Screening: Better prediction of liver cancer risk
A team of UC Davis Health clinicians and data scientists have developed a machine-learning model to better predict which patients are at greater risk of developing the common liver cancer hepatocellular carcinoma (HCC).
Their findings — published in January in the journal Gastro Hep Advances — describe how predictive-learning can aid in providing early HCC risk assessments for patients diagnosed with metabolic dysfunction-associated steatotic liver disease or MASLD, formerly called nonalcoholic fatty liver disease or NAFLD. Around 25% of Americans have some form of MASLD, an accumulation of fat in the liver often linked to metabolic diseases such as type 2 diabetes. The pilot technology may be able to give providers critical information to screen patients more closely and offer more personalized care.
“MASLD can lead to HCC, but the disease is quite sneaky, and it’s often unclear which patients face that risk,” said study co-author Aniket Alurwar, M.S., a clinical informatics specialist at the UC Davis Center for Precision Medicine and Data Sciences. “It doesn’t make sense to biopsy every patient with MASLD, but if we can segment for risk, we can track those people more closely and perhaps catch HCC early.”
Researchers trained machine learning algorithms, which leveraged large datasets to make verifiable predictions, and tested them against de-identified data from 1,561 UC Davis Health and 686 UCSF patients. The study confirmed one of the most reliable markers for HCC risk is advanced liver fibrosis or scarring, characterized by high Fibrosis-4 Index (FIB-4) scores. However, researchers also found four additional risk factors associated with liver function: high cholesterol, hypertension, bilirubin and alkaline phosphatase (ALP). A combination of those factors in one model helped predict HCC risk.
The team found there are multiple pathways to HCC, with high FIB-4 the most obvious. In some cases, patients with low FIB-4 but high cholesterol, bilirubin and hypertension also developed HCC. Under current guidelines, these patients would not receive precautionary care.
“We got 92.12% accuracy when predicting which MASLD patients would develop HCC, which is very good for a pilot model,” Alurwar said. “Patients with low FIB-4 are typically considered low risk and do not get referred for further assessment. By showing which of these ‘low risk’ patients could develop HCC, we can get them referred for biopsies or imaging.”
Researchers plan to advance their accuracy by incorporating more precise data, such as clinical notes, using AI natural language processing to translate written text into data. The team will also test Bedrock, Amazon’s generative AI platform. Eventually, a similar model could be incorporated into electronic health records, or a separate platform, to flag greater HCC risk.
Acute care: Identifying stroke quickly
UC Davis Health has adopted a new technology platform, Viz.ai, to help quickly identify patients suspected of having a stroke. The hospital is the first in the Sacramento region to use the platform, which utilizes image-based AI to analyze CT scans and alert care teams of a potential stroke within minutes. The platform complies with federal patient privacy laws, allowing the care team to communicate securely.
In addition to coordinating care within the hospital, Viz.ai will allow remote and regional hospitals to securely share critical patient images and information with experts at UC Davis’ stroke center. Images will now be shared within minutes, allowing UC Davis clinicians to make more informed decisions before patient transfers.
“Our partnership with Viz.ai is a great example of how we’re empowering our clinicians with the latest tools,” said UC Davis Health’s chief AI advisor Dennis Chornenky. “This is part of a broader AI adoption roadmap that will help us augment, but not replace, the capabilities of our care teams and improve patient experience.”
As of press time, UC Davis Health had successfully launched Viz.ai usage as part of its telestroke program with the Adventist Health + Rideout hospital serving the Yuba-Sutter region – with one life saved already. After a patient presented to Rideout with severe stroke symptoms, Viz.ai detected a large vessel problem within minutes of imaging completion. Alerts to the local emergency physician and UC Davis stroke neurologist allowed for expedited review of the patient’s presentation and imaging, and a rapid determination to pursue thrombectomy. A decision to transfer the patient for treatment occurred within 10-15 minutes of imaging completion.
Time saved is brain saved, and this process saved the patient three to four times the amount of time spent in more traditional triage, Ng noted.
Expansions of Viz.ai usage are underway for our telestroke partnerships with Adventist Health Lodi Memorial Hospital serving San Joaquin County, Howard Memorial Hospital in Mendocino County, and NorthBay Health Hospitals in Solano County.
More examples
Improving early breast cancer detection
With support from the National Cancer Institute, a national research team co-led by Diana Miglioretti, Ph.D., is exploring how to use AI to make breast cancer screening and surveillance more accurate and equitable. The team is using AI to predict which women with no history of breast cancer are at high risk of advanced cancer, and also to determine whether AI detection scores and facility-level interventions can improve outcomes.
Creating de-identified data from clinical notes
AI analyzes clinical notes and longitudinal medical histories to create aggregate, de-identified data for an opioid overdose surveillance dashboard created for public health agencies in Sacramento and Yolo counties. The pilot using the SMART Cumulus digital platform is part of a partnership involving UC Davis Health and its Digital CoLab innovation hub, the CDC Foundation, and Boston Children’s Hospital.
Streamlining physician notes
UC Davis Health is piloting AI scribe technology to decrease clinician burnout by automating note captures with human oversight. AI drafts about 90% of note content with an aim to improve physician morale, information accuracy, patient satisfaction and access to care. The first pilot generated tremendous enthusiasm among participating physicians, Atreja said, and at press time was expanding with system-wide implementation envisioned in the future.
Improving real time decision-making support during surgery
Founded with the help of a prestigious $6.3 million NIH P41 grant, UC Davis’ National Center for Interventional Biophotonic Technologies is advancing two noninvasive optical imaging technologies — both developed here — that measure fluctuations in light from bodily tissues. By adding AI capabilities to analyze data, researchers intend to create new instruments for robust tissue analysis during surgery and for blood flow monitoring.
Developing privacy-preserving machine learning techniques
UC Davis researchers were awarded a four-year, $1.2 million National Institutes of Health grant to generate high-quality “synthetic data,” which can be generated from real-world sources such as images, videos, text or speech in a way that preserves statistical properties without risk of exposing sensitive information.
