Abstract Body: Introduction
Data sharing and analyses of neuroimaging data can be time-consuming and often represent a rate-limiting factor in clinical trial research. This workflow, which involves downloading imaging data from a clinical trial’s electronic data capture (EDC) system, performing biomarker analyses, and then re-entering data into the EDC, can be cumbersome when performed manually. Integrating the EDC’s application programming interface (API) with cloud-computing processes can significantly reduce the time required for these tasks. The objectives of this study aimed to evaluate whether an automated workflow using cloud computing can reduce the processing time while maintaining accuracy for biomarker evaluation compared to manual methods.
Methods
We compared the time required for manual imaging data downloads, biomarker analyses, and data entry into a clinical trial’s EDC with that required for the same processes using an automated method. Additionally, we assessed the accuracy of automated biomarker analyses relative to semi-automated analyses performed manually by an expert reader, specifically focusing on the volumetric quantification of intracerebral hemorrhage (ICH).
Results
The manual process, involving downloading, de-identification, and semi-automated volumetric quantification of ICH, took an average of 12 hours and 57 minutes per CT uploaded to the EDC. Cloud computing completed the same tasks in an average of 8 minutes and 13 seconds. The cloud-based biomarker analysis demonstrated high accuracy, with an average ICH volume difference of -1.29mL (n=214; p=0.0001) compared to the semi-automated method.
Conclusion
In conclusion, the integration of the EDC’s API with cloud-computing processes for imaging ingestion and analysis reduces processing time of neuroimaging data in clinical trials. Additionally, the automated workflow maintained a high level of accuracy in the volumetric quantification of ICH. Although it may take a human reader 30 minutes to process neuroimaging data, limitations can delay these tasks. Multiple scans could be uploaded at once, delaying the time it takes a reader to analyze the images. Imaging data could be uploaded after workhours or on the weekend, postponing the analyses until the next working day. Yet, the automated cloud-based workflow drastically reduces processing time and ensures high accuracy of ICH volumetric calculations, offering a more efficient alternative to traditional manual methods.