How It Works

This section provides insights into how the Platform APIs works.

Overview

The Platform APIs offer a programmatic interface for partners to upgrade medical imaging systems with Qure’s AI diagnostic tools. Here is a high level overview of the integration:

The workflow for using the Platform APIs is straightforward, focusing on three main steps:

• Uploading the DICOM image

• Processing and analysis of the uploaded DICOM image

• Fetching the results

Uploading the DICOM Image

When you upload a DICOM image from your medical system, such as modality or PACS, the Client API Module makes an upload request to the Platform API, targeting either the /studies or /upload endpoint. The Platform API then takes over, processing the uploaded DICOMS. It categorizes them and performs initial checks to validate the format and metadata. Successful validation leads to the files being stored in object storage backed by AWS S3, and their metadata is recorded in Qure’s database. Upon a successful upload and metadata entry, the API confirms the action with a success message, which includes details of the metadata.

Processing and Analysis of the Uploaded DICOM Image

After a successful upload, the processing of the DICOM images proceeds as follows:

1. An image queue is created in which the uploaded images are placed. This queue organizes the images for further processing.

2. An image from the queue is then selected for analysis. The system determines the specific use case for the image based on the image information provided. This use case identification is crucial as it dictates the processing workflow.

3. The system determines the specific use case for the image based on the source information provided with the API request. This use case identification is crucial as it dictates the processing workflow.

4. With the workflow determined, the system generates tasks tailored to the requirements of the analysis. These tasks are the discrete units of work that need to be completed for the image.

5. These tasks are sent to a job manager. The job manager is responsible for distributing the tasks among available worker nodes.

6. Worker nodes are assigned these tasks. Each node is a computational unit that carries out the processing. It executes its assigned tasks, which could involve preprocessing the images, running them through diagnostic models, and generating outputs.

7. Upon completion of the tasks, the outputs, which includes the diagnostic results and any additional assets, are compiled.

8. These outputs are then stored: the assets are sent to object storage, and the results are recorded in the database.

Fetching the Results

The results are fetched for the uploaded DICOMs as follows:

1. The process begins on the client side, where the Client API Module makes a GET request to Qure's servers to retrieve the results for the uploaded image. This request is made to the /results API, with the the SOPInstanceUID of the DICOM appended to the URL.

2. On the Qure servers, the Results API receives the GET request and proceeds to fetch the diagnostic results and associated asset URLs from the database and object storage.

3. The results, along with any additional information such as URLs to the processed images or other assets, are compiled into a JSON response and sent to the client.

4. The Client API Module receives the JSON response and extracts the relevant data. It fetches the assets from their respective URLs, which could point to structured reports, images, or PDF documents.

5. Finally, the modality or PACS system on the client side receives these reports and assets, completing the cycle of requesting and receiving processed diagnostic information for the originally uploaded DICOM images.