Related Experiment Video
Updated: Jul 12, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.1K
Dual adversarial deconfounding autoencoder for joint batch-effects removal from multi-center and multi-scanner
Lara Cavinato1, Michela Carlotta Massi2, Martina Sollini3,4
1MOX, Department of Mathematics, Politecnico di Milano, Piazza Leonardo da Vinci, 32, Milan, 20133, Italy. lara.cavinato@polimi.it.
Scientific Reports
|November 2, 2023
Summary
This study introduces a novel algorithm to harmonize medical imaging data from multiple centers for Hodgkin Lymphoma patients. The method effectively reduces scanner-specific noise and improves prognostic model accuracy, aiding clinical decision-making.
Area of Science:
- Medical Imaging
- Quantitative Image Analysis
- Oncology
Background:
- Medical imaging is crucial for cancer investigation and monitoring.
- Multi-center studies require denoising of confounding factors like scanner/center-specific influences.
- Lymphoma imaging needs to represent multi-site disease spread for effective biomarkers.
Purpose of the Study:
- To address the dual-factor deconfusion problem in medical imaging.
- To propose a deconfusion algorithm for harmonizing Hodgkin Lymphoma imaging data in multi-center settings.
- To improve prognostic models by enhancing imaging data quality.
Main Methods:
- Development of a novel deconfusion algorithm.
- Application of the algorithm to harmonize multi-center imaging data for Hodgkin Lymphoma.
- Evaluation of the algorithm's ability to denoise data and preserve spatial lesion relationships.
Main Results:
- The algorithm successfully denoises data from domain-specific variability (p < 0.001).
- It preserves the spatial relationship between peer lesions (p = 0), a key prognostic biomarker.
- Harmonization significantly improves prognostic model performance (p < 0.001 training, p < 0.05 testing).
Conclusions:
- The proposed algorithm effectively harmonizes multi-center medical imaging data for Hodgkin Lymphoma.
- This approach enables more accurate prognostic assessments and exhaustive patient representations.
- The work facilitates large-scale, reproducible analyses for clinical translation of imaging biomarkers.
