Delivering AI in Healthcare – Platforms vs Packages? Why not both!

Author(s):

  • Becks Simpson, Machine Learning Technical Lead at Imagia 

Personalized medicine is touted as the holy grail of patient care. By supercharging decision support so that the right treatments are prescribed at the right times and diagnoses can be made based on all available data intelligently analyzed, we stand to improve patient outcomes and reduce healthcare costs as well.

This explains the excitement surrounding big data and machine learning initiatives in healthcare in the last few years. With the wealth of data available in hospitals to train on and the cutting-edge algorithms being produced that can distinguish things computers have not been able to before, the push for personalized medicine is greater than ever. Digital health companies must be able to deliver their software successfully into the hospital infrastructure and into the clinical workflow. If they fail to do so, they won’t be able to leverage the data available, train AI to help support clinicians and make an impact on patient outcomes. Herein lies the challenge. Should the software be delivered as discrete, bit-sized portions, typically bundled as packages? Or should they go all in and take a platform approach; an all-in-one end-to-end kind of deal?  Given the advantages and disadvantages of both approaches alone and how complementary the two are, as we’ve found at Imagia, it’s actually better to do both!

The upsides and downsides of packages

Some companies have already started to deploy AI solutions for precision medicine into the clinical context. The most common way they deliver software into a hospital is to distribute the code as a package.  Basically this means smaller, bite-sized pieces are deployed which perform a single task or limit scope of functionality. For example, an AI driven tool for radiology might only allow for smart nodule detection and classification but not interfacing with other forms of data such as reports or electronic health records.

The main benefit of this is that doctors and hospitals can pick and choose which pieces of the workflow they would like to augment without needing to buy a heavy program that is over-engineered for what they want to achieve. However, this type of software will often have its own widget or display which means that the clinician has another window to open and switch between during their workflow. This becomes easily burdensome and adds to the user fatigue already experienced by medical practitioners. Additionally, as the number of distinct software programs in use increases, the more likely there are to be incompatibilities between them. This can also mean that if more functionality is needed, they need to request it through one of the providers if the packages they have cannot function correctly together.

 

Bundling into platforms helps a bit

In an effort to get around these downsides, some companies have turned towards offering their tools and algorithms bundled together as platforms. Compared to packages, platforms provide a cohesive end-to-end experience and don’t suffer the issues of incompatibility that packages sometimes do. Platform-based software can reduce user fatigue as the interactions are through a single user interface with little to no switching between different tools. Since all of the functionality from start to finish of a clinician’s workflow is captured in the platform, there is no longer a concern of incompatibility between pieces.

However this also means that should a user require a new feature that the software provider does not plan to build, they have little recourse to find an alternative. Another downside is that buying a platform is an ‘all or nothing’ solution where the user is forced to accept all the pieces even though they might not need them. Imagine a scenario where an oncologist has some excellent software for dose management and prediction then a new platform based software program is brought in to process patient data, allow for automatic reporting and predict appropriate treatment as part of decision support, but their dose management is terrible! Unless the platform has some stellar integrations, it’s likely that the oncologist will have to either ditch their great, working tool for a sub-par one, or try and manage the interactions between the two themselves. It seems that both packages and platforms both present with shortcomings that may make them imperfect for delivering AI to drive personalized medicine.

 

So why not do both! The benefits of taking both approaches

Given that the advantages of packages and platforms when used separately are complementary, it makes sense to use an approach to software delivery that leverages both in order to mitigate the downsides and take full advantage of the benefits.

At Imagia, we have found that the best solution is to build a platform out of modular pieces that can work individually, or in small groups or, when combined all together, form an end-to-end product. As it is still one piece of software from a single provider versus many from different providers, it is easier to maintain, and has less user fatigue from many widgets and thus better workflow efficiency. The bonus however is that sites can optionally install only the portions of the workflow that a user actually needs. By building a platform-like product out of packages, ensuring standardization and security between components also becomes a non-issue. The modular pieces are designed to integrate with each other or exist standalone so every integrated piece has well-defined interfaces, looks unified when combined together and has rules for integration which makes using it a more seamless experience. It also becomes a one stop shop for all the pieces one might use in their clinical workflow from data access, annotating and measuring, reviewing AI predictions etc, without having to worry about whether the new pieces fit with the existing ones.

 

The pursuit of personalized medicine

As we move towards a future where more and more clinical sites and practitioners take the step towards incorporating AI-driven software into their workflows in the pursuit of personalized medicine, the more solutions and products we will see in the market being deployed into hospitals and medical clinics. With all of these options available, it is more important than ever to work with software providers who understand the implications of delivering with a package-based focus or a platform-based one and especially, those who work at the intersection of both approaches.

To learn more about Imagia’s products and solutions and how you can infuse value throughout your clinical workflows with powerful AI capabilities, get in touch with us by writing to [email protected]  or  completing this form.

Related posts

Application of Homomorphic Encryption in Medical Imaging

Application of Homomorphic Encryption in Medical Imaging

A technical 20-page report for next-generation data governance models. Authors: Francis Dutil, Alexandre See, Lisa Di Jorio and Florent Chan

...
Read more
Preserving Patient Privacy with Homomorphic Encryption

Preserving Patient Privacy with Homomorphic Encryption

Health research necessitates access to vast quantities of medical data and personal health information (PHI). From a patient perspective, pr

...
Read more
Delivering AI in Healthcare - Platforms vs Packages? Why not both!

Delivering AI in Healthcare - Platforms vs Packages? Why not both!

Personalized medicine is touted as the holy grail of patient care. By supercharging decision support so that the right treatments are prescr

...
Read more