By: Mr. Raghavendra Goud Vaggu, Global CEO, Empe Diagnostics

The severity of tuberculosis (TB) makes it a troubling crisis across the globe, especially as it is responsible for millions of deaths around the world. According to the World Health Organisation (WHO), TB was responsible for 1.6 million deaths in 2021, making it the 13th biggest killer and second leading infectious killer only after COVID-19 that year. With 10.6 million people falling ill in 2021 to the dreaded disease, and patients cutting across all demographics, there is need for vigilance.

That same year, about 2.59 million cases were recorded in India by the WHO. At a staggering rate of 188 people per every 100,000, India has a lot to do. Thankfully, tuberculosis is both preventable and curable, especially if it is detected early enough and treated in good time. With emerging machine learning algorithms, TB diagnosis seems to have taken a new dimension with computer-aided diagnostics (CAD) measures now taking centre stage in many climes.

Machine learning and AI for TB diagnosis

The rise of computer-aided diagnostics has certainly added impetus to the drive for better TB diagnosis, especially because of better medical imaging that gives radiologists more precise interpretation of the patient’s chest, blood, spine, or brain, depending on the part of the body that is affected. One of such tools is the CAD model which offers precise diagnosis of the TB cavity and clearly displays areas of interest when observing the chest x-ray image. This is a huge improvement on preexisting CAD systems which could not identify TB cavities, as a result of the lung field’s superimposed anatomic parts.

Tools like deep learning, a component of AI, is also quickly becoming a trend among top diagnostics and health centres that has brought about increased diagnostic precision. TB detection has also been enhanced with more recent systems that support techniques like masking, texture analysis, as well as chest radiograph analysis. Accurate TB diagnosis has evaded radiologists and other diagnostic experts for many years due to the complex nature of the disease and how it can present in different parts of the body, but modern AI tools are changing the story.

Looking ahead in TB diagnosis and treatment

Within the framework of modern TB analysis, the place of data cannot be overemphasized. Fully automated CAD systems are being experimented for their great features, including handcrafted systems and deep features. These systems use pre-trained CNN frameworks and supervised learning to probe the parts of the body that are of interest using obtained data and processing such data to reach conclusive diagnosis. This also comes with the ongoing conversation around the difference between supervised and unsupervised learning and how the choice of the future can shape TB diagnosis. More advanced CAD systems will likely emerge in the future, powered by superior AI.