Software & Data

Dataset records, research implementations, and reproducible workflows.

Datasets

What each release contains, how to access it, and my role in the study.

2026ZenodoAccess by request

Multi-Instrument SERS Dataset

  • 29,545 spectra
  • 20 analyte classes
  • 5 instruments
  • Version 1.0.0

Surface-enhanced Raman spectra from four portable instruments and one laboratory reference. The recordings support research on spectral alignment and analyte classification across devices. See the Zenodo record to request the files.

License
CC BY 4.0
Files
The public record gives instrument counts, preprocessing details and the citation. Use the Zenodo access request to obtain the spectra.
Evaluation
The classification experiment used 50 training spectra per analyte class. The study describes the device-to-reference comparison and evaluation protocol. Study methods and Table 5

My contribution I published the Zenodo record and led model development and evaluation for the accompanying cross-device study.

2022KagglePublic dataset

COVID-QU-Ex Dataset

  • 33,920 chest radiographs
  • 11,956 COVID-19 images
  • Lung segmentation masks

Chest X-rays from COVID-19, other infections, and normal cases. The companion study used lung masks to outline the lungs and a separately annotated subset to localise infection and grade severity.

Files
Kaggle v7 provides chest X-rays with lung masks. The infection subset pairs COVID-positive images with masks from QaTa-Cov19 for infection localisation.
Evaluation
The paper reports five folds with an 80/20 training and test split. One fifth of the training data was reserved for validation in each fold. Experimental setup and Table 1

My contribution I co-curated the chest X-ray benchmark and contributed to model evaluation, interpretation, and manuscript review.

Dataset citation credits Anas M. Tahir, Muhammad E. H. Chowdhury, Yazan Qiblawey, Amith Khandakar, Tawsifur Rahman, Serkan Kiranyaz, Uzair Khurshid, Nabil Ibtehaz, Sakib Mahmud and Maymouna Ezeddin.

2022KagglePublic dataset

Aseptic Loose Hip Implant X-Ray Database

  • 200 images for implant localisation
  • 206 images for loosening classification

Hip implant radiographs for two tasks in the HipXNet study. One subset has expert annotations for locating the implant. The other supports classification of aseptic loosening.

License
CC0 1.0
Files
The study describes expert-annotated implant radiographs for localisation and radiographs labelled fixed or loose for classification.
Evaluation
The paper describes five-fold evaluation with an 80/20 image split for training and testing. HipXNet methods

My contribution I helped develop and evaluate the HipXNet pipeline and contributed to data analysis, interpretation, and writing.

The dataset DOI lists Tawsifur Rahman, Muhammad E. H. Chowdhury, Amith Khandakar, and Yazan Qiblawey as creators. The Kaggle description also credits the wider study team, including me.

Research software

Reusable layers and implementations accompanying my papers.

Reusable libraryVersioned library

SuperONN-Layers

Polynomial 1D and 2D super-operational layers for PyTorch and TensorFlow/Keras, including forward and transposed operations with fixed or trainable offsets.

The release provides the library and its examples. Setup depends on whether you use the PyTorch or TensorFlow backend.

Wearable biomechanicsVersioned code and verified CPU demo

Reconstructing 3D kinetics from smart-insole loading

Estimating 3D forces, moments, and centre of pressure from the vertical loads measured by smart insoles.

The synthetic insole workflow completed on CPU with Python 3.12. The study workflow needs prepared insole or motion recordings, and the article checkpoints are separate from the release.

Medical imagingStudy source snapshot

Kidney stone segmentation in CT

Finding the CT slices most likely to contain kidney stones before tracing their boundaries.

The workflow expects prepared CT slices and stone masks. The public repository provides the study scripts and notebooks. Reproducing the reported Dice scores would require the study inputs and evaluation setup.

Physiological signal processing

Cuffless blood pressure and arterial waveform estimation from PPG and ECG

Deep learning methods that estimate systolic and diastolic pressure and reconstruct continuous arterial blood pressure waveforms from PPG and ECG signals.

Medical imagingStudy source snapshot

Grading prenatal hydronephrosis from ultrasound

Combining multiple fetal kidney ultrasound views into one severity grade, with the kidney outlined before classification.

The workflow uses prepared fetal kidney ultrasound views for region detection, segmentation, and severity grading. Full reproduction requires the study inputs and evaluation setup.

Wearable biomechanics

Machine learning for gait analysis: Tai Chi training and neurodegenerative disease detection

Using foot-sensor recordings and machine learning to separate Tai Chi masters from trainees, then to detect ALS, Huntington’s disease and Parkinson’s disease from walking patterns.

Biomedical devices & simulation

Rapid wall shear stress prediction for abdominal aortic aneurysms

A deep learning surrogate that predicts time-averaged wall shear stress on abdominal aortic aneurysms from 3D vessel geometry.

Intelligent energy systems

Non-intrusive load and occupancy monitoring from smart-meter data

Recovering appliance-level electricity use from a single household meter, then using those appliance events to infer when a home is occupied.

Applied machine learning

Satellite image segmentation for climate and geopolitical risk

A multiclass segmentation pipeline that maps buildings, vegetation, water and other classes from satellite images, built for a team project on risks to water and energy financing in West and Central Africa.

Applied machine learning

Machine learning analysis of COVID-19 vaccine attitudes in Qatar

A survey of 175 residents of Qatar, analysed with statistical tests, machine learning and dominance analysis to find what drives willingness and hesitancy toward COVID-19 vaccination.