A novel Hot-Flash classification algorithm via multi-sensor features integration.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Menopause Research

Background:

• Hot flashes (HFs) are a common symptom of menopause, significantly impacting quality of life.
• Accurate and objective measurement of HFs is crucial for research and treatment.
• Current detection methods often rely on subjective reporting or limited sensor data.

Purpose of the Study:

• To evaluate a novel algorithm for hot flash (HF) classification.
• To assess the feasibility and performance of integrating multisensor features from commercial wearable devices.
• To compare the multi-sensor approach against single-sensor methods and expert evaluation.

Main Methods:

• Feature sets were processed from wrist-based multisensor data, including photoplethysmography, motion, temperature, and skin conductance.
• A Decision Tree algorithm was used to classify physiological-recorded HFs (N=27) from three menopausal women.
• Algorithm performance was assessed against gold-standard HF expert evaluation.

Main Results:

• Skin conductance features alone explained approximately 65% of the variance in HF classification.
• The multi-sensor approach achieved over 90% sensitivity and 95.6% specificity in HF classification.
• The algorithm demonstrated advantages under signal corruption and varying biobehavioral states (sleep vs. wake).

Conclusions:

• The proposed multi-sensor approach shows promise for accurate hot flash classification using commercially available wearable sensors.
• User-centered, automatic detection systems for HFs can advance their measurement and treatment.
• This technology offers a potential for improved monitoring and management of menopausal symptoms.