CardioSpectra Lab · Investigational platform · 2026

See the heart
as it actually is.

A multimodal intracardiac platform that characterizes myocardial tissue (optical, electrical and mechanical) in real time, at the catheter tip. It bridges the gap between non-invasive imaging and conventional intracardiac mapping to enhance both diagnosis and the precision of catheter-based therapy.

Explore the platform Talk to founders
Device class
8 Fr deflectable diagnostic catheter
Modalities
Optical · Electrical · Mechanicalco-located
Software
FusionSuite SaMDlate-fusion AI
Inference
< 250 msacquisition → display
Scroll
01 · The diagnostic gap

Today's Heart: Fully Mapped,
Far from Understood

Electroanatomic systems show where electricity travels. Contact-force catheters report mechanical touch. Non-invasive imaging characterizes tissue, but only before the procedure, never at the catheter tip. None of them, at the point of contact, tell the operator what the tissue actually is. That blind spot spans a broad class of fibrotic and infiltrative myocardial disease: cardiac fibrosis, non-ischemic cardiomyopathy, and infiltrative disease such as amyloidosis01. It is why ablation substrate is still inferred from voltage alone, and why tissue diagnosis often falls back on endomyocardial biopsy, which is invasive and prone to sampling error.

0%
of HFpEF patients carry undiagnosed cardiac amyloidosis, one of many infiltrative diseases that hide from voltage mapping. González-López et al. · Eur Heart J 2015
$0B
Global EP ablation market by 2030. Every substrate-guided procedure is a candidate for point-of-contact tissue data. iData Research · 2024 forecast
0in 1
optical, electrical and mechanical signals co-registered to the same square millimetre of tissue, where today they live in separate systems, if they exist at all. CardioSpectra platform architecture
0M
Cardiac implantable devices placed worldwide each year: the long-tail opportunity for embedded sensing (see roadmap). WHO MedTech registry · 2024
02 · The platform

Multimodal Tissue Characterization
from a Single Contact Point

An 8 French, 110 cm deflectable catheter (up to 180° tip deflection) that integrates a sapphire optical window, four platinum–iridium ring electrodes and a piezoelectric micro-elastography module within a single distal sensing head. Every acquisition is ECG-gated to end-diastole and fires only when catheter–tissue contact is verified stable, so optical, electrical and mechanical data describe the same square millimetre of tissue at the same instant.

Modality 01 / Optical
Hyperspectral Reflectance
450–1000 nm · VIS–NIR

A GRIN-coupled fibre delivers broadband VIS–NIR light through a sapphire window into the contacted tissue and collects the diffuse reflectance back to an on-console spectrometer. The result is a per-contact biochemical and microstructural fingerprint of chromophore composition, scattering and tissue organization. It separates fibrosis, infiltration (including amyloidosis), ischemia and inflammation.04

Range450–1000 nm
OpticsGRIN · sapphire
Working dist.~0.5–1.0 mm
Modality 02 / Electrical
Tetrapolar Bio-Impedance
Complex Z · magnitude + phase

Four circumferential Pt–Ir ring electrodes measure complex impedance (magnitude and phase) across the tissue interface. It reads as a tissue descriptor (cellular integrity, extracellular matrix) and, at the same time, as a real-time metric of catheter–tissue contact quality that gates every acquisition.05

Electrodes4× Pt-Ir ring
Readout|Z| + phase
Doubles asContact metric
Modality 03 / Mechanical
Piezoelectric Elastography
kHz – low-MHz excitation

A miniaturized actuator/sensor pair delivers low-amplitude, kHz-to-low-MHz excitation and reads the tissue's response. From phase lag, amplitude attenuation and resonance shift we derive a relative stiffness index, a physical correlate of fibrosis and infiltration. Actuation is held below established safety thresholds.06

ExcitationkHz–low-MHz
OutputRel. stiffness idx
ActuationBelow safety limit
Specification
Value
Tolerance
Standard
Shaft outer diameter
8 Fr · 2.67 mm
±0.04 mm
ISO 11070
Working length
110 cm
±1 cm
Internal
Deflection
Bidirectional · 180°
±5°
EP standard
Shaft construction
Multi-lumen Pebax · braided SS / nitinol
Cl. VI
ISO 10993-1 / -5 / -10
Acquisition
ECG-gated · end-diastole
Contact-verified
Common clock
AI latency
< 250 ms · acquisition → display
Per acquisition
On-console GPU
Tissue classes
Normal · Fibrotic · Infiltrative · Ischemic · Inflamed
Investigational
Histology-referenced · preclinical
03 · Applications

Tissue Intelligence
for Precision Therapy

The same point-of-contact readout serves two ends. It phenotypes myocardial tissue for diagnosis, and it sharpens the therapy delivered at that exact spot, moving decisions beyond electrical voltage criteria alone.

