Automated brain tumor segmentation and classification for MRI analysis system

Abstract

<span>This paper proposed a new analysis technique of brain tumor segmentation and classification for Fluid Attenuated Inversion Recovery (FLAIR) Magnetic Resonance Images (MRI). 25 FLAIR MRI images were collected from online database of Multimodal Brain Tumor Segmentation Challenge 2015 (BRaTS’15). The analysis comprised four stages which are preprocessing, segmentation, feature extraction and classification. Fuzzy C-Means (FCM) was proposed for brain tumor segmentation. Mean, median, mode, standard deviation, area and perimeter were calculated and utilized as the features to be fed into a rule-based classifier. The segmentation performances were assessed based on Jaccard, Dice, False Positive and False Negative Rates (FPR and FNR). The results indicate that FCM offered high similarity indices which were 0.74 and 0.83 for Jaccard and Dice indices, respectively. The technique can possibly provide high accuracy and has the potential to detect and classify brain tumor from FLAIR MRI database.</span>

Automated tumor segmentation in MR brain image using fuzzy c-means clustering and seeded region methodology

“…Figure 3 Quantitative evaluation of the proposed methods applies two important statistical measurement parameters: Dice [28] similarity coefficient and Jaccard [29] similarity. The performance measures are defined as (11) and ( 12): Where A and B are the two binary images.…”

“…The proposed method provides accurate results compared to well-known fuzzy and other favored methodes for MRI segmentation, such as deformable models, and sets a standard for that type of approach. Norhashimah et al [11] introduced an analysis method of brain tumor segmentation and classification for MR image. To evaluate the proposal method twenty five images were used from BRaTS'15.…”