Poitrasson-Rivière A, Moody JB, Renaud JM, Buckley CJ, Ficaro EP, Murthy VL. Automated deep learning segmentation of cardiac inflammatory FDG PET. Journal of Nuclear Cardiology. 2024;42:102052.
Description:
This article, published in the Journal of Nuclear Cardiology (2024), presents the development and validation of a deep learning–based approach for automated segmentation of cardiac inflammatory uptake on FDG PET imaging. The method was designed to assist in evaluating patients with suspected cardiac sarcoidosis or other forms of myocardial inflammation.
4DM software was used for initial image processing and data preparation, supporting standardization of the imaging inputs used to train and evaluate the segmentation model.
The authors trained a convolutional neural network on FDG PET images labeled by expert interpreters and assessed its performance on a separate test set by comparing automated segmentations to manual references.
Key Findings:
Segmentation Accuracy: The model achieved strong overlap with expert manual segmentations.
Efficiency: Automation significantly reduced the time required for analysis.
Reproducibility: The method produced consistent results across varied patient scans.
Clinical Relevance:
Deep learning tools for segmenting inflammatory activity on FDG PET may enhance workflow efficiency and interpretive consistency in the evaluation of cardiac sarcoidosis. The use of 4DM in preprocessing supports integration of automated analysis within clinical imaging pipelines.
Partners in Research:
INVIA Medical Imaging Solutions and the Division of Cardiovascular Medicine in the Department of Internal Medicine at the University of Michigan collaborated on this research.
