A healthcare worker in Kurukshetra, Haryana, goes door-to-door recording cough sounds of affected people on her smartphone app. The AI-enabled tool runs it through its database of cough sounds of tuberculosis (TB) patients and indicates whether a person is likely to be positive or not. The worker sends the suspect patient information to higher centres for confirmatory testing.
With the Cough Against TB app having already screened around 1.62 lakh patients — in Haryana’s Kurukshetra during the government’s 100-day campaign last year and Mumbai and Mizoram during active case-finding campaigns in 2023 and 2024 — there has been a 13 per cent increase in the number of people diagnosed as compared to normal screening under the national programme. While the tool is still being validated by the Indian Council of Medical Research (ICMR), a pilot deployment has already started in 91 districts across 17 states.
Artificial intelligence is being leveraged in the health sector to reach people in remote areas, enable quicker diagnoses and ensure better treatment outcomes. Government hospitals and national institutes are systematically working towards identifying gaps in national programmes and addressing them with technology. This might help in building last mile access in the country’s public health infrastructure, given the community reach of ASHA (Accredited Social Health Activist) and ANM (Auxiliary Nurse and Midwife) workers. “AI solutions can reach people in the remotest parts of the country. So, we have them in several programmes, including screening for breast cancer, non-communicable diseases, oral health and dermatology,” says, Dr Radhika Tandon, professor of ophthalmology at AIIMS, whose team worked to develop the MadhuNetrAI, the app that evaluates retinal images and identifies early indicators of diabetic retinopathy.
Multiple AI-based solutions have been integrated into the national programme for TB elimination. Both Cough Against TB and AI-enabled X-ray devices can be used for active case-finding in the community instead of waiting for patients to show up for testing.
An AI-enabled hand-held X-ray device, a camera lookalike that can take X-rays within seconds and interpret them, has been effective in primary health centres that have an X-ray machine but still don’t have a radiologist. X-rays can detect white patches on the lungs, indicative of TB, before a person develops symptoms like coughing. At least 473 such portable devices are in use with 1,500 more being added soon. Several of these devices, including those developed by Indian companies, have already been approved by ICMR and have helped in the detection of nearly 2.85 additional asymptomatic cases.
The government’s telemedicine platform e-Sanjeevani — which connects patients at primary health centres to specialists from secondary, tertiary or super-speciality hospitals — now comes with an AI model to guide doctors in their diagnosis. It has provided consultations to over 43 crore patients through its 1.36 lakh spokes at primary health centres and 18,000 hub hospitals.
A portion of this data, based on consultations of over six crore people, has been used in the AI-enabled Clinical Decision Support System (CDSS). This can recognise 300 symptoms for the most common diseases treated on eSanjeevani, such as respiratory tract infections, gastritis, fever and diabetes. “With high accuracy in its top three recommendations, it acts as a smart filter that reduces doctors’ workload. Importantly, doctors maintain complete control; they can accept or reject AI suggestions and their feedback continuously improves the system,” say health ministry sources.
All it takes is for a healthcare worker at the primary health centre to fill out a patient assistance form, listing patient details and symptoms. Then the model guides the health worker through a series of multiple-choice questions to get further details on duration, severity and location of the symptom. Based on this, the system makes a probable diagnosis that can help doctors at hub hospitals offer appropriate treatment and advice.
Once disease surveillance teams scoured newspapers for data. Now an AI model has been developed to scan through news articles in 13 different languages. The model looks not only for reports of an outbreak but also the mention of any unusual cluster of disease or uncommon symptoms. “Earlier, the model would also flag traffic crashes or deaths due to natural disasters. Over time, the algorithm has learnt what we are looking for and accurately flags news of several people getting diarrhoea or fainting,” says Dr Himanshu Chauhan, head of the Integrated Disease Surveillance Programme.
The model has led to a 150 per cent increase in the number of alerts, which are then investigated. “This way we can use the skills of the officers for other activities such as planning a response instead of poring over newspapers,” says Dr Chauhan.
AIIMS doctors, along with the health and education ministry, have developed an AI-based fundus camera that can quickly detect damage to the retina and help in the early diagnosis of diabetic retinopathy — a condition caused by years of high blood sugar levels that damages the retina and can lead to blindness.
“Several primary health centres already have fundus cameras. The AI model can be plugged into any one of them. This means doctors have to invest their time only on patients who are likely to have the condition rather than screening every diabetic,” says Dr Tandon.
She says that the challenge with diabetic retinopathy is that patients may not realise that there has been damage to their retina till it is in an advanced stage. Many, in fact, come in with vision loss. While there are therapies to prevent the damage to the retina, once damaged it cannot really be reversed. “The idea is to have the device at primary health centres where patients are screened and treated for diabetes, so that they also get tested for diabetic retinopathy at the same location at the same time,” says Dr Tandon. Her team is also working on similar tools for other diseases such as glaucoma.
