According to an article published in , adopting a responsive framework can help overcome challenges in applying AI to medical imaging. Canadian Journal of Cardiology.
The researchers, led by MD/PhD candidate Alexis Nolin-Lapalme from the University of Montreal, described an open-source software program they developed integrating AI models into PACS, called PACS-AI. This approach aims to enhance the evaluation of AI models by facilitating their integration and validation with existing medical imaging databases.
“It's like training a team. Training an algorithm needs to be well understood,” Nolen-LaPalme said. AuntMinnie.com. “It's like any medical tool. We need to understand when to use it and how to interpret the results.”
As AI continues to grow in popularity in radiology fields, some departments may face unique challenges in applying the technology to clinical images. These barriers include differences between healthcare system applications, reliance on proprietary closed-source software, and increased cybersecurity risks.
The researchers also highlighted that before AI models can be deployed in clinical settings, they must demonstrate their effectiveness in a wide range of scenarios and be validated with prospective studies. He added that the use of AI techniques in healthcare raises significant legal and ethical issues.
Nolin-Lapalme said a responsible framework is creating tools that will work equally well among clinical groups. This includes training AI models that understand or consider underlying biases when dealing with different patient populations.
“I think there's a lot of interest in AI. People think efficiency is very knowable,” Nolen-LaPalme said. “At times, the results may seem plausible, but critical understanding is key.”
In their review, the researchers describe PACS-AI, an open-source, vendor-agnostic software application that aims to enhance the evaluation of AI models by simplifying their integration and validation with existing medical imaging databases. The goal, the team wrote, is to offer a path toward responsible, fair, and efficient deployment of AI models in healthcare.
This platform acts as an interface between existing clinical PACS and AI models. The researchers highlight that its primary purpose is to allow the automated, near-real-time application of AI models to clinical images for use at the point of care.
The platform offers a web application interface that clinicians can use to search for imaging studies stored in hospital PACS and select a compatible AI model to apply to relevant images. The application backend then collects the relevant images, prepares the data, and calls for AI inference to be performed. The user is then presented with the results in a web interface.
Currently, the PACS-AI platform is only being used for research purposes in several Canadian and US hospitals. Once the models are validated and receive regulatory clearance, PACS-AI will also be used to deploy clinical AI models, the authors wrote. He also noted that clinical use would require regulatory clearance in both countries.
A full description of PACS-AI can be found here.