Work

PLUME-OCT: A quality control tool to enhance statistical power in the analysis of biomarkers from 3D biomedical datasets of OCT images

Year: 2025

Type: article

Abstract: High-dimensional imaging data are central to biomedical research, enabling extraction of quantitative biomarkers for disease modeling and association studies. While quality control (QC) often targets individual 2D slices, many applications rely on the integrity of full 3D volumes. Optical coherence tomography (OCT), used here as a representative modality, illustrates the need for automated QC methods assessing volumetric coherence to support reliable biomarker extraction. We developed Phase-Level Unified Metric Evaluation (PLUME-OCT), a preprocessing tool that quantifies inter-slice misalignments in 3D biomedical image volumes using a Discrete Fourier Transform-based approach. The method was validated against 73,920 pairwise comparisons annotated by four human graders across 40 OCT volumes from the OphthalmoLaus cohort. As a large-scale case study, PLUME-OCT was applied for QC of 88,247 OCT scans from the UK Biobank prior to 3D biomarker extraction and genome-wide association analysis. PLUME-OCT effectively ranked OCT scans by quality, detecting misalignments between OCT B-scans in agreement with human annotators. In large datasets, its use enhanced the detection of statistically significant signals in 3D retinal biomarkers analysis, increasing significant genetic associations by 21.1 % (p = 1.6 × 10 −7 ) and heritability estimates by 19.1 % (p = 5.0 × 10 −12 ) across 30 genome-wide association studies. PLUME-OCT is an open-source, lightweight, and scalable QA tool that enhances statistical power in 3D biomarkers analysis from biomedical imaging datasets. Although demonstrated on retinal OCT, the method applies broadly to other volumetric imaging modalities and can be seamlessly integrated into biomedical informatics workflows for high-quality, large-scale data preprocessing. • Introduces PLUME-OCT for volumetric quality control of 3D OCT scans. • Quantifies inter-slice misalignment via phase-only correlation metrics. • Validated on 73,920 expert annotations across 40 OCT volumes. • Quality control of 88,247 scans increases GWAS gene associations by 21 %. • Open-source tool scalable to massive biomedical imaging datasets.

Source: Computers in Biology and Medicine

DOI: 10.1016/j.compbiomed.2025.111420