Optimizing Tuberculosis Detection Through DenseNet Architecture and Ensemble Learning Techniques

Tuberculosis (TB) continues to be a major world-wide health concern. Immediate and accurate diagnosis is essential for both improving patient outcomes and controlling the illness. Sputum tests and chest X-rays are two examples of traditional diagnostic techniques that frequently depend on expert understanding and which can cause delays and irreg ularities. We describe an automated TB diagnosis method in this work that uses both machine learning and deep learning approaches to integrate clinical symptom analysis with advanced image processing. Our method combines lung X-ray pictures with patient symptoms which includes fever, cough, and weight loss. X-ray image and Symptoms is considered as two important data sources. We use a Convolutional Neural Network (CNN) that has been trained using pretrained models such as ResNet and EfficientNet, and optimized to perform medical imaging. We then enhance the CNN by applying transfer learning. In order to determine the probability of tuberculosis infection, we constantly evaluate the symptoms using ensemble machine learning models like Random Forests and Gradient Boosting Machines (GBM). Diagnostic accuracy has been enhanced by combining the results of the symptom and image analysis into a single predictive model. Our model was trained and validated on a wide range of X-ray images and corresponding symptoms, generating more than 96 percent accuracy in early tuberculosis detection. This approach appears to be very promising and demonstrates the advantages of combining diverse data sources for improved medical diagnosis, especially in settings where access to trained radiologists is minimal. The efficiency of this methodology highlights the importance of multimodal analysis in growing AI- powered healthcare solutions.