A Systematic Review of the Accuracy of Machine Learning Models for Diagnosing Pulmonary Tuberculosis: Implications for Nursing Practice and Implementation.
Kewalin Pongsuwun, Wimolrat Puwarawuttipanit, Sunisa Nguantad and 4 others
PMID 40058367WHAT IT FOUND
Machine-learning tools for pulmonary tuberculosis diagnosis mostly used chest x-rays or CT scans, with convolutional neural networks and support vector machines most common.
No patient outcome or nursing implementation effect was tested.
Key findings
01Twelve studies were included, and all targeted patients with pulmonary tuberculosis.
02Chest radiographs or chest x-rays were the most common feature, used in 5 studies (41.5%), followed by chest CT scans in 4 studies (33.3%).
03Convolutional neural networks were the most common algorithm, used in 5 studies (41.5%), and support vector machines were used in 4 studies (33.3%).
STILL TO COME
How it was doneWhat they foundWhat it means for RNs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The review included only 12 studies, and two did not specify sample size. The included studies did not use standardised evaluation metrics, so algorithm performance cannot be directly compared. The review did not analyse potential biases in data collection or algorithm design. The review did not explore resource constraints, infrastructure requirements, workforce readiness, or other implementation challenges. Only English studies published between 2019 and 2024 were searched, so earlier or non-English work may have been missed. Variations in patient populations and healthcare contexts across studies limit generalizability.
Declared interests
The authors reported no conflicts of interest and nothing to disclose.
The easy way to misread this
Do not read this review as evidence that machine-learning diagnosis improves pulmonary TB outcomes or nursing practice. It counts the models and features used in 12 studies, but it does not report pooled accuracy, patient outcomes, or implementation results.