Artificial Intelligence

DeepSplit: Segmentation of Microscopy Images Using Multi-Task Convolutional Networks

A two-branch U-Net to tackle the problem of undersegmentation (cell merging).

KCML: a machine-learning framework for inference of multi-scale gene functions from genetic perturbation screens

KCML is an intelligent system that allow inference of systematic Inference of multi-scale gene funcitons from genetic screens. It annotate genes with GO terms based on phenotypic similarity. As these annotations are data driven they provide more tissue type-specific gene funcitons.

Discovery of Rare Phenotypes in Cellular Images Using Weakly Supervised Deep Learning

Development of a deep learning algorithm that can automatically discover abnormal phenotypes from an image of heterogenous cells without performing cell segmentation. Discoverd phenotypes can then be highlighted by the algorithm for validation. This have an important application in early detection of cancer where few cancer cells progress into malignancy.