Projects

Projects and focused work

This page brings together standalone projects and focused methods from larger experiences. Each card makes its original setting clear.

Project work

Focused project work

The first three cards are independent technical projects. The remaining cards identify the experience that gave each deliverable or method its context.

Computational imaging methods

Liver MRI Methods

A reproducible multiphasic MRI study covering image geometry, registration inspection, dynamic-contrast features and a model failure that now guides the next debugging work.

  • Multiphasic MRI
  • Registration
  • Segmentation debugging
Bioinformatics workflow

Reproducible RNA-seq Service Workflow

A Nextflow DSL2 service exercise taking a paired public RNA-seq question from validated intake through DESeq2, checked results and a provenance-rich handover.

  • Nextflow DSL2
  • Paired DESeq2
  • Reproducible handover
Self-directed project

Clinical informatics trial operations

A synthetic study workflow connecting protocol requirements, relational data checks, operational metrics and an interactive dashboard.

  • Synthetic study data
  • Relational validation
  • Streamlit dashboard
Final-year project method

Exploratory modelling approach

Cohort construction, ICD/HES-derived feature engineering, model comparison and a staged mutual-information continuation that stopped when the data did not support further searching.

  • Final-year project context
  • Feature pipeline
  • Stopping rules
Health Innovation East deliverable

AI Toolkit

Internal guidance-support material translating AI/MedTech governance, evidence expectations, clinical safety and data requirements into advisory questions.

  • HIE placement context
  • AI/MedTech governance
  • Toolkit authoring
Health Innovation East method

Market intelligence and competitor analysis

Structured market, competitor and horizon-scanning work for live innovation questions, shaped around evidence quality and pathway fit.

  • HIE placement context
  • CB Insights
  • Decision support
Internship method

Computational toxicology data preparation

Dose-toxicity data mapping and structured preparation work from the ConsoneAI/DioScor research internship.

  • Internship context
  • Data mapping
  • Toxicology context

Exploratory methods

Future biological evidence methods

Exploratory methods focused on weak-signal recovery. This strand grows out of the final-year project's rare-disease limits: biologically informed, provenance-aware methods for hypothesis generation, study-design support and measurement prioritisation when the data is sparse and the next evidence decision matters.

Concept-stage methods thinking, not a validated platform or clinical tool.

Weak-signal methodsFind a usable signal without overstating the data.
BiologypriorSparse dataCandidate signal
  1. RecoverUseful signal
  2. PrioritiseNext measurement
  • Recovering useful signal without pretending sparse data is richer than it is.
  • Using biological context to prioritise what to measure next.
  • Generating sharper hypotheses and follow-up study designs.
  • Stress-testing assumptions with negative controls, ablations and disproof-first checks.
  • Keeping observed evidence, derived features and augmented assumptions clearly separate.

Page updated