Evaluation of Gabor Filter, GLCM, and DWT Performance in Brain Tumour Classification
Keywords:
Gabor filter, GLCM, DWT, MRI Images, Brain Tumour, ClassificationAbstract
The brain, vital for body function, can be afflicted by tumours, potentially leading to death or uncontrolled growth, and metastases if left untreated. Hence, automated classification of brain tumour types is crucial for faster treatment, better planning, and patient survival, as manual diagnosis of brain tumour types heavily relies on the expertise and sensitivity of radiologists. Thus, this paper evaluates the performance of Gabor filter, Gray Level Co-occurrence Matrix (GLCM), and Discrete Wavelet Transform (DWT) in identifying normal and abnormal brain tumours using four categories of brain MRI tumours from the Kaggle database. The performance analysis focuses on binary classification to determine the efficacy of each feature extraction method. The study found that Gabor features had a False Positive Rate (FPR) of 7.61%, False Negative Rate (FNR) of 8.57%, sensitivity of 91.43%, precision of 81.36%, and accuracy of 92.13% at 985.34 seconds. GLCM features had an FPR of 9.69%, FNR of 9.52%, sensitivity of 90.48%, precision of 77.24%, and accuracy of 90.36% at 364.74 seconds. DWT features had an FPR of 11.42%, FNR of 11.43%, sensitivity of 88.57%, precision of 73.81%, and accuracy of 88.58% at 275.53 seconds. The GLCM yielded most efficient feature extractor, which can serve as a useful technique and second reader to radiologists in diagnosing a brain tumour to reduce the mortality rate.