RatioLogo
Back

The Cardiac Health Revolution

A groundbreaking new paradigm in healthcare is emerging. It moves beyond simple fitness tracking, using sophisticated algorithms to transform raw cardiac data into precise, early-stage medical diagnoses.

The Controlled Testing Ritual

Setting: A controlled room held between 23°C and 29°C.
Activity: Subjects undergo a precise 20-minute treadmill ritual.

  • First 10 Minutes: Walking at 2 km/h.
  • Second 10 Minutes: Walking at 4 km/h.
    Purpose: This isn't marathon training. It's about generating the clean, raw electrical signatures of the human heart for an algorithm that can detect illness before the first symptom appears.

The Diagnostic Leap

The traditional smartwatch is excellent at counting steps, but remains largely illiterate when translating heart rate variability (HRV) into a medical diagnosis.

New research from the Institut Teknologi Bandung aims to change that. By using a computational framework, their system classifies "stress/depression" and "influenza" states with startling precision.

This signals a critical transition:

  • Past: Wearables as passive data logs.
  • Future: Wearables as active, early-warning diagnostic tools.

The Machine Learning Engine

Support Vector Machines (SVM)

The study applied this type of machine learning to map complex, non-linear biological data into a 3D space. This allows the algorithm to draw a clear, calculable line between "healthy" and "sick" states.

The Results: Accuracy Breakdown

The study yielded impressive, yet distinct, accuracy rates for two different conditions.

Detecting Stress & Depression

  • Key Metrics: Standard Deviation of Heart Rate (SDHR) and RR Intervals (SDIBI).
  • Accuracy: 95% (correctly classifying 19 out of 20 samples).
  • Signature: The model clearly separates the two states.
    • Stressed Subjects: Outputs ranged from 0.784 to 1.660.
    • Healthy Subjects: Outputs plummeted into the negative range, as low as -2.343.

Detecting Influenza

  • Key Metrics: Mean Heart Rate and Mean Inter-Beat Intervals.
  • Accuracy: 85% (correctly identifying 17 out of 20 cases).
  • Challenge: Influenza's early-stage signature often overlaps with healthy states, leading to "data crossing" that can confuse the software.

The Sensitivity of the Heart

The data powerfully highlighted how our cardiac rhythms are a sensitive barometer for our internal state and environment.

Emotional & Physiological Impact

  • Emotions: Anger and fear can spike the heart rate by approximately 8 bpm.
  • Dehydration: Even a modest 1% loss in body weight can drive the heart rate up by 7 bpm.

Current Limitations & The Path Forward

Despite high accuracy rates, this technology is still in its refining stages. Key limitations were identified:

Study Constraints

  • Sample Size: Relied on a small validation set of 20 samples per condition.
  • Ground Truth: Illness classification was based on subjective self-reports, not confirmed by blood tests or viral loads.
  • Missed Cases: Two influenza cases were incorrectly categorized as healthy, despite symptomatic subjects.

The Future Requirement

While the 2D-to-3D mappings provide a sophisticated "Shesop" application framework, the authors conclude that even higher accuracy will require larger volumes of training data. This will help the machine distinguish between a heart that is truly ill and one that is simply tired.

Reference: Wijaya, A. I., Prihatmanto, A. S., & Wijaya, R. (2016). Shesop Healthcare: Stress and Influenza Classification Using Support Vector Machine Kernel. School of Electrical Engineering and Informatics, Institut Teknologi Bandung.