Cardiovascular signal processing & deep learning
I investigate cardiac timing and wearable sensing, and help research and engineering teams turn recordings into reproducible analysis.
Open to PhD opportunities in cardiovascular digital health · Taking signal-processing projects
I’m a cardiovascular signal-processing engineer and researcher. One question connects my work: how the heart’s electrical activity relates to its mechanical and acoustic consequences, and how that relationship can be read from ECG, heart-sound (PCG) and optical-pulse (PPG) recordings.
It began with paediatric heart-sound collection in hospitals and schools in Pakistan, where I learned what cardiologists need from a recording. Since then I have worked on deep-learning models for cross-modal ECG–PCG reconstruction and a wearable research prototype for time-aligned ECG, PPG and motion. I am also developing a probabilistic model of coupled cardiac events.
Alongside research I take on audio and biomedical signal-processing projects — spectral analysis, comparison reports, classifiers and analysis tools. For my PhD, I want to investigate whether signal-derived cardiac timing can constrain patient-specific physiological models, with uncertainty and device differences treated explicitly. My current modelling is statistical and data-driven; physiological modelling and prospective validation are skills I want to develop during a PhD. My doctoral question.
MS thesis · NUST
A conditional GAN with multi-scale attention reconstructs an ECG-like timing reference from heart sounds; phase-aware features then target murmur presence, grade, timing and shape.
Current finding · Oct 2026 In a patient-grouped evaluation, expert-reviewed phase boundaries carried useful information; the tested synthetic-ECG timing route did not establish an advantage over equal-time windows.
Read the studyCAREON · NUST
An ESP32 research prototype with AD8232 ECG, MAX30102 PPG and MPU6050 IMU sensors. I set 250 Hz ECG and 100 Hz PPG sampling targets and specified a 16-bit external ADC fallback for acquisition precision. Timing fidelity and motion sensitivity remain to be evaluated.
The ECG-to-pulse delay includes pre-ejection time, so it is not treated as isolated pulse transit time.
Project detailsQatar University · in development
Copula-based coupled hidden Markov models: one chain for ECG waves (P, QRS, T), one for heart-sound phases (S1, systole, S2, diastole), and a copula for their non-linear dependence.
ApproachPublished · ICAI 2023 · EUSIPCO 2024
ICAI 2023: a time-domain 1D CNN with pitch-shift augmentation for binary paediatric congenital-heart-disease detection. EUSIPCO 2024: band-selective pitch augmentation and MFCC features for a separate five-class valvular/normal heart-sound benchmark.
Both papers
I help researchers, engineers and product teams get reliable answers from messy real-world signals — with clear methods, reproducible outputs and stated limits.
Hire me on FiverrDigital spectra, level, noise-floor and channel comparisons across recording conditions, with a structured report and stated measurement limits.
Filtering, segmentation, feature extraction and recording investigation for PCG, ECG and difficult audio.
ECG, heart-sound and acoustic classification projects, with data, evaluation and deliverables agreed before development.
Sensor integration and signal pipelines for wearable ECG and PPG prototypes.
“He keeps timelines and even for a not so tech person made me understand the steps and the processes.”
“As always Zafar was receptive, involved and exceeded all expectations.”
“Very Professional. Delivered on time. Highly recommend.”
2023 — 2026
Thesis: Cross-Modal Reconstruction of Cardiac Signals: Decoding ECG-PCG Electromechanical Coupling Dynamics. Supervisor Dr. Nazia Perwaiz; co-supervisor Dr. Muhammad Salman Khan. CGPA 3.20/4.00, thesis grade A.
2018 — 2022
Final-year project on paediatric congenital-heart-disease detection from heart sounds (published at ICAI 2023). CGPA 3.21/4.00, project grade A.
Journal statuses below were last reported on 9 October 2026. These manuscripts are not accepted or published; conference papers and congress presentations are listed separately.
Phonocardiogram Classification Based on 1D CNN with Pitch-Shifting and Signal Uniformity Techniques
Automatic Detection of Paediatric Congenital Heart Diseases from Phonocardiogram Signals
PCG-to-ECG Signal Reconstruction Using Conditional GAN with Multi-Scale Attention
Synthetic-ECG-gated phonocardiography for paediatric heart-murmur characterisation: a cross-modal deep-learning method
Parallel multi-scale analysis for paediatric heart disease classification from phonocardiograms
Synthetic ECG-guided temporal gating improves paediatric murmur grading, timing, and shape classification on the CirCor DigiScope dataset
An explainable calibrated concept bottleneck model with evidence-constrained large language model reporting for multi-label valvular heart disease screening from 12-lead ECGs
Grant proposals co-written (submitted, not awarded): HEC Pakistan National Research Programme for Universities, and Qatar University Office of Research Support, both September 2025.
// next step
PhD supervisors, research collaborators and engineering clients are all welcome. A short description of the question or the recordings is the best start.
