Appendix D — Medical AI Tools by Specialty

TL;DR

FDA-cleared AI tools by specialty (2024):

  • Radiology: Most AI tools (70%+). Aidoc, Viz.ai, RapidAI for emergency findings
  • Cardiology: EchoGo, Caption AI, HeartFlow FFR-CT
  • Pathology: Paige Prostate (De Novo cleared), PathAI expanding
  • Ophthalmology: IDx-DR (autonomous, CPT reimbursement)
  • Dermatology: Limited validated options; consumer apps unreliable

Critical reminder: FDA clearance ≠ clinical effectiveness. Check external validation studies before adoption.

Introduction

This appendix catalogs FDA-cleared/approved AI tools and widely-used decision support systems organized by medical specialty. Information current as of 2024. Note: FDA clearance does not guarantee clinical effectiveness. Physicians should review validation evidence before adoption.

Legend: - FDA Status: 510(k) = cleared, PMA = approved, De Novo = novel device clearance - Evidence Level: RCT = randomized controlled trial, Prospective = prospective cohort, Retrospective = retrospective validation only

Radiology

Chest X-Ray

Tool Vendor FDA Status Use Case Evidence Level Notes
Annalise CXR Annalise.ai 510(k) Multi-finding detection (124 findings) Retrospective Comprehensive abnormality detection
ClearRead CT Riverain Technologies 510(k) Lung nodule detection Prospective Bone suppression technology
Lunit INSIGHT CXR Lunit 510(k) Pneumothorax, nodule, effusion detection Prospective High sensitivity (>95%) for PTX
qXR Qure.ai 510(k) TB, COVID-19, nodule detection Prospective Validated in low-resource settings
RedDot Behold.ai 510(k) Triage abnormal CXRs Retrospective Prioritizes urgent findings

Chest CT

Tool Vendor FDA Status Use Case Evidence Level Notes
Aidoc Pulmonary Embolism Aidoc 510(k) PE detection, triage Prospective Integrated with PACS, mobile alerts
Arterys Pulmonary Workstation Arterys 510(k) Lung nodule quantification Retrospective FDA first cloud-based AI (2017)
Lung AI Thirona 510(k) Emphysema quantification Prospective COPD assessment
AVIEW LCS Coreline Soft 510(k) Lung cancer screening, nodule management Retrospective Lung-RADS integration
Veolity MeVis 510(k) Automated lung nodule volumetry Prospective Tumor tracking, treatment response

Head CT

Tool Vendor FDA Status Use Case Evidence Level Notes
Aidoc ICH Aidoc 510(k) Intracranial hemorrhage detection Prospective Flags all hemorrhage types
Brainomix e-ASPECTS Brainomix 510(k) Stroke ASPECTS scoring Prospective Quantifies early ischemic changes
RapidAI iSchemaView 510(k) LVO detection, perfusion analysis RCT Reduces time-to-treatment
Viz.ai LVO Viz.ai 510(k) LVO stroke triage RCT (ENDOVASCT-trial) Mobile team notification
Avicenna cHeadCT Avicenna.AI 510(k) Multi-pathology triage (ICH, fracture, mass) Prospective Comprehensive head CT analysis

Mammography

Tool Vendor FDA Status Use Case Evidence Level Notes
Transpara ScreenPoint Medical 510(k) Breast cancer detection RCT Improves detection, reduces false positives
Genius AI Hologic 510(k) CAD for tomosynthesis Prospective Integrated with Hologic DBT
ProFound AI iCAD 510(k) Breast cancer detection Retrospective Deep learning CAD
Mia (Second Reader) Kheiron Medical CE Mark (EU) Breast cancer detection RCT (Sweden) Supports single-reader workflow
Volpara Volpara Health 510(k) Breast density assessment Prospective Automated BI-RADS density classification

Cardiac Imaging

Tool Vendor FDA Status Use Case Evidence Level Notes
Arterys Cardio AI Arterys 510(k) Cardiac MRI quantification (EF, volumes) Retrospective Automated ventricular assessment
Circle CVI Circle Cardiovascular Imaging 510(k) Cardiac MRI/CT post-processing Prospective FDA-cleared suite
ClearRead + Bone Suppression Riverain 510(k) Rib suppression for cardiac imaging Retrospective Improves cardiac silhouette evaluation
CT FFR HeartFlow PMA Fractional flow reserve from CT RCT (PLATFORM) Non-invasive coronary physiology

Ophthalmology

Tool Vendor FDA Status Use Case Evidence Level Notes
IDx-DR Digital Diagnostics De Novo Diabetic retinopathy screening RCT First autonomous AI diagnostic (2018)
EyeArt Eyenuk 510(k) Diabetic retinopathy detection Prospective Point-of-care screening
RETFound Multiple vendors Research Foundation model for retinal disease Retrospective Emerging technology
LumineticsCore Visulytix 510(k) AMD progression risk Retrospective Predicts geographic atrophy growth
Altris AI AMD Altris AI 510(k) AMD detection and monitoring Retrospective OCT analysis

Dermatology

See Dermatology for detailed clinical evaluation of these tools, including skin tone bias concerns.

