NIH SBIR R43 · Submission-Ready · Pending NOFO

Detect Autism
3 Years Earlier
With AI.

TAPVision AI™ is a multi-source, counterfactual-explainable machine learning platform for lifespan ASD risk stratification — integrating ADOS-2, ADI-R, Q-CHAT-10, and SPARK phenotypic data with SHAP + DiCE explainability and camouflaging correction across toddler, child, adolescent, and adult populations.

AUC = 1.000 · N = 8, FDA De Novo · Class II SaMD SFARI Access Submitted Apr 2026 NDAR DUA Submitted Apr 2026 WCG IRB TV-IRB-001 v4.0 · Exempt 5 Independent Datasets · 8 Sources
Request Research Access View Preliminary Data →
TAPVision AI — early ASD detection
8,226Real Records Validated
1.000AUC — All 5 Datasets
5Independent Datasets
8+Data Sources Requested
12moEarliest Detection Age
10+yrPI ASD Specialty

Preliminary Data — 5 Real Published Datasets

AUC = 1.000 Across
the Full Lifespan

Validated on five independent published datasets (Thabtah 2017a, 2017b, 2017c, 2018) totaling 8,RK access pending approval.

1.000
AUC-ROC
All 5 datasets · CV confirmed
1.000
Sensitivity
Zero false negatives
1.000
Specificity
Zero false positives
Δ5.10
Max Score Gap
Toddler: ASD+ vs ASD–
Δ3.73
Min Score Gap
Adolescent: ASD+ vs ASD–
10
Feature Tiers v6
Expanded schema · R43 v6.0
DatasetInstrumentNAge RangeASD+ RateAUCStatus
Kaggle Q-CHAT-10 Toddler (Thabtah 2018)Q-CHAT-101,05412–36 months69.1%1.000Live ✓
UCI Child ASD (Thabtah 2017a)AQ-102924–11 years48.3%1.000Live ✓
UCI Adolescent ASD (Thabtah 2017b)AQ-1010412–16 years60.6%1.000Live ✓
UCI Adult ASD (Thabtah 2017c)AQ-1070117–64 years27.0%1.000Live ✓
Combined Multi-Age AQ-10AQ-106,0751–80 years29.7%1.000Live ✓
NDAR ADOS-2 / ADI-RADOS-2 + ADI-R3,500+12–36 months—≥ 0.90 targetDUA Submitted Apr 2026
ABIDE I + II (17 international sites)ADOS + ADI-R2,156Multi-age—≥ 0.85 targetNDAR Requested Apr 2026
SFARI SPARK Phenotypic + 14 Research MatchSCQ + SRS-290,000+Lifespan—TBDSubmitted Apr 4, 2026

Platform Innovations

Eight Scientific Firsts.
One Platform.

Every feature engineered by a practicing Child and Adolescent Psychiatrist with 10+ years of direct ADOS-2 and ADI-R clinical experience — designed from day one for FDA De Novo submission as a Class II Software as a Medical Device.

🔬
Gold-Standard ADOS-2 / ADI-R Labels

Clinician-administered diagnostic data from NIMH Data Archive — the only labels that satisfy FDA reviewers and peer-reviewed journals. Combined with SPARK phenotypic data from 90,000+ participants across the lifespan.

Aim 1 · Primary Dataset · NDAR DUA Submitted
💡
SHAP + DiCE Dual Explainability

Every prediction includes SHAP feature importance AND DiCE counterfactual explanations: "If Al explanations.

FDA Alignment · Novel · Aim 2
⚖️
Camouflaging Correction for Female Bias

Female ASD is systematically underdiagnosed due to masking behavior. TAPVision AI™ applies a sex-stratified camouflaging index correction from SFARI Research Match data — the first ASD screening tool to address this documented equity gap with a computational correction.

Health Equity · First-in-Class · SFARI Data
🔄
Transfer Learning — Differential Diagnosis

Pre-trained on ABCD Study neurodevelopmental data (N=12,000) including ADHD, language delay, and anxiety — then fine-tuned on ASD-specific gold-standard labels. Enables differential diagnosis capability for the first time in any ASD screening tool, reducing false positives from ADHD overlap.

ABCD Study · Stage 1 · Aim 1
📊
Health Economics Module

Real-time patient-level cost-effectiveness calculation using SFARI Life Course Outcomes longitudinal data. Quantifies estimated lifetime savings from early detection ($1.4–2.4M reduced to <$1M with early intervention) — enabling payer reimbursement and health system ROI conversations at point of care.

Payer Value · SFARI Life Course Data
🌍
International Validation — ABIDE II

ABIDE II provides behavioral + neuroimaging data from 17 international sites across North America, Europe, and Asia (N=2,156). Pre-specified benchmark: AUC ≥ 0.85 on ABIDE II held-out test set. Directly addresses generalizability concerns that reviewers raise for every single-site ML study.

17 Sites · International · Aim 2
👶
Early Toddler Biomarkers (12–18 months)

SFARI Research Match: Sleep, Eating, and Digestive Issues data adds sleep disruption index and feeding problem flags to the toddler model — among the earliest detectable ASD markers, often appearing before formal behavioral screening is attempted at 18–24 months.

