most experts agree that AI is still on the threshold of an ability to do diagnostic “heavy lifting.”
What do we need in order to bring AI across the threshold so I can broadly benefit the world of PCa? First, software developers need large datasets in order to “train” AI algorithms. In this case, large means tens or hundreds of thousands of MRI scans, biopsy slides, etc. Second, experienced readers must analyze and annotate the images, identifying the targeted region in order to “educate” the program. Third, we need to manage expectations: at this stage in AI development, it is seen as a decision support tool than a decision-making tool. It’s important that patients understand that the judgment of the radiologist, not a computer, is the final say in all matters related to the patient’s well-being.
In short, AI has not yet arrived but make no mistake, it is progressing rapidly. As a pioneer in incorporating AI resources into our Center’s diagnostic pathway, I can attest to a rosy future for the detection, diagnosis, and image guided treatment of AI thanks to the human/computer clinical duet.
NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.