Tuberculosis Detection Using ResNet50: A Deep Learning Approach for Medical Image Classification

In this study, a robust automated system for TB detection from chest X-ray images using deep learning techniques is developed. The dataset, sourced from Kaggle, comprises 4200 images divided into two classes: Normal and Tuberculosis. Image preprocessing steps included resizing images to 256x256 pixels, the normalization of pixel values, and data augmentation techniques to augment data variability and help to avoid overfitting. The dataset was divided into 80% for training and 20% for validating the model. The final accuracy of this fine-tuned Resnet50 model for this task was 91%. A fully connected classifier was used as part of the architecture and performed well in binary classification. Precision (80%), recall (90.9%), F1-score (85.7%), and a confusion matrix were derived for performance metrics of the model, which demonstrated high recall (90.9 %) for TB cases in timely diagnosis. Yet, with a lower precision (81.2%), it is necessary to reduce false positives. Deep learning, as highlighted in this research, is a scalable and accurate solution for TB detection. Future work will address increasing precision, dataset diversification, and the validation of the model for clinical applicability to a broader population. The results suggest the capabilities of AI in bringing augmentation to healthcare efficiency and hastening the path towards better patient outcomes.