Overview:
Our laboratory has recently developed a learning algorithm to assign circadian clock times to RNA-Seq samples sourced from the GTEx database, which includes samples from 48 non-diseased tissues taken from 900 individuals (Talamanca et al., Science, 2023). Within this extensive database, we have identified a set of liver and other tissue histology images that we would like to further investigate, with the goal of drawing correlations between histological features (i.e. cell size, ploidy), circadian times, as well as with other metadata, such as the patient’s BMI and medical history.
Aims:
In this project, the student will first work on utilizing and tailoring standard image segmentation techniques to automate feature extraction from the histology images. Following this first step, the student will build deep convolutional networks or vision transformers to identify pertinent features. Once these features are found, we will aim at establishing correlations between the image-derived data, the circadian clock times, along with other relevant patient information. By combining these data, we hope to discern patterns or correlations that might shed light on the influence of circadian rhythms on liver functionality. Moreover, we are keen on understanding if there are any evident links between extracted features and specific health conditions, as suggested by the patient’s BMI and medical history.
Prerequisites:
- Strong background in deep learning and image processing.
- Basic understanding of RNA-seq analysis and circadian biology is advantageous but not mandatory.
Contact: