Abstract
Accelerated magnetic resonance imaging (MRI) reduces acquisitiontimebutcandegradeclinicallyrelevantimageinformation,potentially affecting downstream diagnosis. In this work, we study knee pathologyclassificationfromhighlyundersampledMRI,focusingonmultipleclinicallyrelevantfindingsincludingmeniscustear,cartilagepathology,andanteriorcruciateligament(ACL)abnormality.Ratherthanreconstructing fully sampled images or training task-specific classifiers directlyondegradedinputs,weproposeKNEE:Knowledge-guidedNeural Embedding Enhancement, a latent-space framework for recovering diagnostic representations from undersampled MRI. KNEE learns a flowbasedtransformationthatmapsundersampledembeddingsintothediagnosticembeddingspaceoffullysampledMRI,enablingdownstreamclassification using representations that better preserve pathology-relevant information.WeevaluateKNEEonacceleratedkneeMRIacrossmultiple pathology classification tasks and compare it with classifiers operating directly on undersampled representations as well as alternative embeddingtranslationmethods.OurresultsdemonstratethatKNEEprovides an effective alternative to image reconstruction, preserving diagnostic performanceunderaggressiveundersamplingwhileavoidingthecomputational cost of reconstructing high-quality images.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SASHIMI_050.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=I17jXNMDVh
BibTex
@InProceedings{AshLir_From_MICCAISAT2026,
author = { Ashlag, Liron AND Cahan, Noa AND Greenspan, Hayit},
title = { { From Undersampled to Diagnostic: Flow-Translated Embeddings for Knee Pathology Classification in Accelerated MRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17258},
month = {pending},
page = {pending}
}