Clinically Validated  ·  Published 2025

Diagnostics without
a specimen

A fingertip-based optical platform that reads your body's biochemical response — no blood draw, no swab, no reagents. A result in ~60 seconds.

See How It Works Request a Briefing
0.896
AUROC vs. PCR
455
Patients Enrolled
~60s
Point-of-Care
3
Issued U.S. Patents
123-CY Device
No Specimen Finger scan only
123-CY Working prototype

The Problem

Every diagnostic requires a specimen.
Until now.

All current diagnostics — PCR, sequencing, immunoassays — require specimen extraction, reagents, and laboratory workflows. This introduces latency, cost, and scaling constraints at exactly the wrong moment.

⏱
Workflow Latency
Specimen collection, transport, processing, and reporting create a chain of delays. Clinicians make empiric decisions while waiting — patients deteriorate in the gap.
📦
Supply Chain Fragility
Reagents and consumables are the first to fail in a surge — the exact conditions that demand the most from diagnostics. Cold chain dependencies compound the risk.
👤
Staffing Burden
Specimen collection and processing add burden during surge when staff capacity is most constrained. Every step requiring personnel is a bottleneck.
🎯
Single-Pathogen Output
Each test answers one question. New threat = new test, new reagent, new supply chain, new training. There is no platform flexibility in the current paradigm.

The 123IV Approach

A new sensing layer for
host biochemistry

Transcutaneous Raman spectroscopy acquires a high-dimensional biochemical signal directly from a fingertip. The platform reads the body's response — not the pathogen — making it threat-agnostic by architecture.

1
Optical Signal Acquisition
830 nm NIR laser directed transcutaneously at the fingertip. 3 × 20-second exposures. No specimen, no swab, no contact fluid. Fully enclosed Class 3B housing, below ANSI MPE safety limits.
2
Spectral Signal Capture
Back-illuminated CCD (SNR 6400:1) captures inelastically scattered Raman photons. Both raw and processed (pure Raman) spectra stored per encounter, timestamped.
3
2000-Dimensional Feature Vector
Each acquisition produces a 2000-dimensional Raman feature vector — a rich representation of systemic biochemical state. StandardScaler normalized on training set only to prevent data leakage.
4
Locked ML Classification
Gradient Boosted Tree classifier (500 estimators, 5-fold CV) identifies host-response patterns. New indications require only software retraining — no hardware redesign, no new manufacturing run.
From laser to Raman spectral output
NO SPECIMEN
123-CY — finger scan to spectral output to classification. Working prototype in clinical deployment at Cornell.

Clinical Validation

Prospective human study.
PCR gold standard.

455-patient real-world dataset. Locked training with prospective validation. Published in a peer-reviewed journal with MIT Open Access.

0.896
AUROC
5-fold cross-validation vs. nasal PCR gold standard
455
Patients
Prospective, IRB-approved study. 4 seasons, all variants
80%
Sensitivity
At specificity 83.7% — clinician-adjustable threshold
83.7%
Specificity
At sensitivity 80% — ROC curve tradeoff available
Study Design
Prospective, observational, non-interventional. Finger scan concurrent with nasal PCR at same encounter — no additional specimen from our device. GCP-compliant, informed consent. 148 COVID-19 PCR-positive, 307 PCR-negative patients across ER, inpatient, and outpatient settings.

Chefitz et al., Spectrosc. J. 2025, 3(1), 6 — MIT Open Access
Platform Robustness — Second Independent Target
Same 830 nm hardware retrained on glucose data demonstrated learnable transcutaneous signal (AUPR 0.58, 205 observations, Type 1 IDDM). COVID-19 and glucose are biologically unrelated targets — detecting both from the same platform via software-only adaptation is the proof that the platform is genuinely threat-agnostic.

Next: Prospective ER study at Weill Cornell Medicine. 350–450 patients. IRB submission pending.

Intellectual Property

Issued patents. Unencumbered freedom to operate.

