New AI Loss Function Using Fuzzy Logic Improves MRI Brain Image Segmentation

A novel loss function for deep learning models that incorporates fuzzy logic to handle uncertainty in medical image segmentation has been introduced. This approach specifically targets the segmentation of brain tissues from magnetic resonance imaging (MRI) scans, a critical task for neurological disease research and medical image analysis. The method combines the categorical cross-entropy (CCE) loss function with a fuzzy entropy component derived from fuzzy logic principles. By accounting for inherent uncertainties in pixel classification, the function aims to improve model optimization. Evaluation was conducted on two publicly available benchmark datasets: IBSR and OASIS. Two widely recognized neural network architectures, U-Net and U-Net++, were used in the experiments. Results indicated that models trained with the proposed loss function achieved superior performance compared to standard approaches. The research was announced on arXiv under the identifier arXiv:2604.16490v1 with a cross-announcement type. The work addresses a pivotal challenge in medical image computing by enhancing the accuracy of distinguishing various brain tissues.