aiqbench analysis
This study investigates whether AI-driven super-resolution techniques applied to FLAIR MRI scans inadvertently erase small white-matter lesions or generate false positives, a critical concern for clinical neurodegeneration diagnostics. The research finds that while all reconstruction methods improve detection over raw thick slices, the dominant risk is the erasure of genuine small lesions rather than hallucination, with specific models like ECLARE performing best at preserving lesion signal. These findings provide essential safety insights for deploying generative AI in medical imaging workflows where diagnostic precision cannot be compromised.
Does FLAIR super-resolution erase or hallucinate small white-matter lesions?
arXiv:2608.06311v1 Announce Type: cross Abstract: White matter hyperintensities (WMH), bright regions on Fluid-attenuated Inversion Recovery (FLAIR) scans are associated with cerebrovascular pathology and neurodegeneration. FLAIR is usually acquired with thick slices in clinical settings, giving it poor through-plane resolution. Super-resolution (SR) is a widely used method for recovering an isotropic volume from an anisotropic scan. Yet whether applying it prior to WMH segmentation preserves lesion content remains unknown: a model may erase small real lesions or hallucinate absent ones. We used 1-mm isotropic high-resolution (HR) FLAIR scans from 29 individuals in the ADNI cohort, each manually segmented for WMH by an expert. Then, we degraded each to simulated 3 and 5 mm through-plane acquisitions. Multi-contrast implicit neural representation (INR), a single-contrast self-supervised model (ECLARE), and cubic interpolation were used to upsample them onto the HR grid. WMH segmentation from a simulated thick slice and the original HR FLAIR set the floor and ceiling, respectively, for the per-lesion analysis. Of four WMH segmentation methods (WMH-SynthSeg, segcsvd, MARS-WMH, TrUE-Net), we ran the analysis under the most sensitive one to small lesions on HR (MARS-WMH) with the evaluation metrics of detection sensitivity, erasure rate (HR-detected lesions lost after reconstruction), and hallucination rate (predicted components absent from both the manual and HR segmentation). The dominant effect of SR was erasure of small real lesions, not hallucination, and it increased with slice thickness, though every reconstruction still improved lesion detection over the raw thick slice. ECLARE recovered small lesion signal best at both thicknesses, while the INR was no better than cubic interpolation.
Abstract:White matter hyperintensities (WMH), bright regions on Fluid-attenuated Inversion Recovery (FLAIR) scans are associated with cerebrovascular pathology and neurodegeneration. FLAIR is usually acquired with thick slices in clinical settings, giving it poor through-plane resolution. Super-resolution (SR) is a widely used method for recovering an isotropic volume from an anisotropic scan. Yet whether applying it prior to WMH segmentation preserves lesion content remains unknown: a model may erase small real lesions or hallucinate absent ones. We used 1-mm isotropic high-resolution (HR) FLAIR scans from 29 individuals in the ADNI cohort, each manually segmented for WMH by an expert. Then, we degraded each to simulated 3 and 5 mm through-plane acquisitions. Multi-contrast implicit neural representation (INR), a single-contrast self-supervised model (ECLARE), and cubic interpolation were used to upsample them onto the HR grid. WMH segmentation from a simulated thick slice and the original HR FLAIR set the floor and ceiling, respectively, for the per-lesion analysis. Of four WMH segmentation methods (WMH-SynthSeg, segcsvd, MARS-WMH, TrUE-Net), we ran the analysis under the most sensitive one to small lesions on HR (MARS-WMH) with the evaluation metrics of detection sensitivity, erasure rate (HR-detected lesions lost after reconstruction), and hallucination rate (predicted components absent from both the manual and HR segmentation). The dominant effect of SR was erasure of small real lesions, not hallucination, and it increased with slice thickness, though every reconstruction still improved lesion detection over the raw thick slice. ECLARE recovered small lesion signal best at both thicknesses, while the INR was no better than cubic interpolation.
10 pages, 2 figures, 3 tables. Accepted at the 11th International Workshop on Simulation and Synthesis in Medical Imaging (SASHIMI 2026), held in conjunction with MICCAI 2026. This is the version submitted for review; the final authenticated version will appear in the Springer LNCS proceedings
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2608.06311 [cs.CV]
(or arXiv:2608.06311v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2608.06311 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 322 words
Related coverage
Does preservation make sense before we know how to revive?
LessWrong · Jun 15, 2026
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
Presentation: Engineering AI for Creativity and Curiosity on Mobile
InfoQ · Jul 21, 2026
Best tablets for note-taking 2026: Expert tested and reviewed
ZDNet · Jul 21, 2026