Mohsen Motie-Shirazi

A field-deployable fetal monitoring system built around a low-cost (~$15) handheld Doppler connected to a smartphone, where we have developed a full on-device pipeline that converts raw one-dimensional audio into clinically meaningful information in real time. The pipeline combines signal processing (filtering, normalization, artifact suppression, and signal-saturation detection) with deep learning models for fetal heart rate tracking, signal quality indices that guide probe positioning, and gestational-age estimation using hierarchical sequence modeling; a mobile computer-vision component uses real-time object detection and digit recognition to transcribe bedside blood-pressure displays into structured data. All inference runs inside an Android application with lightweight, quantized models to deliver sub-second, on-screen feedback that reduces unusable windows, stabilizes fetal heart rate and gestational-age estimates in noisy, real-world conditions, and packages results for seamless handoff to clinical workflows—bringing reliable point-of-care analytics to low-resource settings.

Built from the bench up, this program begins with physical vocal-fold models and controlled airflow, collecting direct intraglottal and subglottal pressures, acoustic recordings, and high-speed video under reproducible conditions. On top of these grounded measurements, we have developed physics-informed models—contact mechanics for collision pressure and finite-element inversion for tissue properties—to generate interpretable, pathology-relevant features such as spatiotemporal opening and closure patterns, symmetry and energy measures, and cumulative collision-pressure "dose." These features are then fed into machine learning models for classification, enabling differentiation of common lesions (for example, nodules, polyps, and posterior glottal insufficiency). The result is a coherent chain—physical modeling → multimodal measurement → physics-informed features → machine-learning classification—that supports diagnosis, therapy planning, and surgical decision-making.

Creation of comprehensive ECG image databases with real-world imaging and scanning artifacts to enable computerized ECG image digitization and analysis. This project provides foundational datasets for developing automated ECG analysis systems.

Development of advanced computational models including 2D finite element models for vocal fold material property inference, Bayesian inference frameworks, and numerical simulations for biomedical applications such as irrigant penetration in dentinal microtubules.