3D Capsule Networks for Brain Image Segmentation

Cascade Residual Multiscale Convolution and Mamba-Structured UNet for Advanced Brain Tumor Image Segmentation

In brain imaging segmentation, precise tumor delineation is crucial for diagnosis and treatment planning. Traditional approaches include convolutional neural networks (CNNs), which struggle with processing sequential data, and transformer models that face limitations in maintaining computational efficiency with large-scale data. This study introduces MambaBTS: a model that synergizes the strengths of CNNs and transformers, is inspired by the Mamba architecture, and integrates cascade residual multi-scale convolutional kernels. The model employs a mixed loss function that blends dice loss with cross-entropy to refine segmentation accuracy effectively. This novel approach reduces computational complexity, enhances the receptive field, and demonstrates superior performance for accurately segmenting brain tumors in MRI images. Experiments on the MICCAI BraTS 2019 dataset show that MambaBTS achieves dice coefficients of 0.8450 for the whole tumor (WT), 0.8606 for the tumor core (TC), and 0.7796 for the enhancing tumor (ET) and outperforms existing models in terms of accuracy, computational efficiency, and parameter efficiency. These results underscore the model's potential to offer a balanced, efficient, and effective segmentation method, overcoming the constraints of existing models and promising significant improvements in clinical diagnostics and planning.

Attention-gated 3D CapsNet for robust hippocampal segmentation

Purpose

The hippocampus is organized in subfields (HSF) involved in learning and memory processes and widely implicated in pathologies at different ages of life, from neonatal hypoxia to temporal lobe epilepsy or Alzheimer's disease. Getting a highly accurate and robust delineation of sub-millimetric regions such as HSF to investigate anatomo-functional hypotheses is a challenge. One of the main difficulties encountered by those methodologies is related to the small size and anatomical variability of HSF, resulting in the scarcity of manual data labeling. Recently introduced, capsule networks solve analogous problems in medical imaging, providing deep learning architectures with rotational equivariance. Nonetheless, capsule networks are still two-dimensional and unassessed for the segmentation of HSF.

Approach

We released a public 3D Capsule Network (3D-AGSCaps, https://github.com/clementpoiret/3D-AGSCaps) and compared it to equivalent architectures using classical convolutions on the automatic segmentation of HSF on small and atypical datasets (incomplete hippocampal inversion, IHI). We tested 3D-AGSCaps on three datasets with manually labeled hippocampi.

Results

Our main results were: (1) 3D-AGSCaps produced segmentations with a better Dice Coefficient compared to CNNs on rotated hippocampi (p = 0.004 , cohen's d = 0.179 ); (2) on typical subjects, 3D-AGSCaps produced segmentations with a Dice coefficient similar to CNNs while having 15 times fewer parameters (2.285M versus 35.069M). This may greatly facilitate the study of atypical subjects, including healthy and pathological cases like those presenting an IHI.

Conclusion

We expect our newly introduced 3D-AGSCaps to allow a more accurate and fully automated segmentation on atypical populations, small datasets, as well as on and large cohorts where manual segmentations are nearly intractable.