Brain Tumor Diagnosis through Deep Learning

Authors- Er. Susil Bhatta, Sharad Kumar Ghimire

Abstract-– Brain tumors are one of the most critical and life-threatening conditions that require prompt and accurate diagnosis for effective treatment. Recent advancements in deep learning have shown promising results in the field of medical diagnostics, particularly in the automated detection and classification of brain tumors from medical imaging data. This studyfocuses on developing a robust and reliable model for brain tumor detection and compare it with pre-trained deep learning models such as InceptionResNetV2. We utilized a comprehensive dataset of magnetic resonance imaging (MRI) scans to train and evaluate various Convolutional neural networks (CNNs). The pre trained models were fine-tuned to adapt to the specific characteristics of the brain tumor dataset, enhancing their ability to accurately identify and classify tumors. The performance of these models was compared using key evaluation metrics, including accuracy, precision, recall, and F1-score. The base model was trained for 20 epochs with a batch size of 32, while the InceptionResNetv2 was trained for 10 epochs under the same batch conditions. Notably, the InceptionResNetv2 exhibited significantly lower training and validation losses of 2% and 1%, respectively, compared to the base model’s losses of 7% and 6%. Both models achieved an overall accuracy of 95%, indicating their comparable effectiveness in classifying instances correctly. However, the base model outperformed the InceptionResNetv2 in precision (96% vs. 94%), suggesting a lower rate of false positives, whereas InceptionResNetv2 demonstrated superior recall (95% vs. 93%), indicating its ability to capture a greater proportion of actual positive cases. Both models attained an identical F1 Score of 94%, reflecting a balanced performance in terms of precision and recall. These findings suggest that while InceptionResNetv2 shows better training and validation performance with enhanced generalization capabilities, the base model is preferable when minimizing false positives is crucial.

DOI: /10.61463/ijset.vol.13.issue2.334