Related Experiment Video
Updated: May 24, 2026
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Preoperative Risk Assessment of Adrenal Metastases in a Multicenter Study: Development of a Robust Federated Learning
Bao Feng1, Zhaole Yu2, Yehang Chen2
1Laboratory of Intelligent Detection and Information Processing, Guilin University of Aerospace Technology, Guilin 541004, China (B.F., Z.Y., Y.C., J.X., Q.H.); Jiangmen Key Laboratory of Artificial Intelligence in Medical Image Computation and Application, Jiangmen Central Hospital, Jiangmen 529030, China (B.F., W.L., E.C.).
A robust federated learning model (RFLM) accurately assesses adrenal lesions using pre-contrast CT scans. This artificial intelligence approach improves preoperative risk assessment, potentially reducing misdiagnosis and the need for enhanced imaging.
Area of Science:
- Radiology and Artificial Intelligence
- Medical Imaging Analysis
- Oncology
Background:
- Preoperative assessment of adrenal lesions faces misdiagnosis risks.
- Artificial intelligence (AI) shows promise in improving diagnostic accuracy.
- Multi-center AI model performance is challenged by data privacy and heterogeneity.
Purpose of the Study:
- To develop a robust federated learning model (RFLM) for differentiating adrenal metastases from benign lesions.
- To evaluate the RFLM's performance using multi-phase computed tomography (CT) data.
- To address challenges of data privacy and non-independent data in multi-center studies.
Main Methods:
- Retrospective analysis of 1187 adrenal lesions from 1100 patients (2008-2021).
- Development of an RFLM combining graph networks and mutual information.
- Experimental evaluation on three-phase CT (pre-contrast, venous, arterial) across multiple centers.
Main Results:
- The pre-contrast phase (PCP)-based RFLM demonstrated superior performance across all four centers.
- Area Under the Curve (AUC) values for PCP-based RFLM ranged from 0.7826 to 0.8895.
- The model showed strong performance and potential clinical utility across diverse datasets.
Conclusions:
- PCP-based RFLM provides effective preoperative risk assessment for adrenal lesions.
- This AI approach can reduce reliance on contrast-enhanced CT, minimizing radiation exposure.
- The RFLM is a robust and reliable diagnostic tool for adrenal lesion characterization.
