Business Impacts
Saved ~70% of product implementation time
Developed stable and robust MLOps frameworks
Enabled 4x faster operations
Executed parallel workloads
Customer Key Facts
- Country : Central European country
- Size : 1,00,000+
- Industry : Biotech, Pharmacutical
Developing a Radiology Medical Image Processing Solution with Google’s Medical Imaging Suite
The client is one of the world’s largest biotech company and a global supplier of transformative innovative solutions that provides in-vitro diagnostics. The client was facing several challenges due to a lack of centralized data management and limited model portability and sharing options. The standalone ML workloads had no centralized control for artifact management or end-to-end pipelines for data ingestion and model saving. The client wanted to develop a robust solution for ingesting, curating, storing, and running MLOps by leveraging Google’s medical imaging suite capabilities.
Challenges
- Development of an ingestion pipeline and processing of DICOM and non-DICOM image files with its validation for data captured from various external sources
- Creating a robust MLOps foundation that can be used for future model migration
- Unavailability of customization in the existing platform for handling non-DICOM files
Technologies Used
Healthcare API
Docker
Vertex AI
Google Cloud
Cloud Storage
CloudRun
Dataflow
Developed an end-to-end accounts payable application
Solution
Quantiphi team created a customized solution using Google's MIS for processing DICOM and non-DICOM image data files that can be used to process radiology images.
Implementation and flow:
Use Case 1 : Radiology image of lung segmentation:
Quantiphi processed DICOM and non-DICOM image data files of lung segmentation to perform annotation tasks leveraging MIS platform
Use Case 2: Radiology image of geographic atrophy (GA)
Quantiphi utilized eye’s GA progression radiology image data for model deployment and served predictions using docker
Results
- Developed customized, smart medical image-processing solution
- Out-of-box solution platform to manage DICOM and non-DICOM image data files
- Ability to process various types of radiology-image data files received from multiple sources leveraging Google’s MIS platform
