Cell Solution Implementation & Automation

Company Name

Location

United Kingdom

Industry

Client

bit.bio, a synthetic biology company, is dedicated to advancing medicine by industrializing the production of human cells, making curative systems more accessible. bit.bio converts induced pluripotent stem cells into human cell types in a single, scalable step, ensuring high purity and consistency.

Challenge

The unique and complex shapes and formats of the structures of interest, combined with the number of conditions to process, made manual analysis extremely time-consuming. The DataArt team addressed this by automating the analysis process. They developed a custom deep learning model in order to overcome the challenges of shape detection, enabling robust and high-throughput morphological analyses.

Solution

DataArt leveraged Vertex AI Services to train a custom computer vision model, by using various models from Model Garden. The best results on the labeled test data were achieved with a custom fine-tuned Mask R-CNN model, boasting 97% segmentation accuracy. They deployed this model for inference and built an automatic retraining pipeline on Jenkins, ensuring it could adapt to newly labeled data.

Highlights

Automated Training

on the new data

Custom Model

With 97% accuracy based on Vertex AI Mask R-CNN

Increased

Efficiency, 40% speed up comparing to manual process

Results

The instance segmentation model provided to the bit.bio team standardizes and automates this part of the workflow, saving researchers valuable time and leading to increased accuracy and efficiency. In addition, the scalable cloud infrastructure ensures adaptability to increasing data volumes, supporting evolving research and an expanding portfolio.

Please provide your contact details, and we will get back to you promptly.

First name*

Last name*

Email*

Job title

Company*

Phone

What would you like to discuss?*

How did you learn about DataArt?

Choose from the list

What is your role?

Choose from the list

I agree to receive DataArt's monthly marketing newsletter.

I have read and agree to the Privacy Policy*