Brain hyperintensities: automatic segmentation of white matter hyperintensities in clinical brain MRI images using improved deep neural network

White matter hyperintensities (WMH) are commonly found in the brains of healthy elderly individuals and have been associated with various neurological and geriatric disorders. Automatic WMH segmentation is essential for evaluating the natural disease course and the efficacy of clinical interventions, especially drug discovery. This work presents a hybrid method, which combines the novel metaheuristic Slime Mold Algorithm with the Harris Hawk's Optimization algorithm to train the Deep Convolutional Neural Network (DCNN) as a robust model to optimize the Kapur’s entropy for WMH segmentation. We named it as Slime Mold Harris Hawk's Optimization-based DCNN (SMHHO-DCNN). The proposed method was tested on the open WMH segmentation challenge MICCAI 2017 and in-house dataset. However, the WMH regions would appear blurry and globally inconsistent. In order to solve the fuzzy problem, we improved the network architecture by incorporating skip connections across the mirrored layers with in encoder and decoder stacks. As a result, the segmented WMH output appears more realistic and coherent with its surrounding contexts, which enables the hybrid attention module to further enhance the precision of WMH localization by extracting the supporting information of high-level characteristics and low-level features. As a result, that the proposed hybrid optimization, skip connections, has increased the segmentation performance of Dice score to 87.2%, precision 90.3%, recall 89.5%, and f1-score 88.8% for MRI dataset which is superior compared to other approaches. The proposed technique outperformed existing state-of-the-art methodologies in both qualitative and quantitative measurements across a wide range of medical modalities. Furthermore, the dataset contains T1-w, T2-w, and FLAIR images, demonstrating the adaptability and robustness of the model.