Smart Diagnosis: AI-Driven Neural Networks for Accurate Medical Response Categorization
Main Article Content
Abstract
The rapid evolution of artificial intelligence (AI) in the healthcare sector has led to increased integration of machine learning tools in medical consultations. While this trend promises enhanced diagnostic capabilities, it also raises critical concerns about the reliability of AI-generated medical advice, particularly when trained on limited or inconsistent datasets. Inaccurate or ambiguous outputs from AI systems can jeopardize patient safety and reduce trust in technology-assisted healthcare. To address these challenges, MEDXNET is introduced as a novel AI-driven diagnostic classifier that distinguishes whether a given clinical response originates from a certified medical professional or from a generative AI system. MEDXNET is built upon a hybrid deep learning architecture that combines Bidirectional Long Short-Term Memory (BiLSTM), transformer-based attention mechanisms, and one-dimensional Convolutional Neural Networks (CNN1D). This architecture is designed to extract both local and global dependencies in complex medical text, improving contextual understanding and classification accuracy. Term Frequency-Inverse Document Frequency (TF-IDF) vectorization is used to convert raw medical text into meaningful numerical representations, enhancing feature extraction for the neural networks. The MEDIC dataset, which consists of annotated medical responses, serves as the primary data source for training and evaluation. To benchmark performance, the model's classification capabilities are compared with traditional architectures, including GRU, LSTM, and standalone BiLSTM. MEDXNET demonstrates superior performance in terms of accuracy and reliability, outperforming these established models. Additionally, a CNN2D variant of the architecture further improves performance, achieving a remarkable classification accuracy of 96.78%, which surpasses all baseline techniques. This intelligent system provides a safeguard for users by validating AI-generated medical outputs against known physician-generated responses, empowering users to make more informed healthcare decisions. By bridging the gap between human expertise and artificial intelligence, MEDXNET contributes to safer, more trustworthy medical interactions in an increasingly AI-assisted clinical environment.
Article Details

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.