A Review on Tuberculosis Pattern Detection Based on Various Machine Learning Techniques

This paper examines various respiratory tuberculosis detection methods, including conventional and enhanced algorithms, across multiple datasets. It evaluates the impact of image enhancement on the preliminary processing of TB chest Xray images and examines various image preprocessing strategies. Three image enhancing methods—Unsharp Masking, High-Frequency Emphasis Filtering, and Contrast Limited Adaptive Histogram Equalization— were evaluated to improve lung segmentation and heat mapping accuracy. The deep feature enhancement network identified in this study significantly improves accuracy, achieving 82.9% Average Precision and 77.6% recall on the TBXllK dataset. The paper also explores an advanced visualization technique that ensures comprehensive analysis and a deeper understanding of the subject matter. The CBAMWDnet model demonstrated outstanding performance in TB identification across multiple datasets, increasing the likelihood of early evaluation and treatment, marking a significant milestone in TB diagnosis.