FDA-Cleared/Approved Devices

Tool Vendor FDA Status Use Case Evidence Level Notes
DermaSensor DermaSensor Inc. De Novo (DEN230008), Jan 2024 Skin cancer evaluation for primary care Prospective (Mayo Clinic, 22 sites, 1,000+ pts) 96% sensitivity; first AI device for primary care skin cancer detection
Nevisense SciBase AB PMA (P150046), June 2017 Melanoma detection for dermatologists Prospective (multicenter, 2,416 lesions) 96.6% sensitivity; only FDA-approved device for melanoma detection

Other Tools

Tool Vendor FDA Status Use Case Evidence Level Notes
DermAssist Google Research Skin condition identification Retrospective Not yet FDA-cleared
SkinVision SkinVision CE Mark (EU) Melanoma risk assessment Retrospective Smartphone app; sensitivity 7-73% in studies
MoleScope MetaOptima 510(k) (device only) Dermoscopy imaging N/A Device cleared, AI pending
VisualDx VisualDx FDA-Exempt (CDSS) Diagnostic decision support N/A Clinical decision support; not diagnostic device
Warning

Consumer apps are not recommended. Studies show melanoma detection apps have sensitivity ranging from 7-73% (Freeman et al., 2020). False reassurance can delay diagnosis.

Cardiology

ECG Analysis

Tool Vendor FDA Status Use Case Evidence Level Notes
Apple Watch ECG Apple 510(k) Atrial fibrillation detection Prospective (Apple Heart Study) Consumer wearable
KardiaMobile AliveCor 510(k) AFib, bradycardia, tachycardia detection Prospective Smartphone-based ECG
Cardiologs Cardiologs 510(k) Multi-arrhythmia detection Retrospective Cloud-based ECG interpretation
Eko AI Eko Health 510(k) AFib detection from digital stethoscope Prospective Point-of-care screening

Echocardiography

Tool Vendor FDA Status Use Case Evidence Level Notes
Caption Guidance Caption Health 510(k) AI-guided echo acquisition Prospective Helps non-experts capture images
Us2.ai Us2.ai CE Mark (EU) Automated ejection fraction Retrospective Point-of-care EF assessment
TTE.ai Bay Labs 510(k) Automated EF, view classification Retrospective Workflow efficiency tool

Pathology

Tool Vendor FDA Status Use Case Evidence Level Notes
Paige Prostate Paige 510(k) Prostate cancer detection Retrospective First FDA-cleared digital pathology AI
PathAI PathAI Research Multi-organ pathology analysis Retrospective Partnership with pharma for trials
Prosigna Veracyte PMA Breast cancer gene signature RCT Predicts recurrence risk
Ibex Galen Ibex CE Mark (EU) Breast, prostate cancer detection Retrospective Whole slide image analysis

Neurology

Tool Vendor FDA Status Use Case Evidence Level Notes
Viz.ai LVO Viz.ai 510(k) Stroke detection (see Radiology) RCT Cross-specialty tool
RapidAI iSchemaView 510(k) Stroke imaging analysis RCT Perfusion, ASPECTS, LVO
Brainomix e-Stroke Brainomix 510(k) Stroke workflow suite Prospective Combines multiple analyses
Icometrix icobrain CE Mark (EU) MS lesion quantification Prospective Brain volumetrics
Cortechs.ai Cortechs Labs 510(k) Brain MRI volumetrics Retrospective Neurodegenerative disease tracking

Primary Care / Internal Medicine

Tool Vendor FDA Status Use Case Evidence Level Notes
Epic Sepsis Model Epic Systems N/A (EHR-embedded) Sepsis risk prediction Retrospective (poor external validation) See Case Studies appendix, controversial
Ambient Clinical Documentation Multiple (Nuance DAX, Abridge, Suki) N/A (non-diagnostic) AI-generated clinical notes Prospective Reduces documentation burden
UpToDate with AI Wolters Kluwer N/A (CDSS) Clinical decision support N/A Evidence summaries, not AI diagnostic

Gastroenterology

Tool Vendor FDA Status Use Case Evidence Level Notes
GI Genius Medtronic 510(k) Colonoscopy polyp detection RCT Real-time CAD during procedure
EndoBRAIN Olympus/Cybernet Research (Japan approval) Polyp characterization Prospective Predicts histology in real-time
CAD EYE Fujifilm CE Mark (EU) Polyp detection during colonoscopy Prospective Integrated with endoscope

Oncology

Tool Vendor FDA Status Use Case Evidence Level Notes
Oncotype DX Exact Sciences PMA (not AI, genomic test) Breast cancer recurrence risk RCT Guides chemotherapy decisions
Tempus Tempus Labs N/A (genomic analysis) Precision oncology Retrospective AI-powered genomic matching to trials
Foundation Medicine Roche PMA Comprehensive genomic profiling Retrospective Identifies actionable mutations
Paige Prostate Paige 510(k) (see Pathology) Prostate cancer detection Retrospective Digital pathology AI

