Cornell University: AI apps bring veterinary data to CVM community

Human medicine has long been able to use artificial intelligence to mine patient data to augment clinicians’ and researchers’ work. Now, thanks to an effort by the information technology team at the College of Veterinary Medicine (CVM), these tools will soon be available to faculty, staff and students for animal research.

“The college has a wealth of contemporary and historical clinical data that we have long wanted to use more effectively for research, education and practice,” said Dr. Lorin D. Warnick, Ph.D. ’94, the Austin O. Hooey Dean of Veterinary Medicine. “This recent work done by the information technology team is a major achievement towards this goal.”

Established as a goal of the CVM strategic plan, college faculty and administration pledged to improve the ability to analyze clinical, research and business data.

“Right now, in the veterinary industry, much of the data is still siloed,” said Scott Ross, assistant director of application development and integration. “We’re trying to address that issue long term.”

The mission has been spearheaded by Dr. Meg Thompson, associate dean for hospital operations and director of the Cornell University Hospital for Animals (CUHA). Very few veterinary institutions have developed comprehensive, searchable databases for daily use in the clinic, classroom and lab.

“We’re trying to answer some really interesting questions,” Ross said. “We’d like to move the needle in veterinary medicine toward more evidence-based decision making. Thanks to Dr. Thompson, the college is beginning to realize the power of the data held in many of its applications.”

Thompson built strategic partnerships with industry leaders, including ezyVet, while simultaneously allowed the technology team to concentrate on innovative solutions over operational concerns.

The data used for these applications includes everything from the 1.4 million clinical cases CUHA and Cornell University Veterinary Specialists have recorded, some going back to the early 1970s – to the 14.2 million diagnostic tests that the Animal Health Diagnostic Center has documented since 2000. It also includes 90 terabytes of pathology slides that were digitized during the pandemic to allow veterinary pathologists to view tissue samples remotely. The project has also incorporated administrative data from human resources and accounting platforms.

The first step with these multimillion datasets was to do “data positioning,” which means “massaging it to make it usable for people,” Ross said. Then the team, including software developers Steve Halasz and Daniel Sheehan, coded three apps to make them easily usable:

Case Search: Released in June 2020, this digital app allows faculty, staff and students to do a Google-like search of the millions of clinical cases recorded since the 1970s using a variety of keywords, including diseases, breeds and owner names – instantly pulling up results that show master problems, medications, labs, diagnosis and more.
Case Experience: Released in September 2021, this app provides a comprehensive dashboard of all clinical cases seen by any clinician or student at the hospital, allowing a per-person breakdown of species and breeds, diseases, medications and procedures, and case timelines. This is particularly useful for students in their clinical rotations and specialty trainees, who can then view and track the breadth of their clinical experiences in one view. It enables both faculty and students to identify potential gaps in competencies and track progress, and will aid in the college’s dedication to competency-based curriculum.
Cohort Builder: Slated for release in 2022, this app will identify groups of patients for research projects using heterogeneous datasets, accurately identifying relevant patients and associated data, improving the quality of studies and research. For example, a clinician scientist who wants to study how successful a surgical technique is in a certain group of dogs would normally have to spend hours manually searching old cases, record individual data fields in a spreadsheet, and then analyze the outcomes of each case. Cohort Builder does that automatically, pulling all relevant cases, so researchers can put together a large sample size for their analysis.
“For all clinical research that involves searching medical records, particularly retrospective cohort studies, this kind of technology is hugely helpful,” said Dr. Robert Goggs, associate professor of emergency and critical care. “We are always looking to maximize the number of patients we can include in our studies while also ensuring we get high quality, complete data. Using AI to mine the wealth of our collective medical records to help answer research questions is a huge time saver.”

In the future, Ross said, the team hopes to mine human resources and administrative data for other types of analyses and continue to build applications that harness the college’s wealth of information.

“We’re excited to put these tools in the hands of our faculty, staff and students to make their day-to-day lives easier,” Ross said.