LesionOrbit
Radiology assistant where the doctor types in plain English (e.g. 'find right upper lobe nodules') and the AI highlights matching regions on chest CT with a confidence score
General radiologists at community hospitals handling mixed-case workloads
- Plain-English query that returns highlighted lesion regions in seconds on chest or abdominal scans
- Confidence flag with a short explanation of why the model is or is not sure
- Side-by-side comparison with prior studies and tracked change over time
- One-click PACS export of the annotated image into the radiology report
BTHA's decoupled language-vision backbone means a single AI segmentation assistant can plug into many scanner-era datasets without retraining, finally making flexible text-query radiology feasible inside hospital workflows
Active market with documented adoption: Jefferson Health/RSNA 2024 pilot showed AI lung nodule detection reduced reading times for community radiologists, and MarketsandMarkets sizes the AI medical imaging market as a major growth segment with multiple commercial deployments.Lung Nodule Detection Algorithm Can Enhance Reading Efficiency | RSNA 2024 ↗Artificial Intelligence in Medical Imaging Market Report 2024-2029 ↗
Crowded niche (Aidoc, Viz.ai, Annalise.ai, Lunit, Qure.ai, Fractify, GE, Siemens) all offer fixed-workflow detection, but none offer natural-language text-query highlighting with confidence scores — that interaction model is genuinely differentiated vs. existing single-purpose detectors.Top Imaging & Pathology AI Companies: 2025 Market Analysis ↗AI Radiology 2026: Vendors, FDA, PACS ↗
Proven willingness to pay: 7-platform comparison shows $120K–$450K/year base contracts plus $0.50–$15 per-study fees; 5-year deals common and integration costs push TCO to ~3x quoted price — strong SaaS economics for a community-hospital radiology tool.Radiology AI Software Pricing in 2025: What 7 Platforms Actually Cost Hospitals ↗The Real Cost of Per-Study AI Pricing in Radiology ↗
Regulatory and reimbursement tailwinds (CMS payment for AI-augmented imaging, expanding FDA clearances) plus radiologist shortages make text-query second-readers a durable 10+ year category, not a fad.Top Imaging & Pathology AI Companies: 2025 Market Analysis ↗
Underlying tech is still arxiv-stage (BTHA, MedSAM variants, contrastive text-image alignment); needs FDA 510(k)/De Novo clearance, PACS/DICOM integration ($50K–$200K, 6+ weeks per hospital), and validation across mixed scanner-era datasets before clinical use.Decoupling Language Guidance from Backbones for Text-Guided Medical Segmentation ↗DescriptorMedSAM: language-image fusion with multi-aspect text ↗