Health systems that embrace artificial intelligence tools can not only improve radiology operations and quality but patient follow-up as well, which can result in greater staff efficiencies, higher patient completion rates and better access and outcomes for patients.
A collaboration between East Alabama Medical Center, Inflo Health and the American College of Radiology Learning Network began using machine language models and advanced natural language processions to extract data from radiology reports to improve its follow-up with pulmonary patients while Stamford Health in Connecticut was able to extend additional radiological measures to all cardiovascular patients through automation.
Also of note this week, Lunit, a vendor of cancer diagnostics and therapeutics, announced that two recent studies evaluating its AI-powered mammography screening found the technology could also estimate the development of breast cancer up to six years before a positive diagnosis.
“If the scores of commercial AI algorithms developed for immediate cancer detection can also estimate future cancer risk, then more accurate and reliable short-term risk estimation could lead to tailored, personalized preventive measures, possibly resulting in earlier breast cancer detection and less-aggressive treatment,” European researchers said in a statement Wednesday.
EAMC improves patient follow-up
The Alabama health organization announced Thursday that through a partnership tracking radiology follow-up with AI — and by involving primary care physicians in acute care communications — transformed its recommendations follow-up rate by 74%.
EAMC partnered with Inflo Health, which leverages radiology-specific language models and advanced NLP, and the American College of Radiology to boost its patient engagement and clinician productivity.
The AI-powered software is executed on measure specifications outlined by the ACR’s ImPower program — which helps organizations build improvement leadership skills and methods to achieve better outcomes — helping EAMC radiologists identify additional imaging recommendations and actionable findings, as well as automating department workflows.
The goal of the collaboration with EAMC was to improve the consistent inclusion of post-scan recommendations for incidentally detected pulmonary nodules and also increase the percentage of exams that received timely follow-up, the organizations said in a statement.
EAMC also implemented the AI software’s appropriateness measures, automating the process of identifying incidental lung nodules that met the inclusion criteria.
The effort significantly streamlined EAMC’s processes, reduced manual effort and boosted staff efficiency, according to Melinda Johnson, the organization’s radiology director.
“This has also enabled us to expand care navigator roles to other clinical areas,” she said in a statement. “This partnership exemplifies how integrating advanced technology with strategic collaboration can set new standards in radiology practices and operational excellence.”
The result was a reduction in manual tasks from five hours per week to just 15 minutes, representing a 95% efficiency improvement, the collaborators said.
To improve patient completion and relay the recommended imaging follow-ups, EAMC addressed operational barriers, including inconsistent communication between acute care and primary care. As a bonus, that effort generated an estimated $9,000 per month in additional revenue.
“Leveraging technology to standardize and optimize clinical workflows requires the concerted efforts of organizations and their software vendors working in tandem so that the solution is built by understanding the problem” added Judy Burleson, ACR’s vice president of quality management programs.
“The quality improvement education and support provided by the ImPower program, coupled with EAMC’s commitment to improve patient outcomes, and Inflo Health’s willingness to adapt their product, made these advancements possible,” she said.
Stamford Health enhances access
Stamford Health, a nonprofit organization serving Fairfield County, Connecticut, announced earlier this month a new automated cardiovascular screening that enables more timely and personalized follow-up care for patients at risk.
Stamford Health’s Heart & Vascular Institute said in a statement that the AI-powered cardiovascular screening tool significantly improves the early detection and management of cardiovascular disease across its patient population.
The institute uses Bunkerhill Health’s advanced algorithm to identify the presence of coronary calcium by calculating the total coronary artery calcium or Agatston score, an indicator of future risk of coronary artery disease in a predefined patient population.
CAC screening would normally require a special order from a physician, but the automated algorithm now runs in the background of all the institute’s non-gated chest CT scans, such as those used in lung cancer screenings.
“We are focused on providing the most cutting-edge, sophisticated care to our patients,” said Dr. Ronald Lee, chair of Stamford Health’s department of radiology.
Patients will automatically receive a CAC score during any non-contrast chest CT scan and when elevated CAC is identified, the patient’s primary care provider or cardiologist is notified of their score and risk.
“This tool enhances our ability to detect early signs of cardiovascular disease and ensures that patients receive the follow-up care they need to prevent serious health outcomes,” added Dr. David Hsi, chief of cardiology and the institute’s codirector.
Testing AI for predictive mammography
Accuracy of mammography screening has long been a challenge with radiology protocols often calling for double scan readings. AI algorithms can mark areas of concern and provide breast-level and examination-level malignant neoplasm scores to aid radiologists in image readings.
Lunit said Wednesday that researchers at the Cancer Registry of Norway and Odense University Hospital in Denmark already using its INSIGHT MMG tools demonstrated the potential to also improve the predictive value of its national breast cancer screening programs, ultimately leading to earlier diagnosis and treatment for women.
The retrospective Norwegian study, Artificial Intelligence Algorithm for Subclinical Breast Cancer Detection, completed in August and published earlier this month in the JAMA Network, analyzed image data from a cohort of 116,495 women aged 50 to 69 years with no prior history of breast cancer.
Norway’s cancer registry, which has a contract with Lunit for research use of AI software, offers digital mammography screening every two years. The patients in the retrospective cohort study underwent at least three consecutive biennial screening examinations performed between September 13, 2004, and December 21, 2018, at nine of the country’s breast screening centers.
Researchers divided the cohort into three groups – women with screening-detected breast cancer on the third study screening round, women with interval cancer diagnosed after the third study screening round and women with no breast cancer diagnosed after three consecutive examinations and six years without cancer diagnosis – finding 1,265 screening-detected cancers and 342 interval cancers.
For those identified with breast cancer — defined as ductal carcinoma in situ or invasive breast carcinoma — the mean absolute AI scores were higher for breasts developing versus those not developing cancer four to six years before their eventual detection. AI scores were also higher and increased more rapidly over the three successive screening rounds for women with a diagnosis of screening-detected cancer versus an interval cancer.
“These findings suggest that commercial AI algorithms developed for breast cancer detection may identify women at high risk of a future breast cancer, offering a pathway for personalized screening approaches that can lead to earlier cancer diagnosis,” according to researchers.
Andrea Fox is senior editor of Healthcare IT News.
Email: [email protected]
Healthcare IT News is a HIMSS Media publication.
Source : Healthcare IT News