Related Experiment Video
Updated: May 31, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
An Artificial Intelligence-Digital Pathology Algorithm Predicts Survival After Radical Prostatectomy From the
Eric V Li1, Yi Ren2, Jacqueline Griffin2
1Department of Urology, Northwestern University Feinberg School of Medicine, Chicago, Illinois.
Locked multimodal artificial intelligence (MMAI) algorithms trained on biopsy slides can predict prostate cancer outcomes after surgery. These AI models show promise in identifying patients who may benefit from additional treatments post-radical prostatectomy.
Area of Science:
- Urology
- Oncology
- Artificial Intelligence in Medicine
Background:
- Clinical variables alone have limited predictive power for recurrence after radical prostatectomy (RP).
- Digital histopathological images offer a rich data source for predictive modeling in prostate cancer.
- Previous multimodal artificial intelligence (MMAI) models were developed using biopsy specimens for radiation treatment planning.
Purpose of the Study:
- To evaluate the predictive ability of locked MMAI algorithms, originally trained on prostate biopsy specimens, for prostate cancer-specific mortality (PCSM) and overall survival (OS) in patients undergoing RP.
- To assess if MMAI models developed for radiation therapy can be repurposed for predicting outcomes in surgically treated prostate cancer patients.
Main Methods:
- Utilized a subset of patients from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Randomized Controlled Trial with digitized RP specimens and survival data.
- Applied locked MMAI algorithms for PCSM and distant metastasis (DM) prediction, originally trained on biopsy slides.
- Employed Cox proportional hazards modeling and Kaplan-Meier survival analysis to evaluate prediction accuracy.
Main Results:
- 1032 patients undergoing RP with a median follow-up of 17 years were analyzed.
- MMAI algorithms for PCSM and DM significantly predicted PCSM (HRs > 1.9, P < .001).
- Both MMAI models also significantly predicted overall survival (OS) (HRs > 1.19, P < .04).
Conclusions:
- Locked MMAI algorithms trained on biopsy specimens successfully predicted clinical outcomes (PCSM and OS) when applied to RP specimens.
- These findings suggest MMAI models can be valuable tools for predicting outcomes in surgically treated prostate cancer patients.
- MMAI and other biomarkers could aid in selecting patients for intensified postoperative treatment, such as androgen deprivation therapy or radiation.
More Related Videos
13:19Microarray-based Identification of Individual HERV Loci Expression: Application to Biomarker Discovery in Prostate Cancer
Published on: November 2, 2013
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
