An information theoretic framework for MRI preprocessing, multiclass feature selection and segmentation of PF tumors

Abstract

In our earlier works, we demonstrated that multiresolution texture features such as fractal dimension (FD) and multifractional Brownian motion (mBm) offer robust tumor and non-tumor tissue segmentation in brain MRI. We also showed the efficacy of these and other features such as intensity and shape factor to delineate cyst from tumor tissue segments. To achieve this goal, we obtained novel multiclass Kullback Leibler Divergence (KLD) feature selection techniques to effectively select features for tumor (T), cyst (C) and non-tumor (NT) tissues types in multimodal MRI. In this work, we propose an information theoretic framework for improved pediatric posterior fossa tumor segmentation. Our proposed method combines all necessary steps such as MRI inhomogeneity correction, feature extraction, multiclass feature selection and T, C and NT tissue segmentation respectively in an integrated framework. Our integrated framework allows one to observe effect of each step in the end tumor segmentation results. Finally, we evaluate our method using eight pediatric patients in T1, T2 and FLARI modalities.