Βy Prof. Daniele Regge, Chief of the Radiology Unit, Candiolo Cancer Institute
Patients diagnosed with localised prostate cancer (PCa) are classified in different risk groups, i.e. low, intermediate or high risk, and this choice will affect treatment and ultimately impact on patients’ survival and overall quality of life. Usually, classification of PCa into a risk class is based on the results of the prostate-specific antigen (PSA), Gleason score retrieved from biopsy, and clinical stage (i.e., TNM). However, systematic biopsy, performed by sampling the gland randomly with retrieval of up 12 tissue cores, underestimates both PCa aggressiveness and tumour extension and may cause pain and local side effects. On the other side, performance of fusion biopsy, where the target is a suspicious region at MRI, is strongly related to the radiologists’ experience.
For all these reasons, there is a compelling need to develop tools that can precisely measure PCa aggressiveness without the side effects of biopsy, and support physicians in the selection of the most appropriate treatment option for each individual patient, taking into account tumour heterogeneity. Use Case (UC) 2 of ProCAncer-I aims to characterize PCa aggressiveness based on MRI by developing an AI signature, i.e. virtual biopsy, providing similar information to that of tissue biopsy.
MRI virtual biopsy might in the future substitute or complement tissue biopsy, limiting the use of the latter to specific subgroups of patients. Moreover, virtual biopsy could hypothetically provide information on cancer aggressiveness of the whole gland and monitor changes in tumour volume and aggressiveness with time. A fully automatic non-invasive tool based on MRI, providing a likelihood score of PCa aggressiveness, has already been developed and validated on 131 patients (149 tumours) from two different institutions. Preliminary findings are encouraging in distinguishing low and highly aggressive PCa (figure) [doi: 10.3389/fonc.2021.718155]. Within the ProCAncer-I consortium we will have the opportunity to develop and validate prostate virtual biopsy on a much larger data-set, including over 5,000 patients
from 13 different European clinical centres.
Figure: Waterfall plot of the PCa aggressiveness radiomics score. From the recent ProCAncer-I publication: Giannini V, Mazzetti S, Defeudis A et al. “A Fully Automatic Artificial Intelligence System Able to Detect and Characterize Prostate Cancer Using Multiparametric MRI: Multicenter and Multi-Scanner Validation“, Frontiers in Oncology 2021 https://doi.org/10.3389/fonc.2021.718155