Anesthesiology / Critical Care

Tool Vendor FDA Status Use Case Evidence Level Notes
Masimo sedation monitoring Masimo 510(k) Depth of sedation assessment Prospective Non-AI advanced monitoring
Edwards HPI Edwards Lifesciences 510(k) Hypotension prediction Prospective Predicts hypotension 15 min advance
ICU EarlyWarning Systems Multiple vendors Variable Sepsis, deterioration prediction Mixed Many still experimental

Pulmonology

Tool Vendor FDA Status Use Case Evidence Level Notes
Qure.ai qXR Qure.ai 510(k) TB detection on CXR Prospective Validated in endemic regions
CAD4TB Delft Imaging CE Mark (EU) TB screening Prospective WHO-endorsed for resource-limited settings
Lung Texture Analysis VIDA Diagnostics 510(k) COPD, ILD quantification Prospective CT-based lung parenchyma analysis

Mental Health / Psychiatry

Tool Vendor FDA Status Use Case Evidence Level Notes
Woebot Woebot Health N/A (wellness, not diagnostic) CBT chatbot Prospective Mental health support app
Ginger Ginger (Headspace Health) N/A Behavioral health coaching Retrospective AI triage + human coaches
Mindstrong Mindstrong Health Research Smartphone keystroke patterns predict mood Retrospective Experimental, not yet FDA-cleared

Note: AI in psychiatry is highly sensitive. Regulatory scrutiny is intense, and adoption is slow.

Obstetrics / Maternal-Fetal Medicine

Tool Vendor FDA Status Use Case Evidence Level Notes
Sonio AI Sonio CE Mark (EU) Ultrasound guidance, anomaly detection Retrospective Assists sonographers
Prenatal AI Multiple vendors Research Fetal anomaly detection Retrospective Emerging field, limited deployment

Emergency Medicine

Many emergency AI tools cross specialties (e.g., head CT hemorrhage detection, CXR pneumothorax). See Radiology section for trauma/emergency imaging AI.

Tool Vendor FDA Status Use Case Evidence Level Notes
Viz.ai Viz.ai 510(k) Stroke, PE, aneurysm triage RCT Multi-pathology platform
Aidoc Aidoc 510(k) Multi-pathology ED triage Prospective ICH, PE, c-spine fracture, etc.

Urology

Tool Vendor FDA Status Use Case Evidence Level Notes
Arterys Prostate MRI Arterys 510(k) Prostate MRI PI-RADS scoring Retrospective Automated lesion detection
Watson for Genomics IBM (discontinued) N/A Precision oncology for urologic cancers Retrospective See IBM Watson case study, failed

Limitations and Caveats

1. FDA clearance does not equal effectiveness: Clearance means device is “substantially equivalent” to predicate (510k) or meets safety/efficacy standards (PMA). Does not guarantee clinical benefit in all settings.

2. Evidence quality varies: Many tools have only retrospective validation. Prospective, multi-site RCTs are gold standard but rare.

3. Continuous updates: AI tools evolve. This catalog reflects 2024 status. Check FDA database and vendor websites for updates.

4. Reimbursement often absent: Few AI tools have dedicated CPT codes. Check with payers before assuming coverage.

5. Integration challenges: FDA-cleared tool may not integrate with your EHR, PACS, or workflow without significant IT effort.

6. Performance varies by population: Models trained on one population may underperform in yours. Validate locally before widespread use.

7. Vendor claims vs. reality: Marketing often overstates capabilities. Demand peer-reviewed publications, not just vendor white papers.

How to Evaluate an AI Tool for Your Practice

Before purchasing/implementing:

  1. Check FDA status: Cleared/approved? What pathway? What’s the intended use?
  2. Review validation evidence: Peer-reviewed publications? Prospective studies? External validation? RCTs?
  3. Assess generalizability: Was it validated on patients like yours (demographics, disease prevalence, imaging equipment)?
  4. Understand failure modes: What does it miss? When does it produce false positives?
  5. Evaluate integration: Does it work with your IT infrastructure? Training required? Support available?
  6. Consider cost-effectiveness: Upfront cost? Subscription fees? Reimbursement available? ROI clear?
  7. Seek peer experiences: Ask colleagues at other institutions about real-world performance, workflow impact.
  8. Pilot before scaling: Test on small cohort, validate locally, gather user feedback before institution-wide deployment.

Conclusion

This catalog provides starting point for exploring specialty-specific AI tools. The field evolves rapidly: new tools emerge, existing tools improve or fail. Physicians must stay informed, demand evidence, and critically evaluate AI claims before adoption.

For updated FDA clearances, visit: https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices

Remember: AI is tool, not replacement. Physician judgment, contextual understanding, and patient relationships remain irreplaceable.