Hypoxic Ischemic Encephalopathy (HIE) is a brain disease that affects thousands of neonates every year. Accurate and immediate diagnoses of HIE are crucial elements for clinical treatment and can reduce the disease’s fatality. Magnetic resonance images (MRIs) are effectively used in the clinical treatment of HIE hence, correct segmentation of lesion regions is quite beneficial. However, segmenting HIE lesions is a challenging task due to their nature (i.e., more diffuse and smaller than other lesions). In this paper we propose a novel mathematical operation, “Reciprocal Transformation”, which transforms the data from one distribution to a more suitable distribution to train a deep learning model better. Combined with the concatenation operation, we observe a significant improvement in our validation dataset. The proposed method achieved a mean dice score of 48.26 in the test set of the challenge. Codes are available at: https://github.com/m-arda-aydn/SegResNet-based-Reciprocal-Transformation .