Researchers at Weill Cornell Medicine, Cornell’s Ithaca campus, and Cornell Tech have developed LILAC, an AI-based system that enhances medical image analysis.
LILAC (Learning-based Inference of Longitudinal imAge Changes) is designed to analyse images over time using machine learning.
Building on its sensitivity and adaptability, the system can be used for a broad range of medical and scientific applications, offering a significant advancement in image analysis.
The researchers demonstrated the system on diverse longitudinal image series, including developing IVF embryos, healing tissue, and ageing brains.
In the study, LILAC identified minute differences between images and predicted outcomes such as cognitive scores from brain scans.
The study was led by Weill Cornell Medicine AI instructor in radiology Heejong Kim.
Study senior author Mert Sabuncu said: “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.
“We expect this tool to be useful especially in cases where we lack knowledge about the process being studied, and where there is a lot of variability across individuals.”
According to the researchers, traditional methods for analysing longitudinal image datasets often require extensive customisation and pre-processing.
LILAC performs these corrections automatically, enhancing flexibility and efficiency.
In a proof-of-concept demonstration, LILAC was trained on sequences of microscope images showing in-vitro-fertilised embryos.
The system accurately determined the chronological order of images with about 99% accuracy, underscoring its ability to detect time-related changes.
LILAC also succeeded in ordering images of healing tissue and detecting group-level differences in healing rates.
It predicted time intervals between MRI images of older adults’ brains and cognitive scores from MRIs of patients with mild cognitive impairment with less error than baseline methods.
The researchers said that LILAC’s adaptability in identifying image features is key for detecting changes in individuals or groups and offers new clinical and scientific insights.
Furthermore, the researchers are planning to demonstrate LILAC in real-world settings to predict treatment responses from MRI scans of prostate cancer patients.
LILAC principal designer Kim said: “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.”
