A Pipeline for ADNI Resting-State Functional MRI Processing and Quality Control

2026-02-03

AI Summary

Resting-state fMRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) are underutilized due to heterogeneous acquisition protocols, missing scans, and misalignment between clinical and imaging timepoints, which collectively limit statistical power. This study addresses these challenges by establishing an end-to-end processing pipeline that, for the first time, integrates multi-phase, multi-site, and multi-scanner rs-fMRI data from ADNI-GO/2/3. Leveraging open-source tools—including Clinica, fMRIPrep, and MRIQC—augmented with custom scripts, the pipeline achieves temporal alignment of clinical and imaging assessments, standardized preprocessing, and automated quality control. The resulting dataset adheres to BIDS-derivatives standards, delivering high-quality time-series data alongside individualized quality reports. This substantially enhances data usability and reproducibility, providing a robust foundation for longitudinal investigations of functional biomarkers and multimodal studies in Alzheimer’s disease.

Technology Category

Computer Vision: Multi-modal VisionMachine Learning: Time-Series/Data StreamsIntelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasetsSecurity and Privacy: Data transparency and provenanceWeb Mining and Content Analysis: Web data integration and cleaning

Abstract

The Alzheimer's Disease Neuroimaging Initiative (ADNI) provides a comprehensive multimodal neuroimaging resource for studying aging and Alzheimer's disease (AD). Since its second wave, ADNI has increasingly collected resting-state functional MRI (rs-fMRI), a valuable resource for discovering brain connectivity changes predictive of cognitive decline and AD. A major barrier to its use is the considerable variability in acquisition protocols and data quality, compounded by missing imaging sessions and inconsistencies in how functional scans temporally align with clinical assessments. As a result, many studies only utilize a small subset of the total rs-fMRI data, limiting statistical power, reproducibility, and the ability to study longitudinal functional brain changes at scale. Here, we describe a pipeline for ADNI rs-fMRI data that encompasses the download of necessary imaging and clinical data, temporally aligning the clinical and imaging data, preprocessing, and quality control. We integrate data curation and preprocessing across all ADNI sites and scanner types using a combination of open-source software (Clinica, fMRIPrep, and MRIQC) and bespoke tools. Quality metrics and reports are generated for each subject and session to facilitate rigorous data screening. All scripts and configuration files are available to enable reproducibility. The pipeline, which currently supports ADNI-GO, ADNI-2, and ADNI-3 data releases, outputs high-quality rs-fMRI time series data adhering to the BIDS-derivatives specification. This protocol provides a transparent and scalable framework for curating and utilizing ADNI fMRI data, empowering large-scale functional biomarker discovery and integrative multimodal analyses in Alzheimer's disease research.

Problem

• resting-state fMRI
• Alzheimer's disease
• data quality
• temporal alignment
• neuroimaging variability

Innovation

• rs-fMRI
• quality control
• BIDS
• multisite harmonization
• Alzheimer's disease