Back

Increased subpial cortical lesion detection at 3 tesla using Inversion Recovery Susceptibility Weighted Imaging with Enhanced T2 Weighting (IR-SWIET)

Sizer, E.; Onyemeh, K.; Kohli, A.; Levit, E.; Roy-Hewitson, C.; Brown, Z.; Low, J.; Feb, K.; Zhang, J.; Ulano, A.; La Rosa, F.; Nair, G.; Reich, D. S.; Shinohara, R. T.; Morrow, S. A.; Solomon, A. J.; Beck, E. S.

2026-08-10 neurology
10.64898/2026.08.07.26359605 medRxiv
Show abstract

Background: Multiple sclerosis subpial cortical lesions are prevalent and associated with disability but difficult to detect on MRI. Inversion recovery susceptibility weighted imaging with enhanced T2 weighting (IR-SWIET) and T1/T2 ratio imaging have been proposed for cortical lesion detection on 3 tesla (T) MRI. Objectives: To assess cortical lesion detection using IR-SWIET and T1/T2 ratio imaging. Methods: Cortical lesions were identified in 20 persons with MS (pwMS) independently on six image sets: T1 weighted (w) magnetization prepared 2 rapid acquisition gradient echoes (MP2RAGE) + T2w fluid attenuated inversion recovery (FLAIR) alone or with T1/T2, IR-SWIET single acquisition (x1), average of two (x2) or median of four (x4) acquisitions, or denoised single acquisition (IR-SWIETx1DN). In 10 additional pwMS with 7T-based cortical lesion segmentations, lesions were identified on MP2RAGE + FLAIR + IR-SWIETx1DN. Results: Median subpial lesions identified on MP2RAGE + FLAIR was 0 (interquartile range (IQR) 2) vs 0 with T1/T2 (IQR 1, p=0.07), 1 with IR-SWIETx1 (IQR 6, p=0.42), 5 with IR-SWIETx2 (IQR 5, p=0.008), 4 with IR-SWIETx4 (IQR 6, p=0.008), and 4 with IR-SWIETx1DN (IQR 6, p=0.008). Versus 7T, IR-SWIETx1DN detected subpial lesions with similar sensitivity to IR-SWIETx2. Conclusions: IR-SWIET, but not T1/T2, improves subpial cortical lesion detection. Denoising may be an efficient and sensitive alternative to multi-acquisition averaging.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.