Christoffer Nellåker

Dr. Nellåker’s group is translating the latest developments in computer vision and computational biology to answer biomedical questions and deepen biological understanding. Broadly the group works with deep learning approaches for image analysis, histology, spatial transcriptomics, and morphometry phenotyping. We apply these through multidisciplinary collaborations with clinical genetics, clinical pathology, clinical microbiology, physics, engineering, genetics and women's and reproductive health groups around the world.

The group has published methods for phenotyping placenta biology from histological imaging - showing that complex placenta biology and cell type distributions can be elucidated using computer vision approaches. We have also shown the utility of deep segmentation approaches to perform automated phenotyping of adipose histology samples and that these can be used to find robust genetic associations. Current work in the group is building high throughput tools for processing of whole slides images in research studies and clinical pathology settings.

We have shown the potential for deep image analysis to be used for extracting diagnostically meaningful information from facial images, but also the challenges with taking such technologies forward into clinical settings. Our research has shown how reproducibility and biases in deep learning approaches can impede the potential benefits of such technologies, and developed some approaches to mitigate such challenges (1,2).

Similar methodological approaches have utility in diverse biomedical fields and the group has a collaboration around microbial imaging and phenotyping to perform rapid antimicrobial resistance testing.