
Projects
Deep Cell
Deep learning-enabled software for single-cell analysis
PI: David Van Valen (Division of Biology and Biological Engineering)
SASE: Raffey Iqbal, 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.
The Van Valen Lab’s cloud-native software ecosystem DeepCell deploys deep learning models to perform single-cell analysis tasks ranging from segmenting cells in pathology images to cell tracking in timelapses of living cells. This ecosystem has multiple components, including CellSAM, a foundational model for cell segmentation based on the Segment Anything Model (SAM) from Meta AI, a biological image annotation tool (DeepCell-Label), and a Kubernetes-backed deployment for batch processing of image data (DeepCell-Kiosk). The lab’s deep learning models render difficult analyses routine and empower new, previously impossible experiments across the life sciences. Large scale, high quality labeled datasets are needed to train models that perform and generalize across tissues and cell cultures.
The Schmidt Academy is working with the Van Valen Lab to develop modern serverless architectures for model deployment. This will include migrating existing cellular image analysis workflows in DeepCell-Kiosk into the serverless architecture and developing sustainable deployment options for the foundation models. Another focus is updating DeepCell’s data management system to better accommodate the ever increasing size of spatial-omics datasets. Due to the huge volume of collected image data needed to train the deep learning models, this work is expected to significantly improve the performance and usability of the DeepCell suite.
The Schmidt Academy has in the past worked with the Van Valen Lab to develop a data labeling web application DeepCell Label. Further work is also planned to enhance DeepCell-Label by improving visualization capabilities, integrating with foundation models such as CellSAM, and exploring new capabilities such as handling 3D image data and annotating neuroimaging.