# arXiv: Interpretable Multimodal MRI-Clinical Fusion for Functional Outcome Prediction After Acute Ischemic Stroke [Canonical reading](https://escapemeridian.com/readings/arxiv-interpretable-multimodal-mri-clinical-fusion-for-a79157b3) Sina Raeisadigh, Henning Müller and Abderrahmane Hedjoudje report a study of whether day-1 multimodal MRI adds prognostic value when predicting three-month functional outcome after acute ischemic stroke. They analyzed 66 patients using MRI data and clinical variables. In their eight-fold stratified group cross-validation, clinical-only modeling performed strongly, while MRI-only models had limited discrimination. FLAIR-clinical fusion had the highest mean cross-validated performance; ADC-clinical fusion followed. Combining ADC, DWI and FLAIR did not consistently outperform fusion of clinical data with a single MRI modality. The authors say these results suggest that FLAIR and ADC representations provide complementary prognostic information beyond clinical variables. Their analysis also identified age, NIHSS subscores, ADC lesion burden and learned FLAIR representations as key contributors; qualitative FLAIR activation maps responded over lesion-like regions. As an Escape Meridian interpretation, the results make a specific case for testing compact MRI representations alongside clinical information, rather than assuming that adding every MRI modality improves prediction. The abstract describes a 66-patient analysis using cross-validation, so these reported results should be understood in that study context. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-09T00:53:28.611Z. First observed: 2026-11-06T05:07:30.693Z. Retrieved: 2026-11-09T00:35:29.121Z. - [Interpretable Multimodal MRI-Clinical Fusion for Functional Outcome Prediction After Acute Ischemic Stroke](https://arxiv.org/abs/2611.08154) ## Limits The abstract reports results from 66 patients using eight-fold stratified group cross-validation; it does not establish whether the reported performance holds in an independent patient cohort. ## Open question Would the reported FLAIR-clinical performance hold in an independent patient cohort? ## Source-stated The study analyzed day-1 multimodal MRI and clinical data from 66 patients to predict three-month functional outcome after acute ischemic stroke. > Sixty-six AIS patients with ADC, DWI, FLAIR, and clinical data were analyzed. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated Clinical-only modeling performed strongly, while MRI-only models had limited discrimination in the reported cross-validation. > Clinical-only modeling achieved strong performance (AUC = $0.899 \\pm 0.151$), while MRI-only models showed limited discrimination (AUC $\\approx$ $0.67$-$0.73$). - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated FLAIR-clinical fusion had the highest mean cross-validated performance among the multimodal configurations reported. > Among multimodal configurations, FLAIR-clinical fusion achieved the highest mean cross-validated performance (AUC = $0.948 \\pm 0.061$), followed by ADC-clinical fusion (AUC = $0.941 \\pm 0.094$). - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated Full ADC-DWI-FLAIR fusion did not consistently outperform clinical fusion with a single modality. > Full ADC-DWI-FLAIR fusion did not consistently outperform single-modality clinical fusion. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated The authors interpret the findings as suggesting that compact day-1 FLAIR and ADC representations provide complementary prognostic information beyond clinical variables. > These findings suggest that compact day-1 FLAIR and ADC representations provide complementary prognostic information beyond clinical variables - [Supporting source](https://arxiv.org/abs/2611.08154) ## Escape Meridian interpretation The abstract reports cross-validation results from 66 patients; whether the reported FLAIR-clinical performance holds in an independent cohort remains a question for further testing. > Sixty-six AIS patients with ADC, DWI, FLAIR, and clinical data were analyzed. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source and contributor context Raeisadigh, Sina, Müller, Henning, Hedjoudje, Abderrahmane · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Multimodal model A model that works with more than one kind of input or output, such as text, images and audio. - [Definition reference](https://developers.google.com/machine-learning/glossary#multimodal-model) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 38baa827-f44c-4c94-87fa-a0ccebf5ca03. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-09T00:53:28.611Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 83466d31e0e365939fb56cf965e7e909dd1cc0b3f7d179d784116c81939d5f94. Source text hash: b9187c7139b270bbf4a5ac52d9bf7bb79b3b25afe26f162449bd0660c044261f. ## Revision and source-check history ### Version 1 — 2026-11-09T00:53:28.611Z Sina Raeisadigh, Henning Müller and Abderrahmane Hedjoudje report a study of whether day-1 multimodal MRI adds prognostic value when predicting three-month functional outcome after acute ischemic stroke. They analyzed 66 patients using MRI data and clinical variables. In their eight-fold stratified group cross-validation, clinical-only modeling performed strongly, while MRI-only models had limited discrimination. FLAIR-clinical fusion had the highest mean cross-validated performance; ADC-clinical fusion followed. Combining ADC, DWI and FLAIR did not consistently outperform fusion of clinical data with a single MRI modality. The authors say these results suggest that FLAIR and ADC representations provide complementary prognostic information beyond clinical variables. Their analysis also identified age, NIHSS subscores, ADC lesion burden and learned FLAIR representations as key contributors; qualitative FLAIR activation maps responded over lesion-like regions. As an Escape Meridian interpretation, the results make a specific case for testing compact MRI representations alongside clinical information, rather than assuming that adding every MRI modality improves prediction. The abstract describes a 66-patient analysis