Diagnosis for a healthy AI system

The use of AI in diagnosis can make a big difference to India's health parameters. For this to succeed though, the government needs a national effort to build electronic medical records

As acronyms go, some are rarely seen together. A recent development though has led to mentioning AI and TB in the same sentence. The ministry of health of India has decided to explore the possibilities of using artificial intelligence to combat tuberculosis in the country. 

In a development that could set the tone for healthcare policy in India, the government is recognising the importance of embracing new approaches. Health ministry has signed up with independent research body, Wadhwani Institute for Artificial Intelligence. The collaboration would include “supporting the National TB programme become AI-ready, which would include, developing, piloting, and deploying AI-based solutions,” the ministry says. 

AI is being accepted by doctors and scientists as an effective tool. The Radiological Society of North America recognised pioneering effort by two doctors that had used AI for TB.  Dr Paras Lakhani and Dr Baskaran Sundaram from Thomas Jefferson University Hospital, Philadelphia, tested the efficacy of automated detection of TB through chest X-rays. The Doctors used trained neural networks to study over 1,000 X-rays of potential patients. An accurate diagnosis of 96 per cent was achieved in the experiment. The algorithms can be trained to distinguish normal X-rays from those with specific conditions. These work faster with higher accuracy. Such solutions can bring improved diagnostic abilities to doctors with limited experience in remote locations. According to the World Health Organisation, TB is one of the top 10 causes of death worldwide. In 2017, 10 million people fell ill with TB, and 1.6 million died from the disease. India has the highest incidence of TB deaths in the world. 

In the private sector, the Centre for Advanced Research in Imaging, Neuroscience and Genomics (CARING) —the research and development wing of Mahajan Imaging, a leading medical imaging service providers, has established a global leadership position in the medical imaging AI space. CARING has created RAID — Radiology AI Deployment — a platform for hosting multiple AI algorithms for the purpose of validating and deploying them in clinical settings. RAID pulls all chest X-rays in real-time performed at one of Mahajan Imaging’s centres and automatically runs a high-sensitivity AI algorithm to detect whether the X-ray is normal or not. If the AI finds the X-ray to be normal, a report is automatically created, which a radiologist can verify before issuing to patients. Over the past few months, more than 5,000 X-rays have been read by this AI, saving critical time for radiologists. The algorithm is being implemented in seven other sites where RAID is installed.

The use of AI for diagnosis and predictive assessment can make a big difference to India’s health parameters. For this to succeed though, the government needs a national effort to build electronic medical records (EMR). Currently, such records of patients are fragmented and stored by various private and government healthcare bodies. Most are not stored in any common protocol. 

Algorithm can be trained to assess all medical records of a single patient and then make predictive analysis. At the national level the health ministry would have to create the relevant policies to collect and store medical data with appropriate privacy and regulatory standards. Using the new technologies can support public health policy in attacking various diseases at improved scale and speed. 

The shortage of experienced doctors and medical teams can also be addressed by technology. New models of heathcare are required where clinical diagnosis systems can be accessed by doctors in cities and villages. AI-based clinical diagnosis systems can support doctors even if they are not specialists. This can bring best in class knowledge to public healthcare. Hopefully the AI TB project is precursor to such a system.