Efficient machine learning models leveraging DCE‑MRI morphological and dynamic features allow accurate breast lesion classification.
Authors
Affiliations (10)
Affiliations (10)
- Università degli Studi di Pisa Dipartimento di Informatica, Largo Bruno Pontecorvo 3, Pisa, Tuscany, 56127, Italy.
- Universita degli Studi di Pisa Dipartimento di Fisica Enrico Fermi, Largo Bruno Pontecorvo 3, Pisa, 56127, Italy.
- Department of Translational Research, Università di Pisa, Via Savi 10, Pisa, Tuscany, 56126, Italy.
- Department of Radiology, Azienda USL Toscana nord ovest, Via Enrico Mattei 21, Massa, Tuscany, 54100, Italy.
- Unità di Fisica Medica, Azienda USL Toscana nord ovest, Via Enrico Mattei 21, Massa, Tuscany, 54100, Italy.
- University of Pisa, Via Roma 67, Pisa, 56126, Italy.
- Istituto Nazionale di Fisica Nucleare, Via Enrico Fermi 54, Frascati, 00044, Italy.
- Physics, University of Pisa, Largo Bruno Pontecorvo 3, Pisa, 56127, Italy.
- Department of Physics, University of Pisa, Largo Bruno Pontecorvo 3, Pisa, 56127, Italy.
- National Institute of Nuclear Physics Section of Pisa, Largo Bruno Pontecorvo 3, Pisa, Toscana, 56127, Italy.
Abstract
We propose an ensemble learning approach to classify malignant versus benign breast lesions leveraging morphological and dynamic features derived from Magnetic Resonance Images (MRI). The analysis has been performed on 164 breast lesions of the publicly available "Advanced MRI Breast Lesions" dataset from The Cancer Imaging Archive, containing T2-weighted and Dynamic Contrast-Enhanced (DCE)-MRI sequences, along with the segmentation masks of suspicious lesions. After extracting radiomic features using Pyradiomics Python package, we computed dynamic features from DCE-MRI kinetic curves, which describe the contrast agent wash-in and wash-out. These features have been defined as the derivatives of image intensity measures, like mean and standard deviation, computed inside the masks on the 5 DCE-MRI time steps. We trained and evaluated an eXtreme Gradient Boosting (XGBoost) classifier, experimenting with different combinations of features and segmentation masks in a stratified 5-fold cross-validation. The best model trained on T2-weighted MRI morphological features achieved an Area Under the Curve (AUC) score of 0.83±0.04 on the independent test set consisting of 20 lesions, while the model using only dynamic features performed an AUC of 0.91±0.03. Despite being obtained on a small test sample, these results show the potential of features derived from DCE images in breast lesions classification.