CardioSense
Cardiology · Diagnostic AI · v1.0

A second opinion
on the heart, in
two hundred milliseconds.

CardioSense fuses eleven clinical signals through a gradient-boosted classifier trained on roughly seventy-thousand cardiovascular records — returning a calibrated probability of disease and a transparent, factor-by-factor breakdown.

Sample readout · MRN-1024
72%
High
Patient over fifty,
sedentary, hypertensive.
SYS 156 · DIA 92 · BMI 31.4
Hypertension
Obesity
Sedentary
≈ 70K
Training samples
11
Clinical inputs
< 200ms
Inference time
0–100%
Calibrated output
What you get

Quantified risk, backed by data.

Every diagnosis returns three artefacts — a number, a story, and a record. Together they support a clinical conversation rather than replace one.

01

Calibrated probability

A continuous probability score — not a binary verdict — so you can triage rather than guess.

02

Vitals breakdown

Each input scored against clinical thresholds: BP, BMI, cholesterol, glucose, lifestyle.

03

Private patient log

Every diagnosis is timestamped and saved to your account — searchable and exportable.

Under the hood

Boosted trees on tabular reality.

CardioSense is built on XGBoost — a gradient-boosted tree ensemble that excels at the structured, mixed-type tabular data that defines clinical screening. The underlying corpus combines the public Cardiovascular Disease Dataset with Framingham-derived features.

Inference is delegated to a dedicated Flask service on Railway, called from a type-safe TanStack server function. The score is returned, displayed, and persisted to a row-level-secured patient record.

Full methodology
Ready when you are

Eleven inputs. One readout.

⚕ For screening and educational use only — never replaces clinical judgement