EXoPERT EMERALD Clinical Study

Overview

The purpose of this study is to establish a multi-center clinical repository of blood samples to support the development and evaluation of an artificial intelligence-based in vitro diagnostic software. The software analyzes surface-enhanced Raman spectroscopy (SERS) profiles of extracellular vesicles (EVs) extracted from human plasma for the early detection of multiple cancers, including lung, ovarian, breast, pancreatic, and colorectal cancers.

The EMERALD study is a multi-center study (Protocol No. EXPT-MC-F-2024) designed to establish a clinical plasma repository for the training and clinical evaluation of the ExoPred deep learning software The study protocol operates in three sequential phases to ensure data integrity and prevent algorithmic overfitting: Phase 1 (Algorithm Training Stage): A minimum of 500 samples are analyzed to develop the algorithm and establish the initial clinical cut-off values, followed by a database lock. Phase 2 (Algorithm Lock-Readiness Assessment): A minimum of 50 newly acquired unknown samples are evaluated to assess model robustness before the final algorithm is formally locked in the Technical File. Phase 3 (Blinded Independent Stage): A minimum of 200 independent samples, strictly excluded from prior phases, are tested under strict blinding to evaluate finalized clinical sensitivity, specificity, and accuracy. The locked software analyzes surface-enhanced Raman spectroscopy (SERS) profiles of extracellular vesicles (EVs) to detect five target malignancies: lung, breast, colorectal, pancreatic, and ovarian cancers. Based on multi-class classification probability scoring, the predicted tissue of origin (TOO) can be presented for up to a maximum of three cancer types. All data handling and automated quality checks follow standard operating procedures (SOPs) within a 21 CFR Part 11 compliant EDC system.

Conditions