On Monday 14th of March 2022, a workshop was held on the role that Artificial Intelligence can play in the healthcare sector. This event was organized by AI4Belgium as part of the European AI Week and featured several speakers including Kevin Françoisse, CEO and co-founder of Sagacify.
Nowadays, the concept of artificial intelligence is widely used but not everyone knows what it is exactly, how it can be defined and how it differs from traditional computer programs. Softwares 1.0 require instructions that tell the computer what to do to execute a task. However, they run into problems when tasks are too complex. AI overcomes this pitfall as it does not need predefined rules. The computer is taught what he has to understand so that he learns by itself and is then able to reproduce the task.
Artificial intelligence can be seen as a mathematical function that maps any input to an output that helps make a decision or automates a task. The latest technology in the AI sphere is deep learning, which can take as input any type of signal, including semantically complex data such as images, text, and time series. Possible outcomes for the health sector cover detecting sepsis, identifying labels in medical imaging, or detecting anomalies in hospital processes. Another example that Sagacify has worked on is an invoice anomaly detection system which ensures that patients are not be charged unnecessarily and that there has been no error in the hospital system.
Opportunities are endless because AI can use any kind of signal, can analyze huge amounts of information in no time, produces very consistent and precise decisions, learns continuously and therefore improves over time.
More concretely, artificial intelligence can help hospitals cut costs by being more efficient at every level in the organization, from the patient’s admission to his journey through the various departments and all the associated back-office services. Most importantly, AI frees hospital staff from administrative tasks so they have more time to focus on patients and, down the line, this helps treat patients better and faster.
In practice, statistics show that only 15% of companies that tried to implement an AI project have been successful. This is because it is extremely difficult to deploy a system in production and to ensure that it generates the expected value. Keys to success are an experienced AI team and a proven methodology. Besides, it is good practice to start with a small project that is more likely to succeed and to build momentum for a series of AI projects. The idea is to start small but with a long vision in mind.
Another challenge is that AI is often thought to be dedicated to internet giants with millions of data points, but also to be expensive or complicated to deploy in production. However, AI is nowadays able to extract value from small datasets, provided that they are of high quality. The implementation of AI projects can therefore be eased by collecting high quality data, but also by having domain experts control the data engineering in collaboration with AI experts, and by using the latest tools such as MLOps platforms. Besides, if hospitals joined forces, they could have more data and reduce costs through economies of scale. The idea of federated learning is that different AI systems can feed each other to improve their performances, while staying independent.
Finally, it is important to take into account that no AI model will be 100% accurate. Depending on the AI’s level of confidence in its output and the cost of error, a quick human validation can be required. This feedback is extremely important as it allows the model to learn from its mistakes and improve over time.
We’ve seen that AI can be a powerful tool to optimize hospitals’ systems. The technology is ready to help hospitals reduce costs, free up time for practitioners, and therefore treat patients better and faster.
Now the only remaining question is: Where to start? Since Sagacify is specialized in helping its customers benefit from AI, it can help hospitals all the way from identifying opportunities to deploying the system in production and ensuring it is successful. Technology now contributes to developing a new patient experience, and hospitals should not hesitate to take the plunge if they do not want to be left behind.