Interview with prof. Daniele Regge (Clinical Coordinator of the ProCancer-I project)
Q: In the case of prostate cancer there are several tests that can help clinicians in the decision process, for example, prediction algorithms based on PSA blood test and biopsy. Why are they not good enough? What is the role of MRI in this context?
A: First, we know that there is a proportion of men with elevated PSA values that do not have cancer, and on the contrary, another group of patients with cancer but with PSA within the normal range. Second, biopsy alone is not sufficient to predict the presence of cancer. Indeed, a biopsy could miss cancer or it could sample only a small unrepresentative area of the tumour, failing to provide important information to guide the clinical decision. For the above reasons urologists now request MRI before the biopsy. MRI scans the whole prostate accurately detecting suspicious lesions, guiding biopsy towards them. Major limitations of MRI are the long reporting times and the need for experienced readers, the latter not readily available. Humans cannot easily perceive and process the huge amount of information embedded in each MRI examination. Artificial intelligence comes into play by supporting radiologists in detecting cancer.
Q: The ProCAncer-I European project aims to develop several artificial intelligence solutions related to the most relevant clinical needs in prostate cancer. Which are at the moment the most important issues when managing prostate cancer patients?
A: Other than detection, one important task of AI could be to distinguish prostate cancers that are born to be bad, from those that are born to be good. In the past, this did not matter since prostate cancer patients were treated with a one-size-fits-all approach. Today the goal is precision care. One of the most important aims of ProCAncer-I is to identify patients with aggressive forms of cancer, that therefore need whole gland treatment (prostatectomy or radiotherapy) from those that harbour indolent cancer and that can safely undergo active surveillance, sparing the side effects of treatments. Recent research has shown that with AI algorithms it is possible to extract and process information derived from prostate MRI that can be related to tumour histology. The capability of MRI to grade cancer without the need for tissue analysis is defined as virtual biopsy.
Q: Given the importance of artificial intelligence applications, which are the advantages that the ProCancer-I project could bring into the real world?
A: The ProCancer-I is an EU-funded project which brings together 20 different European centres of excellence (13 clinical centres, 6 technical partners, and 1 legal team) that will contribute to the creation of ProstateNET, the biggest collection of data and imaging (more than 17,000 cases) related to prostate cancer. This is really important because most trial results are achieved from limited datasets, hampering generalization to other patient cohorts. Using the ProstateNET database, we will not only detect and characterise prostate cancer, as previously said, but we will also aim at predicting which patients will develop metastasis, which will likely relapse and which will suffer the side effects of treatment. These findings will help clinicians to focus on the cases that should be follow-up more closely or that should undergo additional treatments, i.e. administration of adjuvant therapies, lymphadenectomy, etc. This approach could really become a game-changer in the fight against the most common cancer in men.
Q: What should we expect from ProCancer-I and AI in the next five years? Will radiologists still be necessary to the patients or will computers do all the work?
A: In the last few years there has been a lot of talking about AI and how it will affect our everyday life. We are increasingly relying on the computer workforce and the same is happening in hospitals, where AI is gradually becoming our virtual assistant. In prostate cancer, for example, there are studies demonstrating how the combination of radiologists and computers can improve the reporting process, bringing the benefit of a timely and accurate diagnosis in patients with clinically significant disease. When the results of ProCAncer-I will be integrated into a seamless process, cases could be prioritized by placing at the top of the list those needing accurate reporting by the radiologists, so that greater attention could be addressed to those who actually need treatment. In this view computers will do a great job, cooperating with the human workforce and this synergy will reduce the workload for trivial aspects, letting clinicians focus on what really matters, their patients.
