AI is now assisting in surgery, acting as a virtual nursing assistant, and of course revolutionizing image analysis. Not only has AI itself improved significantly with deep learning, so has access and the tools for developing it. There are some amazing open source tools and platforms, like TensorFlow, out there allowing you to build and deploy your own AI models. So why don’t we all just build our own AI?
It’s just not that simple. Despite how fruitful these open source platforms are in their provision, to really make them work and maintain them you need a certain level of expertise. Not just the expertise in your own medical or scientific field, the only requirement for building AI models with Aiforia, but also knowledge of coding, data science, and software development.
Let’s open this up to a construction analogy. To build a house you need tools. A hammer, some nails, drills, etc. However, those tools aren’t enough. You need someone who knows how to use those tools such as a builder. You also need someone who can design the house, an architect or a developer. Not just to make sure it is aesthetically pleasing but also to ensure that it won’t collapse on you. In fact, safety and quality are probably the more important aspects of building a home.
Let’s move away from the construction site and go back to the computer lab. It is also an understatement to say that safety and quality are important when using AI in healthcare, whether it be research or in a clinical setting. A quality management system (QMS) must be in place to monitor and continually improve the performance of your platform. Both the QMS and technology must be maintained. The architects are needed. Well, the data or software architects are needed. Without their help how can a novice in AI have the assurance that what they have built is functioning well and not posing any risks?
Rather than just providing the building tools for you to use on your own, at Aiforia we have bundled everything you need from a computer science perspective together into an easy and ready-to-use platform. No coding, no local hardware, and no data science are needed to build AI models with Aiforia. Our platform just needs to know what you want to find. Annotation Assistant, our active learning tool, even guides you through the training process suggesting regions for you.
A turnkey solution built for medical and scientific users, Aiforia directly addresses the challenges of monitoring safety and efficacy in AI systems built broadly for healthcare usage. Our platform is proven and validated with hundreds of applications in a huge variety of medical and scientific fields, a continuously growing number: from analyzing prostate cancer samples to detecting and quantifying neurons in rat brains. Aiforia’s in-house scientists, with years of experience in pathology, work closely with customers, coders, and AI annotators to ensure the AI models are working as they should be. We also have close ties with the medical and scientific community. The headquarters are located in the middle of the Helsinki medical campus, downstairs from the Institute of Molecular Medicine Finland and next door to one of the largest hospitals in Europe. We are constantly collaborating with our neighbors.
The Aiforia Platform also follows a rigorous quality management system compliant with ISO 13485:2016; an internationally recognized standard. Since 2018 the externally validated Aiforia QMS ensures we are providing medical and scientific services consistently meeting customer and regulatory requirements. As AI becomes a more prominent tool in healthcare so should ensuring the quality of these AI-based devices.
Open source platforms are freely available. Data science, coding, quality management, and scientific support are not. So, while with that freely available code and tools you could try to grind it out to build your own AI models, those other pertinent factors need to be included in your healthcare-related work and sourced from elsewhere. Aiforia provides it all in one pace. All you need is internet to access it.
https://www.ncbi.nlm.nih.gov/pubmed/27730415
http://www.thespinejournalonline.com/article/S1529-9430(17)30851-3/fulltext