Enhancing clinical trial imaging workflow with AI integration.

Abstract

e13662 Background: AI is revolutionizing clinical trials by assisting several stages of the trials, including patient recruitment, creation of clinical study reports and quality control that can enhance efficiency, accuracy, and decision-making processes. Over 50% of all clinical trials and 70% of oncology trials utilize imaging as an endpoint. The clinical trial imaging workflow involves numerous manual tasks, which, when scaled to thousands of imaging uploads per day, highlight the necessity for efficient processing. Voiant Hub has implemented, and utilizes, AI to enhance the image processing workflow for determining oncology clinical imaging endpoints. Methods: Key activities in imaging clinical trials include Image Quality Control (QC) and Image Preparation, supported by trained operators before independent assessment by a reader. Automating these steps with AI drastically reduces the time from image upload to independent review from days to minutes. AI-assisted QC addresses challenges like human errors, delays, regulatory variances, and image quality. Uploaded images are tested against multiple deep learning-based AI models to identify modality, anatomy, sequence, and contrast, generating data for image quality assessment. The assessment report is automatically uploaded to the Voiant Hub portal within minutes. Results: Preliminary performance results show high accuracy rates, with AI models achieving up to 99.5% accuracy on Image QC. For instance, in glioma brain and solid tumor body trials with CT & MRI imaging, the AI model achieves 99.5% (n = 205) and 93% (n = 245) accuracy for contrast & sequence identification (T1, T2, Flair), and 97.3% (n = 302) and 95% (n = 757) accuracy for anatomy coverage. AI-powered lesion tracking further reduces the time spent by the reader in measuring lesions and assessing treatment efficacy and disease progression. Conclusions: The integration of AI in clinical trial imaging workflows significantly enhances efficiency, accuracy, and speed, contributing to more successful results in clinical trials.

Ähnliche Arbeiten