A method is proposed for predicting long-term treatment failure using “eigentumors”: principal components computed from volumes surrounding breast tumors in contrast-enhanced images. The dataset contains pre-treatment scans of 563 consecutively included patients with early-stage breast cancer with median follow-up of 86 months. Principal components of washin and washout in box-shaped regions surrounding the tumors were computed, and LASSO and logistic regression were used to construct a model for predicting the probability of treatment failure. ROC analysis yields a bootstrapped performance of 0.73, and bootstrapped Kaplan-Meier survival curves based on the model’s outcome show significant separation (χ=32.89, P < 0.0001).
This abstract and the presentation materials are available to members only; a login is required.