Non-invasive gliomas grading using swarm intelligence algorithm / Muhammad Harith Ramli

Gliomas is a malignant brain tumor that can be graded into four grade which are grade I, II, III and IV. Nowadays, the invasive technique used to diagnose the gliomas is found to lead to certain errors. The other problem on current diagnosis method is time consuming in order to manually annotate and segment the MRI image of gliomas.
Thus, this study of non-invasive gliomas grading using Swarm Intelligence algorithmis proposed to grade the gliomas into its grade. The developed prototype might assist the doctors and radiologists to diagnose the gliomas without consuming too much time and reduce the error in diagnosis. Framework development methodology is used to make sure the prototype is successfully developed according to the schedule. The basic feature extraction of minimum, maximum and mean of gray level values are used as the parameter to develop the prototype. Swarm intelligence (SI) algorithm is implemented because there are lot of previous works which prove that the SI is good for segmentation and classification. Bat algorithm is chosen for the development of the prototype for segmentation and classification purpose. There are five general steps implemented to develop the prototype. From the testing conducted, it shows that Bat algorithm produced moderate performance towards gliomas grading and segmentation. A few future recommendations such as Gray level co-occurrence matrix, Markov random field or Gabor filter to improve the feature extraction results. The other recommendation are to increase the number of random pixels for initial population step of Bat algorithm. The developed prototype is useful for the doctors and radiologists to grade the gliomas non-invasively and reduce the time taken to diagnose gliomas in the future.