using cross-validation, so these reported results should be understood in that study context. ## Original source arXiv cs.AI · Article Original source date: Nov 5, 2026 (2026-11-05). Date basis: Date basis not provided. Added to Escape Meridian: 2026-11-09T00:53:28.611Z. First observed: 2026-11-06T05:07:30.693Z. Retrieved: 2026-11-09T00:35:29.121Z. - [Interpretable Multimodal MRI-Clinical Fusion for Functional Outcome Prediction After Acute Ischemic Stroke](https://arxiv.org/abs/2611.08154) ## Limits The abstract reports results from 66 patients using eight-fold stratified group cross-validation; it does not establish whether the reported performance holds in an independent patient cohort. ## Open question Would the reported FLAIR-clinical performance hold in an independent patient cohort? ## Source-stated The study analyzed day-1 multimodal MRI and clinical data from 66 patients to predict three-month functional outcome after acute ischemic stroke. > Sixty-six AIS patients with ADC, DWI, FLAIR, and clinical data were analyzed. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated Clinical-only modeling performed strongly, while MRI-only models had limited discrimination in the reported cross-validation. > Clinical-only modeling achieved strong performance (AUC = $0.899 \\pm 0.151$), while MRI-only models showed limited discrimination (AUC $\\approx$ $0.67$-$0.73$). - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated FLAIR-clinical fusion had the highest mean cross-validated performance among the multimodal configurations reported. > Among multimodal configurations, FLAIR-clinical fusion achieved the highest mean cross-validated performance (AUC = $0.948 \\pm 0.061$), followed by ADC-clinical fusion (AUC = $0.941 \\pm 0.094$). - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated Full ADC-DWI-FLAIR fusion did not consistently outperform clinical fusion with a single modality. > Full ADC-DWI-FLAIR fusion did not consistently outperform single-modality clinical fusion. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source-stated The authors interpret the findings as suggesting that compact day-1 FLAIR and ADC representations provide complementary prognostic information beyond clinical variables. > These findings suggest that compact day-1 FLAIR and ADC representations provide complementary prognostic information beyond clinical variables - [Supporting source](https://arxiv.org/abs/2611.08154) ## Escape Meridian interpretation The abstract reports cross-validation results from 66 patients; whether the reported FLAIR-clinical performance holds in an independent cohort remains a question for further testing. > Sixty-six AIS patients with ADC, DWI, FLAIR, and clinical data were analyzed. - [Supporting source](https://arxiv.org/abs/2611.08154) ## Source and contributor context Raeisadigh, Sina, Müller, Henning, Hedjoudje, Abderrahmane · Article author These are credits stated by the source; they do not establish a profile match or authorship of every passage. ## Source media No suitable media was found by the supported page extraction. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Multimodal model A model that works with more than one kind of input or output, such as text, images and audio. - [Definition reference](https://developers.google.com/machine-learning/glossary#multimodal-model) ## arXiv arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed. - [Explore this context](/entities/arxiv) - [Reference](https://arxiv.org/) - [Reference](https://info.arxiv.org/help/api/index.html) ## Editorial provenance Version 1. Publication ID: 38baa827-f44c-4c94-87fa-a0ccebf5ca03. Prepared by gpt-6-luna; grounding review by gpt-6-luna. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-09T00:53:28.611Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: material\_development. Observation: 83466d31e0e365939fb56cf965e7e909dd1cc0b3f7d179d784116c81939d5f94. Source text hash: b9187c7139b270bbf4a5ac52d9bf7bb79b3b25afe26f162449bd0660c044261f. ### uncertain — 2026-11-09T00:53:28.611Z — anthropic/claude-opus-5.5 Publication: 38baa827-f44c-4c94-87fa-a0ccebf5ca03 attribution — pass: The authors (Sina Raeisadigh, Henning Müller, Abderrahmane Hedjoudje) match the arXiv listing. The interpretive claim about complementary prognostic information is attributed to the authors, and the Escape Meridian inference is labeled as such. > Authors:Sina Raeisadigh, Henning Müller, Abderrahmane Hedjoudje summary — uncertain: The summary is faithful but hard for a non-specialist to follow. It never says that FLAIR, ADC and DWI are types of MRI scan, or what functional outcome, NIHSS or discrimination mean. It also omits the AI method: pretrained ResNet image embeddings with linear support vector machines. Without that, a reader may not see why compact or learned FLAIR representations matter or how the paper relates to AI. Smallest repair: one plain clause saying a pretrained image model turned each MRI scan type into compact features that were combined with clinical data, plus tooltips for FLAIR, ADC, DWI, NIHSS and AUC. > Three-dimensional pretrained ResNet embeddings were extracted from each MRI modality, reduced using fold-isolated principal component analysis date — pass: The originalPublished date of 2026-11-05 matches the v1 submission date shown on arXiv. The feed date of 2026-11-06 is correctly not used. > \[v1\] Thu, 5 Nov 2026 15:14:58 UTC (328 KB) support — pass: Each claim matches the abstract. The draft does not overstate the result: it reports cross-validated means, keeps the 66-patient context, and frames the independent-cohort question as open. The Escape Meridian interpretation, which argues against assuming that adding every MRI type helps, is conditional and rests on the full-fusion finding. No real-world clinical reliability is implied. > Full ADC-DWI-FLAIR fusion did not consistently outperform single-modality clinical fusion. media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/a79157b3e478dec11553d86cf47946ab8ba8fde11b3e44293becf1db44bcce5e/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/arxiv-interpretable-multimodal-mri-clinical-fusion-for-a79157b3.json)