Enhancing the Accuracy of Region of Interest Detection in Multi-Lesion Brain Tumors Using a Hybrid Deep Learning Network

Abstract

Precise detection of Regions Of Interest (ROIs) is essential in multilesion brain tumors for diagnosis and treatment planning. Most conventional methods, such as CNN and SVM, usually have a trade-off between precision and recall, misjudging tumor regions, and reducing diagnostic reliability. This paper proposes a Hybrid Deep Learning Network (HDLN) to address some limitations of conventional methods. The proposed HDLN combines CNNs for feature extraction with Long Short-Term Memory (LSTM) networks to integrate contextual information for more robust and accurate tumor region identification. This hybrid method combines the best of CNN and LSTM to improve accuracy in ROI detection. The proposed HDLN was thoroughly tested, indicating significant improvements over traditional methods, achieving a 0.15% increase in accuracy, a 0.12% increase in precision, and a 0.10% improvement in recall compared to CNN and SVM approaches. These advances indicate that the proposed HDLN can contribute to the field of ROI detection in medical images, offering clinicians a more credible and time-saving tool. Optimized parameters signified higher detection accuracy and reduced false positives and negatives. This paper describes the architecture, implementation, and thorough evaluation of the proposed HDLN, offering a prospective candidate to improve clinical workflow, diagnosis, and treatment planning for patients with brain tumors.