/End-to-end-CNN-and-Hybrid-CNN-RF-Brain-Tumor-Detection

This project employs TensorFlow to develop a CNN-based brain tumor detection system. Moreover, a hybrid model, combining a pre-trained CNN as a feature extractor with a LightGBM Classifier, achieved even better performance, underscoring the efficacy of hybrid approaches in medical image analysis.

Primary LanguageJupyter NotebookMIT LicenseMIT

End-to-end-CNN-and-Hybrid-CNN-LGBM-Brain-Tumor-Detection

In this project, I've harnessed the power of CNNs within the Tensorflow framework to detect and classify brain tumors from magnetic resonance images. By meticulously applying data augmentation techniques, the CNN model achieved an impressive 97% accuracy on previously unseen test data. To comprehensively assess its superiority, I compared its performance with a LGBM classifier trained on hand-crafted texture features extracted from images. Intriguingly, the project reached new heights by deploying a hybrid model, employing a pre-trained CNN as a feature extractor and coupling it with a LGBM Classifier. This unique fusion outperformed both the standalone deep-learning model and the traditional classifier, solidifying the potential of hybrid approaches in medical image analysis.