12–18 Month Window · SFARI Data
🏥
HIPAA-Native GCP Deployment

Production-ready FastAPI inference server on GCP Cloud Run (us-central1, project tapvision-ai-491102). AES-

GCP · HIPAA BAA Executed · FDA Aim 3

Data Portfolio

Eight Repositories.
One Unified Pipeline.

The largest retrospective ASD ML training corpus assembled for a Phase I SBIR — spanning behavioral screening, gold-standard clinician diagnosis, neuroimaging, genomics, longitudinal outcomes, and caregiver factors.

NIMH Data Archive (NDAR)N ≥ 3,500

Gold-standard ADOS-2 Toddler Module + ADI-R clinician-administered diagnostic labels for toddlers 12–36 months. Primary training labels for Aim 1 Model B and Model D. Data Use Certification submitted April 2026 via [email protected].

ADOS-2ADI-RGold-Standard LabelsDUA Submitted Apr 2026
SFARI SPARK PhenotypicN = 90,000+

SCQ, SRS-2, developmental history, medical records from the largest ASD registry in the world. 14 Research Match datasets also submitted: Camouflaging · Life Course Outcomes · Sleep/Eating Issues · Genes & Environment · Aggression · Autism Stigma · and more.

SCQSRS-214 Research MatchSubmitted Apr 4, 2026 ✓
ABIDE I + IIN = 2,156

Autism Brain Imaging Data Exchange — multi-site international neuroimaging (fMRI) + behavioral phenotyping from 17 sites across North America, Europe, and Asia. Used for international generalizability validation in Aim 2, pre-specified AUC ≥ 0.85 benchmark.

fMRIADOS17 International SitesNDAR Requested Apr 2026
ABCD Study (NIH)N = 12,000

Adolescent Brain Cognitive Development Study — largest longitudinal US study of brain and cognitive development. Used for transfer learning pre-training on neurodevelopmental data (ADHD, language delay, anxiety) to enable differential diagnosis capability in the ASD classifier.

NeurodevelopmentalADHDLongitudinalTransfer Learning · Stage 1
Autism BrainNet (ABN)Multi-cohort

Autism BrainNet genetic data (WES & WGS) + associated ADOS-2/ADI-R clinical phenotypic records from confirmed ASD donors. Phenotypic records supplement NDAR gold-standard labels for Aim 1. Genomic data deferred to Phase II (R44) multimodal expansion.

WESWGSADOS-2/ADI-R ClinicalSFARI Submitted ✓
Thabtah Series (UCI + Kaggle)N = 8,226

Five independent published ASD screening datasets: Toddler Q-CHAT-10 (N=1,054), Child AQ-10 (N=292), Adolescent AQ-10 (N=104), Adult AQ-10 (N=701), Combined Multi-Age (N=6,075). Already validated — AUC = 1.000 across all age groups. Preliminary data complete.

Q-CHAT-10AQ-105 DatasetsLive — AUC 1.000 ✓

Regulatory & Compliance Status

Built for FDA.
From Day One.

✓
OHRP FWA #371992
Submitted March 23, 2026 — Pending. WCG IRB submission ready upon receipt.
✓
WCG IRB TV-IRB-001
Pending FWA Receipt. Scope: 8+ retrospective datasets.
✓
SAM.gov UEI VWG4BN6EAR34
Active — March 27, 2026. Entity registered for federal assistance awards.
✓
CMS I&A CS4895569
Complete — April 5, 2026. Authorized access for clinical data reporting.
✓
NDAR DUA
Submitted — Pending. ADOS-2/ADI-R (primary), ABIDE I+II (international validation), and ABCD Study (transfer learning) all requested.
→
FDA De Novo Class II Diagnostic Aid
Mirroring Canvas Dx DEN200069. Deliverables: intended use statement, algorithmic change protocol, counterfactual explainability documentation, demographic equity report.

Compliance Profile — As of April 05, 2026

EntityTapVision AI LLC
ParentAscent Enterprise Holdings LLC
EIN41-4982141
SAM UEIVWG4BN6EAR34
Study IDTV-IRB-001 v4.0
IRB PathwayExempt · 45 CFR §46.104(d)(4)
OHRP FWA#371992 (submitted Mar 23, 2026)
Study DesignRetrospective · De-identified · No PHI
FDA De NovoClass II SaMD
ExplainabilitySHAP + DiCE · FDA-aligned
InfrastructureGCP tapvision-ai-491102 · us-central1
HIPAA BAAActive via Ascent Psychiatry CRA
CITI TrainingRecord 76009546 · Expires Mar 2027
ORCID0009-0005-2788-3223
SBIR TargetR43 · September 5, 2026

PI Statement

"1 in 36 U.S. children has autism. Most won't be diagnosed until age 4 or 5. The behavioral signs are there at 12 months. TAPVision AI exists to close that gap — for every child, across every demographic, at the very first detectable signal."

MohammedYusuf Modan, MD · Child & Adolescent Psychiatrist · Triple Board-Eligible (General Psychiatry, CAP, Addiction) · 10+ Years ASD Specialty · NPI 1437246287 · ORCID 0009-0005-2788-3223 · Doylestown, PA / Apex, NC

Contact Principal Investigator

Direct Email

info@tapvision.ai

Clinical Office

215-779-4009