Idea
Deep learning platform automating rheumatoid arthritis joint damage scoring from hand X-rays for rheumatologists and radiologists.
Research Paper
Core Innovation
This paper introduces ARTSS, a multi-stage deep learning framework that automates Total Sharp Score assessment from full-hand X-rays. It uniquely handles challenges like joint disappearance and variable image lengths using Vision Transformer models, improving accuracy and reducing reader variability compared to manual scoring.
Market Size (TAM)
$2–10B TAM, $1–2B SAM; assumption: global rheumatoid arthritis diagnostics and imaging analysis market with growing AI adoption.
Potential Customers & Pain Points
- Rheumatologists Needing Consistent Joint Damage Assessment
- Radiologists Seeking Faster Accurate Scoring
- Hospitals Aiming To Reduce Diagnostic Variability
- Clinical Researchers Requiring Standardized Imaging Metrics
Business Model
SaaS platform with subscription fees for hospitals and clinics; API licensing for integration with radiology software; tiered pricing based on usage volume.
Competitive Landscape
- RheumaCare AI
- BoneXpert
- VUNO Med
Implementation Challenges
- Regulatory approval for clinical use
- Integration with existing hospital imaging systems
- Clinician trust and adoption of AI scoring
Validation Strategy
- Conduct multi-center clinical trials comparing ARTSS to expert readers
- Obtain regulatory clearance for diagnostic support use
- Pilot deployments in rheumatology clinics to gather real-world feedback
Research Paper Overview
Automated Radiographic Total Sharp Score (ARTSS) in Rheumatoid Arthritis: A Solution to Reduce Inter-Intra Reader Variation and Enhancing Clinical Practice
Summary
This study presents ARTSS, a deep learning framework to automate Total Sharp/Van Der Heijde Score (TSS) assessment from full-hand X-rays in rheumatoid arthritis patients. ARTSS addresses challenges like joint disappearance and variable image lengths through four stages: image pre-processing, hand segmentation, joint identification, and TSS prediction using models including Vision Transformer. Evaluated on 970 patients with high accuracy and low error, ARTSS reduces inter- and intra-reader variability, saves time, and supports better clinical decisions.