THINQ

THINQ™ is a software medical device that automates the volumetric quantification and visualization of segmentable brain structures from 3D T1-weighted MRI scans, employing a Docker-based batch processing system. It provides clinicians with neuromorphometric reports comparing patient brain volumes to age and gender-matched reference percentile data, serving as a supplemental tool for neurological assessment.

THINQ™, developed by CorticoMetrics LLC, is a software-only medical device designed for the automated quantitative analysis of human brain structures from 3D T1-weighted MRI scans. Its primary purpose is to perform automatic labeling, visualization, and volumetric quantification of numerous segmentable brain structures. The system processes DICOM-formatted MRI data and generates a comprehensive neuromorphometry report in PDF format. This report includes absolute volume measurements for approximately 30 brain structures, these volumes expressed as a percentage of intracranial volume, and visual depictions of the segmented brain areas with color-coded overlays. A key feature is the comparison of these patient-specific volumetric measurements against an internal, age and gender-matched reference percentile database, derived from a normative dataset considering scanner manufacturer and magnetic field strength. Operating as a non-interactive module within a Docker container, THINQ™ is typically deployed in high-performance computing environments for batch processing. It requires specific Linux and Docker versions, along with significant CPU and RAM resources. The input MRI scans must adhere to strict DICOM tag requirements and acquisition parameters, ideally following ADNI MRI protocols, from Siemens, GE, and Philips scanners at 1.5T and 3T. In addition to the PDF report, THINQ™ produces DICOM encapsulated JPEG images showing brain segmentation overlays for visual verification by clinicians. Performance testing demonstrated high accuracy against expert manual labels, with Dice similarity coefficients for major subcortical structures typically between 0.75 and 0.90, and good reproducibility with absolute percent differences for major structures generally between 0.8% and 3% on test-retest data. However, the software is a supplemental tool, not diagnostic, and its accuracy can be affected by incorrect scan protocols, artifacts, or atypical brain anatomy; it has contraindications for specific patient populations and imaging types.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does THINQ's segmentation algorithm specifically address the challenges in accurately quantifying deep grey matter structures, an area where other automated methods have shown limitations?; What is the composition and diversity (e.g., ethnicity, clinical conditions beyond healthy controls) of the reference dataset used for generating normative percentile data, and how is potential bias mitigated?; Beyond the listed Dice coefficients and absolute percent differences, what is the inter-scanner and intra-scanner variability of THINQ's volumetric measurements for specific brain structures when using MRIs acquired under slightly varying, clinically realistic protocols?; How does THINQ's automated quality assurance module identify and handle suboptimal input MRI data (e.g., motion artifacts, poor contrast) before or during processing, and what is the failure rate due to input data quality?; Given THINQ does not segment lesions, how might its volumetric outputs be confounded in patients with unannotated white matter lesions or other subtle pathologies common in neurological disorders like multiple sclerosis or early Alzheimer's disease?