AI applications to PACS systems – challenges and understanding the impact on workflows

“It is essential for healthcare organisations/hospitals to understand the performance and impact of AI solutions prior to placing them into the production workflow”, said Omar Sunna, Director, Global Product Management, GE Healthcare.
09:25 PM

The use of AI in medical imaging or Picture Archiving and Communication System (PACS), such as GE’s Centricity Universal Viewer, can be a great way to reduce the ever-increasing workloads of radiologists and clinicians, since AI is particularly effective in detection and segmentation of any irregularities in imaging data and providing radiologists/clinicians additional diagnostic confidence. However, this is easier said than done and many healthcare organisations/hospitals, depending on their existing infrastructure and data maturity, may find it daunting to implement AI applications in their PACS systems. 

Common challenges in implementing AI applications to PACS systems

“One of the common challenges faced by healthcare organisations/hospitals in looking to implement AI applications to their PACS systems is the handling of multiple point solutions of algorithms without a consistent way to implement these AI solutions into the overall workflow,” said Omar Sunna, Director, Global Product Management, GE Healthcare.

It is essential for healthcare organisations/hospitals to understand the performance, impact, and expected outcomes from the various algorithms in AI solutions prior to placing them into the production workflow. Understanding that Pneumothorax impacts nearly 74,000 Americans every year, GE healthcare has recently introduced an AI algorithm that could help radiologists prioritise the review of critical cases. This algorithm was designed to identify pneumothorax on a chest X-Ray and notify a radiologist to this potential finding, the expected outcome is to speed up the timely diagnosis of a potentially life-threatening condition. 

Another challenge for healthcare organisations/hospitals will be for them to understand which AI solutions are going to be the most impactful to the overall reading workflow and patient care, given the wide plethora of options available in the market. “There is a need to have a pre-production evaluation mechanism to all clinicians to examine the overall accuracy and repeatability of the AI solution before promoting into clinical practice and integrating the algorithms in their clinical workflow,” added Sunna.

Tips on building a tool kit for radiologists for imaging analysis/diagnosis

Sunna recommends that radiologists stay focused on patient outcomes as part of their selection of AI tools for their workflow and ensure that the solutions that are adopted are not just “point” solutions for a given output, but systemically implemented as part of the larger workflows in a healthcare setting. 

 “This means that the AI solution is not just presented to the viewer as additional views for the radiologist but has a number of solutions that can be implemented in the workflow to help drive proper critical condition triage, prioritise the right examinations to be read, and close the loop well with the clinicians about how to best convey this information in a rich report,” he said.

Ideally, the implementation of AI solutions in PACS systems would provide the radiologist additional diagnostic confidence and also to help triage critical cases in order to have them read by the right radiologist at the right time.  

Observations on the future developments of PACS and imaging solutions in healthcare

Having almost two decades of experience in the field of PACS and imaging solutions in the healthcare industry, Sunna believes that the use of AI will progressively become the standard of care to ensure that certain patient clinical conditions are well understood at the time of diagnosis.  

“There will also be continued efficiency gains in terms of identifying clinical conditions and automatically placing these findings into an enriched, structured report so that the clinicians can effectively consume a richer set of diagnosis. In addition, there will be an expansion of AI solutions into both Digital Imaging and Communications in Medicine (DICOM) imaging findings and also, looking at multiple disparate sources of data to enable better clinical decision support,” he explained.

Omar Sunnar is the speaker for Keynote Plenary 8 - GE Artificial Intelligence for Clinical Care - Powered by Edison at the upcoming HIMSS AsiaPac19 Conference in Bangkok, Thailand. His session is scheduled from 10.30am-11.15am on October 9, 2019 at World Ballroom B, Level 23.

GE Healthcare is a Diamond Sponsor for the HIMSS AsiaPac19 Conference. Their booth is located at G1 at the Exhibition Hall, Level 23.