This new European Radiology study aimed to develop a deep-learning model for classifying myocardial iron overload (MIO) using magnitude T2* multi-echo MR images from 496 thalassemia major patients. Two 2D convolutional neural networks (CNN), MS-HippoNet (multi-slice) and SS-HippoNet (single-slice), were trained using 5-fold cross-validation. The model showed strong performance with a multi-class accuracy of 0.885 and 0.836 on test sets, and a good agreement with radiologists’ classifications (κ values of 0.771 and 0.614, respectively). The networks demonstrated effective MIO classification, offering a promising tool for automated MRI data analysis in clinical settings.
Key points:
- Question MRI T2* represents the established clinical tool for MIO assessment. Quality control of the image analysis is a problem in small centers.
- Findings Deep learning models can perform MIO staging with good accuracy, comparable to inter-observer variability of the standard procedure.
- Clinical relevance CNN can perform automated staging of cardiac iron overload from multiecho MR sequences facilitating non-invasive evaluation of patients with various hematologic disorders.
Authors: Vincenzo Positano, Antonella Meloni, Lisa Anita De Santi, Laura Pistoia, Zelia Borsellino, Alberto Cossu, Francesco Massei, Paola Maria Grazia Sanna, Maria Filomena Santarelli & Filippo Cademartiri
