Liam Chalcroft

Liam Chalcroft

PhD, University College London
London · Email

I'm a machine learning researcher working on medical and spectral imaging. I'm currently the founding computer vision scientist at Prospectral, where I lead the machine learning work.

Before that I did a PhD at University College London on stroke lesion segmentation in clinical MRI, mainly on making segmentation models robust to differences between scanners and sequences.

Medical image generation

Tokenizer-Generator Coupling in Medical Image Generation

NeurIPS 2026, poster · Liam Chalcroft

A controlled comparison of tokenizers, generators and sampler settings for low-resolution medical-style images.

The best quantizer changes with the generator. Sampler selection also changes the apparent ranking, and reconstruction quality measured by PSNR is not a reliable selection criterion on its own.

Unconditional 64 × 64 images, evaluated with non-clinical generation metrics.

Read the studyExperimentsmedtokenizersmedlatents

Original two-panel tokenizer-generator comparison: seed variation and FID measurements across generator families.
Original study figure. The vocabulary-1024 block compares three training seeds; the rest of the grid has one. Source and study scope · Full-size figure

Vision-language evaluation

GAZE

Duaa Alim, Mogtaba Alim and Liam Chalcroft
AIiH 2026, oral paper

Viewer-level tools and literature retrieval let a VLM inspect rare brain MRI iteratively. The evaluation considers diagnosis, localisation and captioning together.

Retrieval effects depend on the model: gains in diagnosis can coincide with losses in localisation. The paper examines these trade-offs alongside structured prompting and tool use.

Read the papergazeRelated RCR abstract

Tools and evaluation in GAZE

  • Viewer tools

    The model can request viewer-level operations, including zoom, windowing, contrast and edge detection.

  • Retrieval tools

    Retrieval tools query PubMed literature and Open-i radiological images.

  • Joint evaluation

    Structured outputs are scored jointly for diagnosis, localisation and captioning.

Method overview from the public workflow documentation and paper.

Open-source software

Libraries I maintain and the implementations behind the research.

All code and contributions

At Prospectral

I lead the machine learning work at Prospectral.

Gradient-manifold alignment scheduling for physics-guided diffusion

SPIE Photonics Europe 2026

Paper and materialsOral presentation

Earlier research

My PhD at UCL concerned stroke lesion segmentation in clinical MRI, with a focus on robustness across scanners and sequences.

ThesisInvited yESMInar talkAll publications

  1. Domain-Agnostic Stroke Lesion Segmentation Using Physics-Constrained Synthetic Data

    L. Chalcroft, J. Crinion, C.J. Price, J. Ashburner.

    MICCAI 2025.

    paperarXivcode

    2025

  2. Synthetic Data for Robust Stroke Segmentation

    L. Chalcroft, I. Pappas, C.J. Price, J. Ashburner.

    MELBA 2025.

    paperarXivcode

    2025

  3. DeepISLES: a clinically validated ischemic stroke segmentation model from the ISLES'22 challenge

    E. de la Rosa, M. Reyes, S.-L. Liew, et al. Including L. Chalcroft.

    Nature Comms 2025.

    papercode

    2025

Recent work

Contact

liamchalcroft@gmail.com