Author Correction: Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials

NPJ Digit Med 2023 Deep Learning 1 Explanations View Original
Original Paper (PDF)

Unable to display PDF. Download it here or view on PMC.

Plain-English Explanations
Pages 1-2
Author Correction: Clarifying Tissue Sample Types Used in the Original Study

This document is an author correction to a previously published research article titled "Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials." The correction addresses a specific inaccuracy in how tissue samples used in the study were described.

The original article incorrectly stated that only pretreatment biopsy samples were used to train and validate the multi-modal AI (MMAI) models. In reality, the study also used posttreatment prostate tissue alongside pretreatment samples. All references to "pretreatment biopsy samples" and "pretreatment prostate biopsies" in the Methods and Results sections were updated accordingly.

To assess whether this correction affected the validity of the conclusions, the authors repeated the model validation using only pretreatment tissue cases (n=931, a slightly smaller subset than the original validation set). The results showed similar performance across all six clinical endpoints -- confirming that the core findings remain valid even when excluding posttreatment tissue from the analysis.

Key performance metrics from the pretreatment-only validation include: AUC of 0.83 for 5-year distant metastasis prediction (vs. NCCN AUC of 0.72), AUC of 0.78 for 10-year distant metastasis (vs. 0.69), and AUC of 0.77 for 10-year prostate cancer-specific survival (vs. 0.67). These figures consistently demonstrate the MMAI model's superior predictive ability compared to standard clinical risk tools.

TL;DR: This author correction clarifies that the original AI study used both pretreatment and posttreatment prostate tissue, and confirms that removing posttreatment cases does not change the key findings.
Citation: Open Access, . Available at: PMC9947124.