New AI System LILAC Accurately Predicts Outcomes by Analyzing Image Changes Over Time
New York, NY (February 27, 2025) – Researchers at Weill Cornell Medicine, in collaboration with Cornell’s Ithaca campus and Cornell Tech, have developed a novel AI-based system named LILAC (Learning-based Inference of Longitudinal imAge Changes). This innovative system analyzes sequences of images taken over time, accurately detecting changes and predicting outcomes in a variety of contexts.
LILAC, which is based on machine learning, demonstrated its capabilities across diverse longitudinal image series, including developing IVF embryos, healing tissue after wounds, and aging brains.
“This new tool will allow us to detect and quantify clinically relevant changes over time in ways that weren’t possible before, and its flexibility means that it can be applied off-the-shelf to virtually any longitudinal imaging dataset,”
said study senior author Dr. Mert Sabuncu, vice chair of research and a professor of electrical engineering in radiology at Weill Cornell Medicine and professor in the School of Electrical and Computer Engineering at Cornell University’s Ithaca campus and Cornell Tech.
Studies using LILAC have shown its ability to identify even subtle differences between images taken at different times and predict related outcome measures, such as cognitive scores from brain scans. According to Dr. Mert Sabuncu, the new tool can detect and quantify clinically relevant changes over time, offering flexibility for use with various longitudinal imaging datasets.
Traditional methods for analyzing longitudinal image datasets often require extensive customization and pre-processing. LILAC, however, is designed to be much more flexible, automatically performing corrections and identifying relevant changes.
“This enables LILAC to be useful not just across different imaging contexts but also in situations where you aren’t sure what kind of change to expect,”
said Dr. Heejong Kim, LILAC’s principal designer and the study’s first author and an instructor of artificial intelligence in radiology at Weill Cornell Medicine.
In one demonstration, the researchers trained LILAC on hundreds of sequences of microscope images showing in-vitro-fertilized embryos as they developed. LILAC accurately determined which image was taken earlier in randomized pairs, achieving about 99% accuracy. LILAC also successfully ordered pairs of images of healing tissue and detected group-level differences in healing rates. Similarly, the system predicted time intervals between MRI images of older adults’ brains, as well as individual cognitive scores from MRIs of patients with mild cognitive impairment.
The researchers showed that LILAC can be adapted to highlight the image features most relevant for detecting changes in individuals or differences between groups. Dr. Sabuncu anticipates the tool will be invaluable in cases where there is limited knowledge about the process being studied and significant variability across individuals. The researchers plan to demonstrate LILAC in a real-world setting to predict treatment responses from MRI scans of prostate cancer patients.
The LILAC source code is freely available at https://github.com/heejong-kim/LILAC.
Funding for this project was provided in part by grants from the National Cancer Institute and the National Institute on Aging, both part of the National Institutes of Health, through grant numbers K25CA283145, R01AG053949, R01AG064027 and R01AG070988. For aging brain experiments, data were provided by OASIS-3: Longitudinal Multimodal Neuroimaging: Principal Investigators: T. Benzinger, D. Marcus, and J. Morris; NIH P30 AG066444, P50 AG00561, P30 NS09857781, P01 AG026276, P01 AG003991, R01 AG043434, UL1 TR000448, and R01 EB009352.
