HYBRID SEGMENTATION FRAMEWORK ON BRAIN TUMOR DETECTION IN MEDICAL IMAGES

Brain tumor is an intracranial mass made up by abnormal growth of tissue in the brain or around the brain that limits its functionalities. Brain tumor diagnosis can be quite difficult because of its diversity in shape, size, and appearance and as a result, finding accurate measurement to its diagnosis can as well be critically difficult. This study developed hybrid segmentation framework for brain tumor images in medical imaging through the fusion of threshold and watershed approaches as the hybrid segmentation framework. The image was preprocessed using the Gaussian filtering technique for filtration. Enhancement was achieved using the image enhancement technique of MATLAB. The performance of the hybrid algorithm was evaluated based on Accuracy, Precision, Recall, F-measure, G-measure and False Alarm Rate. A comparative analysis was done to compare the hybrid, watershed, and threshold approaches based on the performance measure. The hybrid framework was found to perform better for all the performance measures with the accuracy value of 0.8250, precision value of 0.8889, recall value of 0.8571, F-measure value of 0.8729, G-measure value of 0.8729 and the false Alarm rate value 0.2500. Hybrid image segmentation framework was effective compared to watershed and threshold approaches and it is recommended for brain tumor analysis in medical image based on high value of accuracy.