Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• LoV3D achieves 93.7% three-class diagnostic accuracy on a subject-level held-out ADNI test set, which is a 34.8% improvement over a no-grounding baseline.result 95%
• LoV3D achieves 97.2% two-class diagnostic accuracy, representing a 4% improvement over state-of-the-art methods.result 95%
• LoV3D achieves 82.6% region-level anatomical classification accuracy, which is a 33.1% improvement over vision-language model baselines.result 95%
• LoV3D demonstrates high generalizability with zero-shot transfer achieving 95.4% accuracy on MIRIAD dataset and 82.9% three-class accuracy on AIBL dataset.result 90%
• LoV3D reduces hallucination risks by enforcing label consistency, longitudinal coherence, and biological plausibility through a stepped pipeline.technical 85%
• LoV3D's training process uses a clinically-weighted Verifier that enables Direct Preference Optimization without human annotations.method 90%
• LoV3D might require extensive computational resources for training and face initial skepticism from clinicians due to reliance on automated systems.limitation 80%
• LoV3D produces a three-class diagnosis (Cognitively Normal, Mild Cognitive Impairment, or Dementia) along with a synthesized diagnostic summary from longitudinal brain MRI.technical 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card