iDetect — Non-Invasive COVID-19 Detection
Scalable COVID-19 detection using eye images and explainable AI — delivering 97% sensitivity, 350ms inference, and a strategic acquisition.
Why Existing Solutions Fell Short
Hospitals and public health organizations faced critical constraints during the pandemic that demanded a new approach.
Invasive, Slow Testing
Nasopharyngeal swabs caused patient discomfort and slow turnaround times, limiting testing capacity during surge periods.
Limited Throughput
Constrained testing capacity impacted public health responses and hospital patient flow during peak demand.
High Operational Costs
Dependency on supply chains for swabs, reagents, and specialized equipment drove up per-test costs.
Black-Box ML Models
Lack of explainability in existing AI solutions hindered clinical adoption and regulatory approval pathways.
These constraints directly affected healthcare operations
Reduced Screening
Limited testing capacity
Extended Stays
Longer hospital admissions
Lost Revenue
Missed diagnostic opportunities
Slow Adoption
Black-box AI distrust
From Discovery to Deployment
A product-focused, regulatory-aware engineering approach: rapid discovery, robust development, and clinical validation.
Discovery & Clinical Alignment
We started with deep stakeholder engagement to understand clinical requirements and define success metrics.
- Conducted interviews with clinicians, lab managers, and compliance officers
- Defined clinical endpoints: sensitivity, specificity, and latency targets
- Designed human-in-the-loop validation for investigational device requirements
Technical Architecture
Enterprise-grade ML infrastructure built for clinical deployment and regulatory compliance.
Data Ingestion
Secure upload, de-identification, and preprocessing pipeline for ocular images.
Hybrid Model
Custom CNN backbone + SVM ensemble, trained with augmentation and demographic balancing.
Explainability
LIME + Guided Backprop overlays surface features for clinician interpretation.
Serving
Dockerized model endpoints on Kubernetes, autoscaling based on request load.
Observability
MLflow + Prometheus + Grafana for model metrics, drift detection, and latency tracking.
Security
HIPAA-compliant Azure tenancy for PII, SOC 2 Type II controls in place.
AWS EC2
Azure
Python
PyTorch
MLflow
Docker
Kubernetes
CNN
SageMaker
MilvusAWS EC2AzurePythonPyTorchMLflowDockerKubernetesCNNSageMakerMilvusQuantifiable Results
Real metrics from clinical deployment and commercial validation.
Before → After Transformation
Faster Triage
Reduced lab dependence and faster patient routing
Lower Costs
Reduced operational testing expenses
Patient Comfort
Non-invasive testing improves experience
I used to spend 2 hours a night charting. Now I'm home for dinner. This isn't just software; it's a lifestyle change.
See Patient Impact in 30 Minutes
Book a personalized demo to see how CodefyU can build similar solutions for your healthcare organization.