An Efficient Level Set Speed Function Based on Temperature Changes for Brain Tumor Segmentation

Abstract

In clinical routine, accurate segmentation of brain tumors from Magnetic Resonance Images (MRI) plays an important role in diagnostic; it is a challenging and difficult task as brain tumors have various appearance properties. In this study, a modified level set speed function for accurate brain tumor segmentation applied on thermal images to reinforce brain tumors segmentation using MRI is presented. Tumor cells have high temperature compared to healthy cells, due to the high metabolic activity of abnormal cells. To calculate the thermal image we have used Pennes BioHeat Transfer Equation (PBHTE) resolved using Finite Difference Method (FDM). By analyzing the tumor thermal profile, the temperature is higher in the tumor center and is reduced as we move to the tumor borders; we have used this physical phenomenon in level set function for tumor segmentation. The proposed approach is tested in synthetic MRI images containing tumors with different volumes and locations. The obtained results showed that \( 10.29\,\% \) of brain tumor segmented correctly by level set method in the thermal image as a tumor part, contrarily in T1 which is segmented as healthy tissue, the same for T1c and Flair with \( 4.32\,\% \) and \( 22.58\,\% \) respectively. Therefore, the temperature can play an important role to improve the accuracy of brain tumor segmentation in MRI.