Multi-channel MRI Embedding: An Effective Strategy for Enhancing Brain Tumor Segmentation

A brain tumor is a collection or a mass of abnormal cells in the brain. One of the most important diagnostic tasks in medical image processing is the brain whole tumor segmentation. It assists in quicker clinical assessment and early detection of tumor development, which is crucial for the lifesaving treatment procedures of patients. Because, the human skull encloses the brain very rigidly and any growth inside this restricted place can cause severe health issues. For instance, brain tumors often can be malignant, if they are not detected at an early stage. An early detection of brain tumors requires careful and intricate analysis for surgical planning or treatment. To diagnose such malignancies, most physicians employ Magnetic Resonance Imaging (MRI) as a fundamental tool. A manual analysis of the MRI data is known to be time-consuming; approximately, it takes up to eighteen hours per sample. Thus, the automatic segmentation of tumors has become an optimal solution for this problem. Studies have shown that this technique provides better accuracy and is faster than manual analysis, resulting in patients receiving the treatment at the right time. This work introduces an efficient strategy, called multi-channel MRI embedding to enhance the accuracy of deep learning-based brain whole tumor segmentation. The experimental analysis on the Brain Tumor Segmentation 2019 (BraTs'19) dataset shows significant improvement. The embedding strategy surmounts the state-of-the-art approaches with an improvement of 2% without any timing overheads.