Bi-LSTM neural network for enhanced radiotherapy: Detection of tissue parameters

Abstract

This study aims to simulate brain tissue and its elements using the MCNPX nuclear code and analyze X-ray emissions for tumor detection. This non-invasive approach offers a promising method for the accurate diagnosis of brain tumors, enabling timely treatment. Brain tissue, including tumor regions, was simulated at 70 keV, a common radiology energy. X-ray emissions were evaluated for variations in tissue density, elemental composition, thickness, and depth. The emitted X-ray doses were divided into 100 segments for detailed analysis. These data were then processed using an optimized Bi-LSTM neural network. This approach enabled precise segmentation and analysis of the X-ray data, improving tumor detection accuracy. Results showed significant differences in X-ray emission patterns between normal and tumor tissues. Tumor tissues exhibited distinct signatures, which were effectively captured and analyzed by the Bi-LSTM model. The model distinguished tissue types based on density, composition, thickness, and depth (1–5 cm). This study demonstrates that combining X-ray emission analysis with a Bi-LSTM neural network provides an effective method for brain tumor detection. The proposed approach may enhance non-invasive diagnostic capabilities and improve patient outcomes through early detection.