Limitation of resources offers opportunities for AI-enabled, scalable tools to be deployed, especially in developing economies. This can impact clinical decision-making as well as healthcare delivery in the near future.

By Suresh Ramu

The rate of new data and research being churned out today is unprecedented. There is a need to be able to adapt this to clinical management and treatment rapidly. Decision-making tools that use Natural language processing (NLP) or other parts of Artificial Intelligence (AI) are being welcomed by leading clinicians.

AI is rapidly being deployed into clinical domains, such as decision making using data and patterns. For example, in radiology and histopathology, especially in developing countries like India where resources are not easily available.

The advantages
AI is always expected to perform at a superior level as compared to human models. For example, errors of 10 percent in reading radiology images, especially if these are false negative (wherein you miss reporting an anomaly) would not be acceptable. But similar errors by radiologists would not be easily measurable or even considered important. So AI algorithms that get approved to be used can be considered reliable, scalable, and cost-effective.

Similarly, because these tirelessly and non-emotionally keep learning (as compared to humans), the adoption of AI will continue to increase. We have seen a lot of this already embedded in medical equipment and medical imaging software.

Resources constraints are opportunities for AI-enabled, scalable tools to be deployed, especially in developing economies. This can impact the way we see clinical decision-making and healthcare delivery in the near future. In my opinion, the areas of handheld point of care diagnostics, wearables, and low-cost imaging will be the most exciting areas for AI deployment.

The challenges
During the current pandemic, AI has been helping us identify a lot more patterns of disease transmission, mechanism of action, treatment, and new drug and vaccine development. AI algorithms are trying to find ways to use X-ray and CT scan images to diagnose COVID-19 non-invasively. Yet, the nature of the virus and its mutations, the diversity in epidemiological responses across different regions in the world has not made it easy.

These concerns are very much applicable in cancer care as well. The complexity of various types of cancers, the biology of each cancer and our knowledge of it, and the long-term clinical follow-up required to prove the efficacy of AI algorithms are important challenges to overcome.

Public health institutions have a scale, size, and volumes of patients that could be enormous engines for the development of automation tools. However, many of them lack the discipline of management and the technology or clinical expertise required for this.

The private healthcare delivery sector in India has played that role and herein lies the opportunity for collaboration to accelerate the development and deployment of AI tools. This requires finding support, easier regulation for the public institutions to work with health tech startups (not necessarily tier 1 suppliers), and legal frameworks that allow for the data to be used and consumed for the public good. Data privacy concerns can be mitigated with better governance.

(The author is Co-founder of Cytecare Cancer Hospitals, Bengaluru, India)