How Mass General is using AI to improve its cardiovascular imaging workflows
Artificial intelligence (AI) is increasingly being incorporated into cardiovascular imaging, but at Mass General Brigham, some of the most practical applications are focused on streamlining the imaging workflow rather than replacing or double-checking what radiologists and cardiac imagers already do.
Brian Ghoshhajra, MD, MBA, medical director of radiology and clinical operations for integrated imaging services at Mass General Brigham and an associate professor of radiology at Harvard Medical School, discussed the health system’s approach to AI and automation with Cardiovascular Business.
Ghoshhajra said imaging can be viewed as a chain that begins with ordering and scheduling, moves through protocoling and quality assurance and ends with interpretation and reporting. While much of the attention surrounding AI has focused on interpretation, he said there may be significant opportunities to improve the earlier steps of the process.
At Mass General Brigham, for example, automation has been used to reduce the amount of manual work required for protocoling cardiovascular CT examinations. Ghoshhajra said the system has grown from performing about five cardiac CT examinations per day when he was a fellow 18 years ago to approximately 85 to 100 studies per day across multiple hospitals using a hub-and-spoke model.
He said looking at the patient record and reviewing the order to determine if the test is appropriate can eat up about two hours of a highly trained clinician’s time each day. With that issue in mind, AI was developed that allow referring physicians to select specific examinations and acquisition strategies. This has enabled the health system to eliminate about 90% of manual protocoling, according to Ghoshhajra.
Some examinations, such as calcium scoring, can move through the workflow without additional protocoling, while more complex studies can be ordered using standardized menus designed around the needs of referring specialists.
The same philosophy extends to patient preparation. Nurses follow standardized algorithms and instructions, allowing physicians to concentrate on decisions requiring their expertise.
“We're much more ready now to implement AI when we're using the physicians to do high-value work,” Ghoshhajra said.
The health system also has developed automated processes around CT fractional flow reserve (CT-FFR). A specific phrase placed in a radiology report will trigger a workflow after the attending physician signs the report. A software bot identifies the phrase, notifies the 3D laboratory and initiates the transfer of DICOM images to the appropriate vendor. Once the analysis is completed, the results are returned to PACS for interpretation.
Ghoshhajra noted that not every automation is technically operated byAI or machine learning. However, he views these tools as part of a broader strategy to remove repetitive tasks from the imaging workflow and allow physicians and technologists to work at the top of their license.
The approach becomes more complicated when introducing commercial AI algorithms. Mass General Brigham uses a governance committee closely connected with its broader informatics infrastructure to evaluate new technologies. The process considers not only whether an algorithm could improve cardiovascular imaging, but also evaluates the IT resources required to implement and maintain it.
Ghoshhajra said AI performance also must be monitored over time because algorithms can experience performance drift due to changes in clinical environments.
He anticipates that cardiovascular imaging training programs will increasingly incorporate informatics and data science so clinicians can understand how AI systems perform in real-world practice.
“We're all worried about losing jobs, but I think there's going to be whole new careers that are created for this," Ghoshhajra said,
For Mass General Brigham, the immediate value of AI may therefore be less about replacing physicians and more about redesigning the imaging chain so specialists spend more of their time on the decisions that require clinical expertise.
Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: [email protected]