AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Decrypting “Cryptogenic” Epilepsy: Semi-supervised Hierarchical Conditional Random Fields For Detecting Cortical Lesions In MRI-Negative Patients

Bilal Ahmed, Thomas Thesen, Karen E. Blackmon, Ruben Kuzniekcy, Orrin Devinsky, Carla E. Brodley

JMLR 2016 · Label-efficient and robust learning › Semi-supervised learning · approach: Semi-supervised medical segmentation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use HCRF multiscale outlier model instead of z-score thresholding (cortical thickness)lesion detection · MRI-negative FCD, top-10detection rate: 55% -> 70%Improved
use HCRF instead of z-score (GWC)lesion detection · MRI-negative FCD, top-10detection rate: 45% -> 50%Improved
use HCRF instead of z-score (curvature)lesion detection · MRI-negative FCD, top-10detection rate: 45% -> 65%Improved
use HCRF instead of z-score (sulcal depth)lesion detection · MRI-negative FCD, top-10detection rate: 60% -> 60%No clear effect
use HCRF instead of z-score (sulcal depth)lesion detection · MRI-negative FCDprecision: 6.0124 -> 8.7008Improved
use HCRF instead of z-score (cortical thickness)lesion detection · MRI-negative FCDrecall: 0.8035 -> 1.1140Improved
aggregate four features by averaging probabilities, HCRF vs baselinelesion detection · MRI-negative FCD, top-10detection rate: 55% -> 60%Improved
aggregate four features by maximum, HCRF vs baselinelesion detection · MRI-negative FCD, top-10detection rate: 50% -> 65%Improved
combine all four features (average or max) instead of cortical thickness alone (ablation)HCRF · MRI-negative FCD, alpha=1detection rate: 70% -> 65%Hurt
select only cortical thickness and curvature (feature selection) before combining (ablation)HCRF · MRI-negative FCD, alpha=1detection rateImproved
tune ranking tradeoff alpha with thickness+curvature averaged (ablation)HCRF · MRI-negative FCDdetection rate: 65% -> 75%Improved
increase alpha toward 1 (rank by cluster size) (ablation)HCRF GWC · MRI-negative FCDdetection rateHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy