Fully Automated Skull Stripping from Brain Magnetic Resonance Images Using Mask RCNN-Based Deep Learning Neural Networks
Humera Azam,
Humera Tariq,
Danish Shehzad
et al.
Abstract:This research comprises experiments with a deep learning framework for fully automating the skull stripping from brain magnetic resonance (MR) images. Conventional techniques for segmentation have progressed to the extent of Convolutional Neural Networks (CNN). We proposed and experimented with a contemporary variant of the deep learning framework based on mask region convolutional neural network (Mask–RCNN) for all anatomical orientations of brain MR images. We trained the system from scratch to build a model… Show more
“…However, the reliance on large datasets and the inherent complexity of deep learning models may present challenges in resource-constrained environments and interpretability. Azam et al (2023) introduced an innovative deep learning-based skull removal method using Mask-RCNN, showcasing superior performance compared to traditional approaches. While it offers enhanced accuracy and automation, its computational intensity and model complexity may pose challenges for resource-constrained environments.…”
Brain tumors present a significant medical challenge, often necessitating surgical intervention for treatment. In the context of postoperative brain MRI, the primary focus is on the resection cavity, the void that remains in the brain following tumor removal surgery. Precise segmentation of this resection cavity is crucial for a comprehensive assessment of surgical efficacy, aiding healthcare professionals in evaluating the success of tumor removal. Automatically segmenting surgical cavities in post-operative brain MRI images is a complex task due to challenges such as image artifacts, tissue reorganization, and variations in appearance. Existing state-of-the-art techniques, mainly based on Convolutional Neural Networks (CNNs), particularly U-Net models, encounter difficulties when handling these complexities. The intricate nature of these images, coupled with limited annotated data, highlights the need for advanced automated segmentation models to accurately assess resection cavities and improve patient care. In this context, this study introduces a two-stage architecture for resection cavity segmentation, featuring two innovative models. The first is an automatic skull removal model that separates brain tissue from the skull image before input into the cavity segmentation model. The second is an automated postoperative resection cavity segmentation model customized for resected brain areas. The proposed resection cavity segmentation model is an enhanced U-Net model with a pre-trained VGG16 backbone. Trained on publicly available post-operative datasets, it undergoes preprocessing by the proposed skull removal model to enhance precision and accuracy. This segmentation model achieves a Dice coefficient value of 0.96, surpassing state-of-theart techniques like ResUNet, Attention U-Net, U-Net++, and U-Net.
“…However, the reliance on large datasets and the inherent complexity of deep learning models may present challenges in resource-constrained environments and interpretability. Azam et al (2023) introduced an innovative deep learning-based skull removal method using Mask-RCNN, showcasing superior performance compared to traditional approaches. While it offers enhanced accuracy and automation, its computational intensity and model complexity may pose challenges for resource-constrained environments.…”
Brain tumors present a significant medical challenge, often necessitating surgical intervention for treatment. In the context of postoperative brain MRI, the primary focus is on the resection cavity, the void that remains in the brain following tumor removal surgery. Precise segmentation of this resection cavity is crucial for a comprehensive assessment of surgical efficacy, aiding healthcare professionals in evaluating the success of tumor removal. Automatically segmenting surgical cavities in post-operative brain MRI images is a complex task due to challenges such as image artifacts, tissue reorganization, and variations in appearance. Existing state-of-the-art techniques, mainly based on Convolutional Neural Networks (CNNs), particularly U-Net models, encounter difficulties when handling these complexities. The intricate nature of these images, coupled with limited annotated data, highlights the need for advanced automated segmentation models to accurately assess resection cavities and improve patient care. In this context, this study introduces a two-stage architecture for resection cavity segmentation, featuring two innovative models. The first is an automatic skull removal model that separates brain tissue from the skull image before input into the cavity segmentation model. The second is an automated postoperative resection cavity segmentation model customized for resected brain areas. The proposed resection cavity segmentation model is an enhanced U-Net model with a pre-trained VGG16 backbone. Trained on publicly available post-operative datasets, it undergoes preprocessing by the proposed skull removal model to enhance precision and accuracy. This segmentation model achieves a Dice coefficient value of 0.96, surpassing state-of-theart techniques like ResUNet, Attention U-Net, U-Net++, and U-Net.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.