Core IP covers both the hardware measurement method and the computational classification pipeline. All patents assigned to 123IV, Inc. No known third-party blocking positions identified.

Core Patent
Sensing & Measurement Interface
Finger-insertion optical interface enabling transcutaneous Raman acquisition. Pulse-oximeter form factor. Ambient light minimization. Non-invasive, no specimen required.
US 11,452,454
Core Patent
Spectroscopic Classification Pipeline
Machine learning-based pattern classification applied to host-response Raman spectral signatures. Covers the computational pipeline from spectral vector to clinical output.
US 11,304,605
Core Patent
Platform Technology
Broader platform protection covering the integrated optical acquisition and AI/ML classification system. Foreign patents also issued.
US 12,121,321
Additional Platforms: Autonomous IV Infusion System (3 issued U.S. patents: US 10,906,180; 11,872,708; 12,121,765 + foreign patents) · Intraoperative Cancer Margin Probe — multimodal optical device for real-time tumor margin assessment (patent pending).

Platform Applications

One hardware unit.
Software-defined indications.

New pathogen or condition = new ML module retrained on paired data, deployed remotely. No hardware redesign, no new manufacturing run, no new 510(k) per new target.

🦠
Clinically Validated
Infectious Disease Detection
COVID-19 detection via host-response biochemical signature. AUROC 0.896 in 455-patient prospective study vs. nasal PCR. Published peer-reviewed 2025.
📊
Signal Validated
Glucose / Metabolic Monitoring
Transcutaneous glucose signal demonstrated on same hardware (AUPR 0.58, 205 obs). Platform robustness proof — biologically unrelated target, software-only adaptation.
🏥
Active Deployment
ER Triage & Respiratory Illness
Prospective study at Weill Cornell Medicine Emergency Department. 350–450 patients, concurrent SOC comparator. Multi-site, high-throughput ER environment.
⚡
In Development
Early Sepsis Detection
Serial spectroscopic measurements to detect biochemical changes 2–6 hours before clinical sepsis criteria. Cornell ER longitudinal study, 500–1,000 patient target. BARDA AOI 7.4 aligned.
🌍
Pipeline
Threat-Agnostic Surge Readiness
Any pathogen triggering host immune/metabolic response is detectable. Same hardware pre-positioned — new threat emerges, ML module retrained and deployed remotely. BARDA stockpile model.
🔬
Pipeline
ICU Longitudinal Monitoring
Continuous biochemical trajectory monitoring. Same device as ER triage — software-defined extension adds a biochemical layer to the existing vital signs monitoring stack.

Team

Physician–engineer team with clinical, optics, and AI/ML execution.

ABC
Allen B. Chefitz, MD
Founder & CEO
Surgeon and lead inventor. IRB execution, ER deployment, clinical workflow design. Holds 6+ issued U.S. patents across multiple medical device platforms.
ELC
Elisa Long Chefitz, MS, MD
Co-Founder
Clinical strategy and translational medicine. Study design, regulatory pathway development, and clinical trial architecture for medical device platforms.
GV
Gregg Vesonder, PhD
Strategic Advisor
Systems engineering, AI/ML architecture. Advises on machine learning methodology, systems integration, and platform scalability strategy.
Collaborators & Partners AI/ML: MIT CSAIL · Stevens Institute of Technology  |  Clinical: Weill Cornell Medicine (Emergency Department)  |  Safety Testing: Sunrise Labs, Bedford NH

Contact

We're open to a conversation.

Whether you're a clinical partner, investor, or strategic collaborator — tell us what you're looking for and we'll respond directly.

✉
abc@123IVrobot.com
📞
914-980-2310
🌐
123IVrobot.com
📄
Read the published paper →
Chefitz et al., Spectrosc. J. 2025, 3(1), 6 (MIT Open Access)
Request a Briefing
For investors, clinical partners, and BARDA/government inquiries. We'll respond within one business day.

Your message will open in your email client. Alternatively, email abc@123IVrobot.com directly.