NEW DELHI – India, which has the highest incidence of tuberculosis (TB) in the world and the second highest of COVID-19, is looking to artificial intelligence (AI) to help detect and classify cases of both and lower the cost of diagnosis.

Lockdowns and fears of COVID-19 contagion have led to a disruption in TB diagnosis and treatment services in the mostly low- to middle-income countries with the highest burdens. Patients often are tested for either of the two conditions, which have similar symptoms, such as persistent fevers and coughs. However, co-infections often are missed.

Deep learning software, when used in diagnosis, could help address this challenge. With that in mind, India is tapping into its strengths in information technology.

One example is the work of Mumbai-based Qure.ai, which specializes in deep learning algorithms to interpret radiology images and has developed Qxr, a chest X-ray screening tool built with deep learning to automate the interpretation of chest X-rays in less than a minute.

The Qxr tool classifies chest X-rays as normal or abnormal, identifies abnormal findings and highlights them on the X-ray image. It can be used as a point-of-care screening tool, followed by bacteriological confirmation, to speed up TB diagnosis from several weeks to a few hours, said Qure.ai’s CEO and founder Prashant Warier. The algorithm is trained to detect classical primary pulmonary TB and simultaneously screen for chronic obstructive pulmonary disease, lung malignancies in high-risk populations and some cardiac disorders. Qure.ai also has developed Qscout, an AI-powered pandemic response app built on the company’s experience on an app used for global TB screening.

The potential of AI, especially deep learning, to increase TB detection rates was discussed at the annual conference of the International Union Against Tuberculosis and Lung Disease last October. Over the past two years, organizations such as PATH, FHI 360 (formerly Family Health International), Philippines Business for Social Progress and other programs funded by the U.S. Agency for International Development, TB REACH and The Global Fund have adopted Qxr technology, said Warier.

Multiple studies conducted by the Stop TB Partnership and the McGill International TB Centre have found that Qxr outperformed other algorithms and resulted in cost benefits of between 55% and 65%.

PATH, one of the early users of AI in TB, has worked with Qure.ai for almost two years. For instance, as part of an 11-month program in Maharashtra state, Qxr TB was used to screen more than 9,000 people and led to a 20% increase in notifications, a 13% increase in case detections and a 50% rise in cases confirmed by microbiological culture tests.

Another company, Pune-based Deeptek Medical Imaging Pvt. Ltd. has screened about 100,000 people for TB using AI under an ongoing project to free Chennai and its population of 7 million from the disease using mobile diagnostic vans and an AI-based tool to screen for underlying chest disease and generate a radiology report. The tool helped evaluate chest X-rays and classify them into patients likely or unlikely to have TB, Deeptek founder Amit Kharat told BioWorld.

“The diagnosis time is drastically reduced,” Kharat said. AI facilitates accurate, instant diagnosis, eliminating the need for a detailed report from radiologists.

Deeptek had carried out a retrospective study in Nanavati Hospital, Mumbai, on 9,098 chest radiographs of 3,180 patients and found that “our AI models matched the radiologist’s performance in the diagnosis of COVID-19 from (chest X-rays) and obtained sensitivity matching (that of) imaging experts,” Kharat said.

Kharat added that unlike RT-PCR tests that give positive and negative results for COVID-19, AI tools help quantify the amount of lung tissue affected by disease and correlate it with morbidity, aiding in better decision-making related to treatment, such as whether to admit a patient to a hospital.

AI screening of chest X-rays for COVID-19 and integrating with structured reporting “can be an excellent tool” to streamline workflow and hasten report turnaround time, says Kharat.

The Deeptek study showed that AI screening of chest X-rays sped up turnaround times and can be a useful prescreening tool.

Another company working on this space is Hyderabad-based Docturnal Pvt. Ltd., which developed TimBre, a screening app for lung TB that has been used across 25 pilot studies on 5,000 subjects, said founder Rahul Pathri. The company developed the Covawe algorithm for COVID-19 that was used from May to July 2020 to screen 2,000 people at home. The model showed 77% sensitivity and 86% specificity.

“We now plan for a bi-directional screening mechanism with a single cough sample screening for both TB and COVID-19 in one go,” said Pathri.

And AI-based tools can be affordable. For example, a single screening using TimBre costs INR50 (US$0.68) per screening, excluding hardware filters or surgical masks. “Covawe is free for now, and we plan to charge INR25 per screen post clinical trials,” Pathri said.