Q-Learning-Based Approach to Detect Tumor in Human -Brain

• Published in: 2024 International Conference on Communication, Computer Sciences and Engineering (IC3SE) pp. 1 - 5
• Main Authors: Mani, Chandra, Aeron, Anurag, Rajput, Kapil, Kumar, Sandeep, Jain, Arpit, Manwal, Manika
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.05.2024
• Subjects: Accuracy; Brain; Brain tumor; Magnetic resonance imaging; MRI; Q Lerning; Q-learning; Sensitivity; Support vector machines; Transforms
• DOI: 10.1109/IC3SE62002.2024.10592886

The brain's own aberrant and unregulated cell division is what causes brain tumors. In the tumor growth exceeds 50%, the patient will not be able to recover. As a result, rapid and precise brain tumor identification is required. Brain image processing research started to focus heavily on brain tumors. Hence, a novel methodology of image processing is proposed for fast and accurate brain tumor detection. Reinforcement-based Q-learning technique is proposed for classifying brain tumor MRI images. Every MRI image is divided into 8 × 8 blocks and the features like mean, variance, skew ness, kurtosis, entropy, contrast, correlation, energy, and homogeneity are extracted. Ground truth matrix is generated by using discrete wavelet transform with which MRI image divided blocks are classified as 0 and 1 where 1 is the tumored block and 0 is the normal block. Reward mechanism in Q-learning is used to further classify the tumored and normal brain images. The efficiency of the classification is compared with other existing algorithms like Dense Net, ensemble methods, and SVM and CNN in accordance with performance parameters like accuracy, sensitivity, and specificity. The accuracy of the proposed algorithm is 97%, sensitivity is 96% and specificity is 97% proving that the proposed algorithm is highly efficient in classifying brain tumor MRI images.