{"id":3146,"date":"2026-11-17T10:12:42","date_gmt":"2026-11-17T03:12:42","guid":{"rendered":"https:\/\/research.binus.ac.id\/airdc\/?p=3146"},"modified":"2026-11-17T10:12:42","modified_gmt":"2026-11-17T03:12:42","slug":"machine-learning-analysis-of-longitudinal-biomarker-trajectories-for-post-stroke-recovery-prediction","status":"publish","type":"post","link":"https:\/\/research.binus.ac.id\/airdc\/2026\/11\/machine-learning-analysis-of-longitudinal-biomarker-trajectories-for-post-stroke-recovery-prediction\/","title":{"rendered":"Machine Learning Analysis of Longitudinal Biomarker Trajectories for Post-Stroke Recovery Prediction"},"content":{"rendered":"<p>Predicting neurological recovery after acute ischemic stroke (AIS) remains a major clinical challenge. Although static admission biomarkers offer limited prognostic value, the longitudinal trajectory of inflammatory and hemostatic markers over the subacute period may provide richer predictive information. We analyzed a clinical trial dataset of 52 AIS patients who received oral DLBS1033 as adjunctive therapy. We took serial measurements of TNF-\u03b1, IL-6, D-Dimer, MMP-9, and carotid intima-media thickness (CIMT) on Days 0, 14, and 28. We engineered trajectory features, absolute change, rate of change, trapezoidal area under the curve (AUC), and variability per biomarker and combined them with baseline clinical data. Five machine learning classifiers were evaluated under repeated stratified 5-fold cross-validation across three feature sets of increasing temporal richness, plus a Spearman-guided top-7 feature selection step. Random Forest with the selected features achieved the best cross-validated AUC-ROC of 0.872 \u00b1 0.122\u2014a 16.6 percentage point improvement over static baseline approaches, with in-sample bootstrap confirmation of 0.962 (95 % CI: 0.909\u20130.994). Notably, 5 of the 7 selected features are trajectory derived, indicating that temporal dynamics provide additional prognostic information. The leading trajectory predictors were MMP-9 trajectory variance and 14-day delta, TNF-\u03b1 AUC, and CIMT variance. Unsupervised clustering revealed two clinically distinct patient phenotypes that were separable by trajectory pattern, but not baseline scores. These findings demonstrate that longitudinal biomarker trajectories, when appropriately selected, substantially outperform static admission models for post-stroke recovery prediction.<\/p>\n<p>Authors:<br \/>\nJason Sebastian Sulistyawan, Mahmud Isnan, and Bens Pardamean.<\/p>\n<p><em>The 11th International Conference on Computer Science and Computational Intelligence (ICCSCI)<\/p>\n<p><\/em><a href=\"https:\/\/www.researchgate.net\/publication\/407211374_Machine_Learning_Analysis_of_Longitudinal_Biomarker_Trajectories_for_Post-Stroke_Recovery_Prediction\">Read Full Text<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Predicting neurological recovery after acute ischemic stroke (AIS) remains a major clinical challenge. Although static admission biomarkers offer limited prognostic value, the longitudinal trajectory of inflammatory and hemostatic markers over the subacute period may provide richer predictive information. We analyzed a clinical trial dataset of 52 AIS patients who received oral DLBS1033 as adjunctive therapy. [&hellip;]<\/p>\n","protected":false},"author":14,"featured_media":3147,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[],"class_list":["post-3146","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-publications"],"_links":{"self":[{"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/posts\/3146","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/users\/14"}],"replies":[{"embeddable":true,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/comments?post=3146"}],"version-history":[{"count":1,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/posts\/3146\/revisions"}],"predecessor-version":[{"id":3148,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/posts\/3146\/revisions\/3148"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/media\/3147"}],"wp:attachment":[{"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/media?parent=3146"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/categories?post=3146"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/research.binus.ac.id\/airdc\/wp-json\/wp\/v2\/tags?post=3146"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}