Projects
Deep Cell
Deep learning-enabled software for single-cell analysis
PI: David Van Valen (Division of Biology and Biological Engineering)
SASE: Tom Dougherty, Scholar
Single-cell imaging captures the temporal and spatial variations that underlie life, from the scale of single molecules to entire organisms. These images record cell morphology, signaling and gene expression in their native configurations. Despite the richness of single-cell images, extracting measurements from images is intensive in labor, time, and capital. Classical image segmentation methods have enabled many discoveries, but are limited in robustness and number of applications. Recent advances in computer vision and deep learning have provided the required accuracy and flexibility, but the unique data and hardware acceleration requirements can place deep learning beyond the reach of many life scientists.
The Van Valen lab has aimed to address this problem through DeepCell, a cloud-native software ecosystem for deploying deep learning models on large imaging datasets. The DeepCell platform achieves single-cell analysis tasks from segmenting cells in pathology images to cell tracking in timelapses of living cells. Through robust and accessible deep learning-enabled software, we render difficult analyses routine and empower new, previously impossible experiments across the life sciences field.
The Schmidt Academy is collaborating with the Van Valen group to enhance their platform DeepCell for deep learning-enabled single-cell image analysis. The DeepCell platform aims to combine model inference and data curation to become a full human-in-the-loop solution for developing deep learning methods that meets the needs of the life science community.
Figure 2: A human-in-the-loop to dataset construction and cloud computing facilitate a scalable solution to live-cell image analysis. (a) Combining crowd sourcing and a human-in-the-loop approach to dataset annotation enables the construction of an ImageNet for live-cell imaging. By annotating montages, crowd contributors both segment and track single cells in live-cell imaging data. This data leads to models that are used to process additional data; expert annotators use Caliban to correct model errors and identify cell division events. The resulting data is then used to train a final set of deep learning models to perform cell segmentation and tracking. (b) Integration of a cell tracking service into DeepCell 2.0. Datasets are uploaded to a cloud bucket; once there, a tracking consumer object facilitates interactions with deep learning models via Tensorflow serving for segmentation and tracking. The implementation within the Kubernetes engine includes an autoscaling model that monitors resource utilization and scales the compute resources accordingly.
