Leprosy can cause irreversible nerve damage if it isn’t diagnosed and treated in time. Unfortunately, leprosy is often underdiagnosed or misdiagnosed because many medical professionals don’t have the training to recognise it. But lately, there have been some exciting new developments in this field. Here’s how emerging artificial intelligence and machine learning technologies are helping to improve leprosy diagnosis.
MaLeSQs: Analysing LSQ Answers with Machine Learning1,2
The Leprosy Suspicion Questionnaire (LSQ) was developed by the Brazilian Health Ministry to help with the timely diagnosis of leprosy. The LSQ acts as a screening tool to detect new cases. It is composed of 14 simple yes/no questions that the patient can answer with the help of a health professional.
The LSQ asks the patient to mark whether they are experiencing the following 14 symptoms:
1. Do you feel numbness in your hands and/or feet?
2. Tingling (pricking)?
3. Anesthetised areas in the skin?
4. Spots on the skin?
5. Stinging sensation?
6. Nodules on the skin?
7. Pain in the nerves?
8. Swelling of hands and feet?
9. Swelling of face?
10. Weakness in hands?
11. Hard to button shirt? Wear glasses? Write? Hold pans?
12. Weakness in feet? Difficulty wearing sandals, slippers?
13. Loss of eyelashes?
14. Loss of eyebrows?
While developing the questionnaire, experts noticed that a specific combination of marked symptoms was related to new cases. This has led to the development of a machine learning tool called MaLeSQs, which can determine whether or not a patient needs to be further evaluated for leprosy based on their combination of answers. By using a method called “Shapley values”, the MaLeSQs developers also figured out which symptoms/answers are most important for detecting leprosy, with nerve-related symptoms standing out as the most important.
After a patient fills out the questionnaire, MaLeSQs uses machine learning to analyse the answers and determine whether or not they are likely to have leprosy. The tool automatically processes the responses and tells the health professional whether the person should be tested further. If it identifies a patient as “LSQ Positive,” it means they might have leprosy and should go for further evaluation. If the person is “LSQ Negative,” they don’t need further tests for leprosy – they most likely have a different condition.
MaLeSQs has shown good sensitivity and specificity. During testing, it correctly identified 85.7% of actual cases and correctly ruled out 69.2% of non-cases. It also had a 98.3% rate of confirming cases when a patient didn’t have leprosy.
Using MaLeSQs to assist with leprosy diagnosis reduces the need for expensive tests, trained specialists and complex equipment, making it a practical and affordable solution for diagnosing leprosy in remote or resource-limited areas.
Comparisons with other diagnostic methods, like blood tests and image-based tools, showed that MaLeSQs is both effective and cost-efficient. While image-based tools may achieve higher accuracy, they require many more resources, whereas MaLeSQs relies only on a simple questionnaire and machine learning, making it very easy to use.
Experts in Brazil hope to expand testing further, to validate the effectiveness of MaLeSQs and improve its accuracy even more.
AI4Leprosy: Harnessing the Power of Imaging Tests3,4
The Novartis Foundation is collaborating with Microsoft AI4Health and the Oswaldo Cruz Foundation (Fiocruz) in Brazil on the AI4Leprosy initiative. They aim to develop a screening assistance tool to accelerate leprosy diagnosis.
The AI4Leprosy tool relies on imaging test results – photographs of skin taken under controlled conditions, like in a studio setting using high-quality cameras. These images focus on the visible skin changes caused by leprosy, such as spots or lesions. The tool then uses AI techniques called neural networks to recognise patterns in the images and identify subtle signs of leprosy that might be missed by the human eye.
After processing the images, AI4Leprosy uses other machine learning methods to analyse the results and make predictions about whether the patient is likely to have leprosy.
AI4Leprosy achieved near-perfect results in a controlled setting, with a sensitivity of 89% (catching almost all real cases) and specificity of 100% (no false alarms).
Despite its high accuracy, this tool has some limitations. Unlike MaLeSQs, AI4Leprosy requires specialised equipment (like high-quality cameras) and controlled environments to take the images, which may not be practical in remote or resource-limited areas. Until imaging technologies catch up, AI4Leprosy is better suited to settings where imaging resources and professionals are available.
New Technologies Give Us Hope for the Future
References:
1. Mendonça Ramos Simões M, Rocha Lima F, Barbosa Lugão H, et al. Development and validation of a machine learning approach for screening new leprosy cases based on the leprosy suspicion questionnaire. Sci Rep. 2025;15(1):6912. doi:10.1038/s41598-025-91462-6.
2. Leprosy suspicion questionnaire (LSQ). Figshare. doi:10.1371/journal.pntd.0009495.g001.
3. AI4Leprosy. Novartis Foundation. Accessed March 21, 2025. https://www.novartisfoundation.org/transforming-population-health/healthtech-innovation/ai4leprosy.
4. Barbieri RR, Xu Y, Setian L, et al. Reimagining leprosy elimination with AI analysis of a combination of skin lesion images with demographic and clinical data. The Lancet Regional Health – Americas. 2022;9:100192. doi:10.1016/j.lana.2022.100192.