Eliminating Bias in Leukemia Diagnosis
Carsten Marr, Director of the Institute of AI for Health, leads the LeukoBIAS project to make AI diagnostic tools for leukemia more fair and accurate. Foundation models—advanced AI systems trained on vast datasets—have already transformed medicine by enabling the detection of patterns and anomalies across diverse medical data. However, these tools can also perpetuate biases on patient age, sex, or ethnicity, potentially leading to unequal outcomes.
Using data from over 6,000 leukemia patients, Marr’s team is developing a framework to identify and correct these biases, ensuring that AI systems provide equitable and reliable diagnoses.
“We are at a critical point where AI is becoming essential in medicine,” says Marr. “By making sure that AI tools are fair, especially for serious conditions like leukemia, we can ensure regulatory compliance and better outcomes for patients.”
What are Foundation Models?
Foundation models are artificial intelligence systems trained on vast datasets in an unsupervised fashion. Their ability to generalize across different applications has transformed industries, including healthcare, where they enable accurate image processing and diagnostic support without requiring task-specific training. Foundation models excel at identifying patterns in medical data, such as detecting abnormalities in blood samples or tissue scans, making them highly valuable for diagnostics.
Transforming Disease Diagnosis with AI-Powered Cilia Analysis
Janna Nawroth’s Ai4Cilia project harnesses AI to analyze the rhythmic motion of cilia—microscopic, hair-like structures vital to respiratory and reproductive health. Defects in ciliary movement can signal disease like chronic obstructive pulmonary disease (COPD) or infertility, but these abnormalities are difficult to detect and often overlooked in clinical practice.
A team around Nawroth has developed a microfluidic platform to standardize the collection of data from ciliated cells, which can be harvested in the clinical via minimally invasive tissue brushing. By using AI to analyze complex ciliary patterns, the platform promises to improve diagnostics and facilitate drug discovery.
“AI-powered ciliary beat analysis has the potential to transform how we diagnose and treat complex conditions,” says Nawroth. “Integrating this tool into healthcare can provide more accurate and timely solutions for patients.”
