AI in Healthcare – Snapshot Part I of VI


Of all the areas in which artificial intelligence can benefit the world, healthcare, in my opinion, is at the top. In the simplest possible terms there are two main reasons why artificial intelligence can be transformational in healthcare.

First, there is a lot that is broken and fixing that can help millions of people, reduce the cost of healthcare, improve patient care, and help fight disease.

Second, we need to accelerate our efforts in finding new and more effective cures for a long list of diseases – cancer, Alzheimer to name a few.

A Model for Healthcare

Besides being the editor of the AI Post, I also serve as the Executive Director of Society of Artificial Intelligence in Medicine and Healthcare.

Health CareI feel greatly blessed that I have had the opportunity to work in the entire value chain of the healthcare field. From preclinical to translational and clinical – and in hospital system administration – and with Accountable Care initiatives – and the launch of a payer company – my experience spreads the entire value chain. This broad experience of healthcare gave me a better appreciation for the need to approach solutions at an end-to-end value chain basis.

To approach the analysis of the industry on a value chain basis, I have developed a framework for capability development for artificial intelligence. Known as SAIMAH Framework (see figure 1), we use it to approach the marriage of various areas of the healthcare value chain with various solution types. NOTE: If you use this framework, please cite the SAIMAH website

Why the framework?

Imagine building a house where each part of the house (e.g. bathroom, living room, bedroom etc.) was designed by a different person and there is no coordination between the designers. Can you imagine what an awkward looking house you’d get? Both the IT approach to healthcare – and even the general business approach to the industry – has been like that. The lack of coordination across various areas of the value chain lead to tremendous waste in the industry and increase patient suffering.

Now that artificial intelligence will define the future technology growth of the industry, it is important to approach solutions from that perspective.

The SAIMAH AI Healthcare Framework

For the society, I have developed a framework for high level classification of companies (startups) in various areas. It is known as the SAIMAH Framework.

The classification scheme is important to understand and coordinate the activities across the entire field. An upcoming comprehensive report of the AI Post will use the same framework to study the market.

saimah framework

The importance of the SAIMAH Framework

When we try to optimize the performance of a system, in this case the healthcare value chain, we cannot be successful if we approach it by maximizing the performance of a single part. Just as our knowledge and experience from supply chain management tells us that we cannot improve the performance of a factory if we improve the performance of one area in manufacturing and ignore the rest (since it will only mean that the downtime of the next area would increase – the bottleneck phenomenon), healthcare will only improve when the entire system is taken as a whole and the efficiency of the entire system is increased.

For example, an area that I find of utmost importance is the divide that exists between preclinical and translational research. This is an area in dire need of attention and transformation. Preclinical research focuses on exploring early stage drug (technically what is studied at preclinical is not a drug) options and experiments are performed with mice, monkeys, and other animals. If a drug shows promise, it moves to the next level (translational) and after significant clinical trials and testing it finally gets approved for human use. Thousands of drugs start and only a few make it to the end. The ratio is 5000 to 1 (see NT Times article link below). It is a well-known fact among the research community, and is often characterized by the phrase “valley of death” – the misfortune of thousands of drugs dying down before ever reaching infancy is considered a necessary fateful journey. The key issue is that it doesn’t have to be.

The Sub-Optimized Value Chain

Even though it seems like that, but the failure that happens during the “valley of death” is not should not be viewed as that a particular drug failed during the human trials or was determined to be unsuitable for the human trial. The real question is that whether it was possible to have known that before even a drug started its journey? Could the whole wasted effort have been avoided?

The reality is that without proper data and collaboration across the industry, so much effort is wasted, so much repetition exists, and so many smart people work on projects with predictable outcomes that it is mind-boggling.  Lack of collaboration and not-invented-here syndrome are rampant. Big pharmaceutical and university collaboration is rare. In these silos, the research processes are primitive and even when they are advanced they are often driven by a desire to seek grant vs. to create real value.

As we go forward, we will publish several articles on these issues, but for now consider the impact that artificial intelligence can make in the research process:

  • Increase collaboration among universities and research centers all over the world
  • Expand, add content to, and globally standardize Molecular Imaging and Contrast Agent Database (MICAD) list. See Chopra et al. (Chopra et al. 2012)
  • Enhance collaboration between pharmaceutical companies and universities
  • Automated matches between grants and most needed or highest values projects
  • Forensic analysis to determine why drugs failed and then use the content to develop forward facing analysis
  • Organizing research

In future articles we will cover clinical and hospital systems.


Chopra, A., Shan, L., Eckelman, W. C., Leung, K., Latterner, M., Bryant, S. H., &Menkens, A. (2012). Molecular Imaging and Contrast Agent Database (MICAD): Evolution and Progress. Molecular Imaging and Biology, 14(1), 4–13.