Diagnostic value
Dx
Real-time tissue phenotyping
Intraprocedural characterization of tissue composition, cellular integrity and stiffness at the tip, beyond what voltage mapping can see.
Dx
Diffuse & heterogeneous disease
Detects cardiac fibrosis, non-ischemic cardiomyopathy and infiltrative cardiomyopathies; amyloidosis is one example within this class.
Dx
A complement to CMR & biopsy
A potential complement to cardiac MRI, and a way to characterize tissue without relying solely on endomyocardial biopsy, which is invasive and prone to sampling error.
Therapy guidance
Tx
Sharper substrate maps
In electrophysiology, augments substrate mapping to better distinguish pathological from normal myocardium.
Tx
More precise, tissue-sparing therapy
Designed to inform where energy is delivered, with the potential to improve ablation lesion targeting, reduce unnecessary tissue injury and support individualized procedural strategy.
Tx
The readout informs the action
Diagnosis and therapy at the same point of contact: the tissue readout directly guides the clinician-delivered therapy at that spot.

Status · CardioSpectra is a diagnostic adjunct that guides and enhances clinician-delivered therapy. It does not itself treat, and it does not yet integrate therapeutic capability or closed-loop guidance, a future direction (see roadmap). Clinical validation studies are warranted.

04 · Intelligence · FusionSuite

Three Signals, One Classification.
Unlocking a New Era of Precision

FusionSuite is the GPU-accelerated SaMD console that turns raw tri-modal data into an interpretable readout. A 1D convolutional network encodes the optical spectrum; dedicated encoders handle impedance and elastography; their features are fused late (preserving each modality's signal) into a probabilistic tissue classification with a calibrated uncertainty score. End to end, acquisition to display, in under 250 milliseconds.

01
Modality-specific encoders
A 1D CNN reads the optical spectrum; fully-connected encoders handle impedance and elastography. Each learns its own signal before anything is combined.
02
Late fusion, by design
Features are fused late rather than early, so the classifier stays robust even when one modality is degraded: poor contact, motion or noise.
03
Probabilistic output + calibrated uncertainty
Every contact returns class probabilities and a calibrated confidence score, in under 250 ms, acquisition to display. Uncertainty is always shown.
04
Locked & logged
Models are locked at runtime; version IDs, raw data, derived features and quality metrics are recorded per acquisition for reproducibility and audit.
05 · Safety & trust · Nexus

Confidence-Aware
by Design

Nexus is the trust layer around every acquisition. Sampling fires only when ECG is stable and contact is verified; low-quality records are excluded from inference and flagged for reacquisition. Hardware and software interlocks inhibit acquisition on unstable rhythm, poor contact or subsystem faults, while thermal sensors and current limiters cap energy delivery. If the AI or console fails, the system falls back to acquisition-only mode: raw data, no automated interpretation. Every record is encrypted, exportable and written to an immutable audit trail, explicitly aligned with Software-as-a-Medical-Device (SaMD) expectations.

Gated
Quality-based inference eligibility
Interlocked
ECG · contact · thermal · current
Fail-safe
Acquisition-only fallback on fault
Auditable
Encrypted, immutable trail · SaMD-aligned

Designed to support, not replace, clinician judgment. Raw data is always accessible, and uncertainty is always shown.

06 · Roadmap

From First Contact
to Continuous Insight

Forward-looking, and clearly labelled as such. The capabilities below are in development, not current features. Each depends on clinical validation studies that are warranted before clinical use, and on evidence that today is largely preclinical or limited-cohort.

→
Electroanatomic-map integration
Interoperability with existing electroanatomic mapping platforms, so tissue data overlays the maps operators already use.07
→
Closed-loop procedural guidance
Toward real-time, closed-loop guidance for catheter-based therapy, including patient-specific in-silico modelling of the procedure.08
→
Implantable-device integration
Miniaturization toward embedding sensing in pacing / defibrillator leads for continuous, longitudinal monitoring.
→
Longitudinal, patient-specific AI
Patient-specific models over time, with explainability and clinician-in-the-loop learning. Privacy-preserving, multi-site training is a longer-term direction.13
07 · Evidence base

Evidence-Based
by Design

CardioSpectra sits at the intersection of four mature research lines: fibrotic and infiltrative myocardial disease, multimodal intracardiac sensing, patient-specific cardiac modelling, and medical AI with privacy-preserving learning. Every architectural decision is anchored to peer-reviewed work below.

