The process of manual microscopy image analysis in digital pathology is still time-consuming and expensive. For our client, leading Danish pharma company, we utilized deep learning models, and successfully addressed this challenge by enabling faster and more cost-effective analysis of large-scale medical image datasets, enhancing drug testing, and development efficiency.

The Challenge

The manual analysis of microscopic images in digital pathology has traditionally been a slow and costly task. Our client faced this hurdle, with their biotechnology experts spending extensive time identifying treatment effects and tumor measurements. The existing deep learning models were inadequate for handling the large image sizes and complexity involved, leading to slower drug development cycles.

The Solution

To overcome this challenges, we built a custom digital pathology pipeline. The underlying platform was the Databricks Lakehouse, built on top of a public cloud to ensure scalability and empower researchers to deploy their own models. This approach was harnessed to process extensive wet lab data, accommodating image sets as large as 10 GB each with high-speed efficiency. The system features comprehensive support for a variety of imaging formats, equipping digital pathology teams with the tools to train and implement new models as needed. With this solution in place, the Digital Pathology Pipeline can be adapted for a range of use cases, streamlining the path from research to clinical application.

The Benefits

This solution allows for quick processing of large image files, which speeds up the research and development of new drugs up to ten times. The platform's design also makes it easy to scale up, handling more data as needed without losing performance. For researchers, the flexibility to use and create their own models means they can address a wide variety of cases effectively. This leads to better use of time and resources, making the analysis of pathology images both faster and less expensive.