The integrated radiomic-clinicopathologic nomogram (RadClip) was a better prognosticator of biochemical recurrence-free survival and adverse pathology than other standard tools.
An artificial intelligence (AI) tool showed early promise at predicting biochemical recurrence following radical prostatectomy (RP) in men with prostate cancer, according to a study published in EBioMedicine.1,2
The AI tool, which is an integrated radiomic-clinicopathologic nomogram (RadClip), uses AI algorithms to evaluate “subtle differences in heterogeneity and texture patterns inside and outside the tumor region on pre-operative MRI to predict patient outcome following surgery.”
The study demonstrated that RadClip was a better prognosticator of biochemical recurrence-free survival (bRFS) and adverse pathology (AP) than other standard prognostic tools, including CAPRA and the Decipher genomic test.
“This tool can help urologists, oncologists, and surgeons create better treatment plans so that their patients can have the most precise treatment,” Lin Li, a doctoral student in Case Western Reserve’s Biomedical Engineering Department and a member of the team that developed the tool, stated in a press release. “RadClip allows physicians to evaluate the aggressiveness of the cancer and the response to treatment so they don’t overtreat or undertreat the patient.”
The retrospective analysis included 198 patients with prostate cancer treated across 4 institutions between 2009 and 2017. The institutions included Cleveland Clinic, The Mount Sinai Hospital, University Hospitals, and the Hospital of the University of Pennsylvania.
All patients received pre-operative 3 Tesla MRI followed by RP and had available follow-up data including post-surgery serum PSA levels. Patients were excluded if they had received neoadjuvant or adjuvant therapy, received radiotherapy as the definitive treatment, or had PSA persistence after RP.
Using statistical models, the investigators compared methods to determine which approach was the best predictor of bRFS and AP in this population. Concordance index (C-index) was the comparison measure for bRFS prediction and AUC was the comparison measure for AP prediction.
At a median follow-up of 35 months, the C-index for RadClip (0.77) was higher than the C-index for CAPRA (0.68) and Decipher (0.51). The C-index was comparable between RadClip and CAPRA-S (0.75). Further, RadClip’s AUC for predicting adverse pathology (0.71) was higher than bother Decipher’s (0.66) and CAPRA’s (0.69).
“We’re bringing together and connecting a variety of information, from radiologic scans like MRI to digitized pathology specimen slides and genomic data, for providing a more comprehensive characterization of the disease,” Anant Madabhushi, PhD, CCIPD director, Donnell Institute Professor of Biomedical Engineering at Case Western Reserve and the study’s senior author, stated in the press release.
“Genomic-based tests cost several thousand dollars and involve destructive testing of the tissue,” Madabhushi added. “Prognostic predictions from an MRI scan provide a non-invasive method for making both short-term and long-term decisions on treatment.”
The authors listed several limitations of their study, including that it was a retrospective analysis; the study used biochemical recurrence as a surrogate marker for metastasis because follow-up time post-prostatectomy was not long enough; and the study was “prognostic and not predictive of added benefit of neoadjuvant or adjuvant therapy.”
Going forward, the investigators suggest that clinical trials are needed to show whether RadClip can be used to identify which patients undergoing prostatectomy should receive additional treatment.
1. Artificial intelligence tool for reading MRI scans could transform prostate cancer surgery and treatment. Posted online January 14, 2021. https://bit.ly/2KfmbU9. Accessed January 14, 2021.
2. Li L, Shiradkar R, Leo P, et al. A novel imaging based Nomogram for predicting post-surgical biochemical recurrence and adverse pathology of prostate cancer from pre-operative bi-parametric MRI. EBioMedicine. 2020;63:103163. doi: 10.1016/j.ebiom.2020.103163.