A · Fibrotic & infiltrative myocardial disease
  1. 01
    González-López E, Moñivas-Palomero V, Escobar-López L, et al. Wild-type transthyretin amyloidosis as a cause of heart failure with preserved ejection fraction. European Heart Journal · 2015 · 36(38):2585-94
  2. 02
    Maurer MS, Schwartz JH, Gundapaneni B, et al. Tafamidis treatment for patients with transthyretin amyloid cardiomyopathy. New England Journal of Medicine · 2018 · 379(11):1007-16
  3. 03
    Ruberg FL, Grogan M, Hanna M, Kelly JW, Maurer MS. Transthyretin amyloid cardiomyopathy: JACC state-of-the-art review. JACC · 2019 · 73(22):2872-91
B · Multimodal intracardiac sensing
  1. 04
    Singh-Moon RP, Marboe CC, Hendon CP. Near-infrared spectroscopy integrated catheter for characterization of myocardial tissues. Biomedical Optics Express · 2015 · 6(7):2494-2511
  2. 05
    Schwan HP. Electrical properties of tissue and cell suspensions. Advances in Biological and Medical Physics · 1957 · 5:147-209
  3. 06
    Couade M, Pernot M, Messas E, et al. In vivo quantitative mapping of myocardial stiffening and transmural anisotropy during the cardiac cycle. IEEE Trans. Medical Imaging · 2011 · 30(2):295-305
C · Patient-specific cardiac modelling & in-silico EP
  1. 07
    Boyle PM, Zghaib T, Zahid S, et al. Computationally guided personalized targeted ablation of persistent atrial fibrillation. Nature Biomedical Engineering · 2019 · 3:870-879
  2. 08
    Plank G, Loewe A, Neic A, et al. The openCARP simulation environment for cardiac electrophysiology. Computer Methods and Programs in Biomedicine · 2021 · 208:106223
  3. 09
    Rodero C, Strocchi M, Marciniak M, et al. Linking statistical shape models and simulated function in the healthy adult human heart. PLOS Computational Biology · 2021 · 17(4):e1008851
  4. 10
    Niederer SA, Lumens J, Trayanova NA. Computational models in cardiology. Nature Reviews Cardiology · 2019 · 16(2):100-111
D · Medical AI & privacy-preserving learning
  1. 11
    Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods · 2021 · 18(2):203-211
  2. 12
    Raissi M, Perdikaris P, Karniadakis GE. Physics-informed neural networks for solving forward and inverse problems involving nonlinear PDEs. Journal of Computational Physics · 2019 · 378:686-707
  3. 13
    Rieke N, Hancox J, Li W, et al. The future of digital health with federated learning. npj Digital Medicine · 2020 · 3:119
  4. 14
    Dayan I, Roth HR, Zhong A, et al. Federated learning for predicting clinical outcomes in patients with COVID-19. Nature Medicine · 2021 · 27:1735-1743
  5. 15
    Abadi M, Chu A, Goodfellow I, et al. Deep learning with differential privacy. Proc. ACM CCS · 2016 · 308-318
08 · Pipeline

From Bench to Bedside,
Candidly Sequenced

We work the way investors and the FDA both reward: software MVP and benchtop sensor rig progressing in parallel, integrating around month nine. Bench evidence first; alpha catheter second; pre-submission meeting before any biocompatibility spend.

Q4 2025
Provisional patent
Tri-modal catheter, FusionSuite and Nexus filed. Continuation-in-part drafted May 2026.
2026 · Now
Bench rig + software MVP
Three-modality benchtop sensor rig acquiring ex-vivo cardiac data; FusionSuite skeleton on public MRI cohorts.
Q4 2026
Integrated demo
Bench-acquired data assimilated into the FusionSuite classifier. Methods preprint, non-provisional filed.
2027
Alpha catheter · FDA Q-Sub
CDMO build of the integrated 8 Fr catheter. Pre-submission with the FDA. ISO 10993 begins.
2028 +
GLP preclinical · FIH
Porcine EP studies. De Novo / 510(k) preparation. First-in-human readiness.
09 · Founders

Built by Physicians
who Feel the Gap

CardioSpectra Lab was founded by Krtin Singhal, MD, and a team of practicing physicians who experience the point-of-contact gap CardioSpectra aims to close. The multi-specialty team is drawn from the Division of Cardiac Electrophysiology / Cardiology and Internal Medicine, practicing at Brookdale University Hospital Medical Center, Brooklyn, NY, and is actively expanding through engineering, ML and regulatory leadership.

Krtin Singhal, MD
Founder · Inventor · CEO

Cardiologist and inventor of the CardioSpectra platform. Filed the founding provisional in October 2025. Leads clinical, regulatory and platform strategy.

Adarsh Balaji, MD
Cardiology Fellow

NewYork-Presbyterian Brooklyn Methodist Hospital, Brooklyn, NY.

Vikaskumar Patel, MD
Internal Medicine Physician

Brookdale University Hospital Medical Center, Brooklyn, NY.

Nayan Doobay, MD
Internal Medicine Physician

Brookdale University Hospital Medical Center, Brooklyn, NY.

Konstantin Kecman, MD
Chief of Clinical Development

Owns the FDA pathway, pre-submission strategy and the design-history file. Bridges the engineering programme to ISO 13485, ISO 10993 and the GLP preclinical lab.

10 · Get in touch

Closing the loop
between
seeing and treating.

CardioSpectra is an investigational platform, currently raising. We respond personally to physicians, investors and prospective collaborators within two business days.

Direct line to Krtin Singhal, MD · Founder
For physicians
Briefing pack & investigator-site interest
For investors
Data room access & founder call
For partners
CDMO, hospital site & strategic collaboration
Press kit
Bios, brand assets, fact sheet