Projects at the Institute of Cancer Research (ICR)

Led Workflow language Training and Python programming Teaching at ICR

Mar 2018 - Feb 2020

Delivering Nextflow workflow language training: At the ICR, I led and co-ordinated a training programme to train coding researchers and scientific software developers (research software engineers) in the use of Workflow languages. Upon identifying ways in which reproducibility could be improved in computational work, I identified the need a training and established workshops for researchers and coding staff. Workflow languages are best-practice in scientific software development, particularly where HPC (High-Performance Computing) platforms are used. Workflow languages provide the means to ensure reproducibility in computational workflows and pipelines by providing interoperability, platform agnosticism (in case the distributed infrastructure changes), standardisation and reproducibility (using package management with software used pinned to specific versions).

  • Alnasir, J (2019). Fostering Reproducibility, Standardisation, Fault-tolerance and Deployability in Computational Pipelines using Nextflow Workflow Language - Adoption and Training at the ICR. Mathematical Foundations in Bioinformatics (MatBio '19), King's College London. [ MatBio '19 ]

DICOM medical image anonymiser using Deep-learning EAST text-detection algorithm

Mar 2018 - Feb 2020

I implemented and delivered a project that applies a pre-trained convoluted neural network to anonymise medical DICOM images, liaising with the project stakeholder to ascertain their requirements. These images are scans that contain patient identifiable information that has been burned into the pixels which needs removal to enable ethical use and sharing of the images for research. The project is fully automated and containerised deployment (implemented using Docker) provides easy invocation of the functionality and integration into other medical image management platforms (such as PAX-IT). Development of this type of algorithm is non-trivial, therefore an existing algorithm was tested — EAST: An Efficient and Accurate Scene Text Detector (Zhou et al) — and incorporated in the python implementation to handle the DICOM image format and obfuscation the patient identifiable information.

Copyright © 1999-2023 Dr. Jamie J. Alnasir, all rights reserved. Legal Disclaimer