Factorizer: A scalable interpretable approach to context modeling for medical image segmentation.

Convolutional Neural Networks (CNNs) with U-shaped architectures have dominated medical image segmentation, which is crucial for various clinical purposes. However, the inherent locality of convolution makes CNNs fail to fully exploit global context, essential for better recognition of some structures, e.g., brain lesions. Transformers have recently proven promising performance on vision tasks, including semantic segmentation, mainly due to their capability of modeling long-range dependencies. Nevertheless, the quadratic complexity of attention makes existing Transformer-based models use self-attention layers only after somehow reducing the image resolution, which limits the ability to capture global contexts present at higher resolutions. Therefore, this work introduces a family of models, dubbed Factorizer, which leverages the power of low-rank matrix factorization for constructing an end-to-end segmentation model. Specifically, we propose a linearly scalable approach to context modeling, formulating Nonnegative Matrix Factorization (NMF) as a differentiable layer integrated into a U-shaped architecture. The shifted window technique is also utilized in combination with NMF to effectively aggregate local information. Factorizers compete favorably with CNNs and Transformers in terms of accuracy, scalability, and interpretability, achieving state-of-the-art results on the BraTS dataset for brain tumor segmentation and ISLES'22 dataset for stroke lesion segmentation. Highly meaningful NMF components give an additional interpretability advantage to Factorizers over CNNs and Transformers. Moreover, our ablation studies reveal a distinctive feature of Factorizers that enables a significant speed-up in inference for a trained Factorizer without any extra steps and without sacrificing much accuracy. The code and models are publicly available at https://github.com/pashtari/factorizer.

New multiple sclerosis lesion segmentation and detection using pre-activation U-Net.

Automated segmentation of new multiple sclerosis (MS) lesions in 3D MRI data is an essential prerequisite for monitoring and quantifying MS progression. Manual delineation of such lesions is time-consuming and expensive, especially because raters need to deal with 3D images and several modalities. In this paper, we propose Pre-U-Net, a 3D encoder-decoder architecture with pre-activation residual blocks, for the segmentation and detection of new MS lesions. Due to the limited training set and the class imbalance problem, we apply intensive data augmentation and use deep supervision to train our models effectively. Following the same U-shaped architecture but different blocks, Pre-U-Net outperforms U-Net and Res-U-Net on the MSSEG-2 dataset, achieving a Dice score of 40.3% on new lesion segmentation and an F1 score of 48.1% on new lesion detection. The codes and trained models are publicly available at https://github.com/pashtari/xunet.

Classification of multiple sclerosis clinical profiles using machine learning and grey matter connectome.

Purpose: The main goal of this study is to investigate the discrimination power of Grey Matter (GM) thickness connectome data between Multiple Sclerosis (MS) clinical profiles using statistical and Machine Learning (ML) methods. Materials and Methods: A dataset composed of 90 MS patients acquired at the MS clinic of Lyon Neurological Hospital was used for the analysis. Four MS profiles were considered, corresponding to Clinical Isolated Syndrome (CIS), Relapsing-Remitting MS (RRMS), Secondary Progressive MS (SPMS), and Primary Progressive MS (PPMS). Each patient was classified in one of these profiles by our neurologist and underwent longitudinal MRI examinations including T1-weighted image acquisition at each examination, from which the GM tissue was segmented and the cortical GM thickness measured. Following the GM parcellation using two different atlases (FSAverage and Glasser 2016), the morphological connectome was built and six global metrics (Betweenness Centrality (BC), Assortativity (r), Transitivity (T), Efficiency (E g ), Modularity (Q) and Density (D)) were extracted. Based on their connectivity metrics, MS profiles were first statistically compared and second, classified using four different learning machines (Logistic Regression, Random Forest, Support Vector Machine and AdaBoost), combined in a higher level ensemble model by majority voting. Finally, the impact of the GM spatial resolution on the MS clinical profiles classification was analyzed. Results: Using binary comparisons between the four MS clinical profiles, statistical differences and classification performances higher than 0.7 were observed. Good performances were obtained when comparing the two early clinical forms, RRMS and PPMS (F1 score of 0.86), and the two neurodegenerative profiles, PPMS and SPMS (F1 score of 0.72). When comparing the two atlases, slightly better performances were obtained with the Glasser 2016 atlas, especially between RRMS with PPMS (F1 score of 0.83), compared to the FSAverage atlas (F1 score of 0.69). Also, the thresholding value for graph binarization was investigated suggesting more informative graph properties in the percentile range between 0.6 and 0.8. Conclusion: An automated pipeline was proposed for the classification of MS clinical profiles using six global graph metrics extracted from the GM morphological connectome of MS patients. This work demonstrated that GM morphological connectivity data could provide good classification performances by combining four simple ML models, without the cost of long and complex MR techniques, such as MR diffusion, and/or deep learning architectures.