Liam Chalcroft | Publications

GAZE: Grounded Agentic Zero-shot Evaluation with Viewer-Level Tools and Literature Retrieval on Rare Brain MRI

2026-08-26T00:00:00+01:00

Vision-language models (VLMs) read an image and produce text in a single forward pass, whereas radiologists typically inspect an image several times and consult the literature before writing a report. We introduce GAZE (Grounded Agentic Zero-shot Evaluation), a framework that lets a medical VLM work in this iterative way by calling viewer-level tools (zoom, windowing, contrast, edge detection) and two retrieval tools backed by the U.S. National Library of Medicine (PubMed for medical literature, Open-i for radiological images), with structured outputs validated against a schema and full tool-call traces recorded for auditability. On NOVA, a benchmark of 906 brain MRI cases covering 281 rare neurological conditions, GAZE reaches 58.2 mean average precision (mAP) at intersection-over-union (IoU) 0.3 for lesion localisation and 34.9% Top-1 diagnostic accuracy under a joint protocol that scores captioning, diagnosis, and localisation from the image alone, without task-specific fine-tuning. Before any tool is used, structured prompting and schema-validated outputs already improve over the published Gemini 2.0 Flash baseline (20.2 to 29.4 mAP@0.3), so framework design is itself an experimental variable. Tool use helps rare pathologies disproportionately: the fraction of cases with IoU > 0.3 rises from 17% to 58% for diagnoses with three or fewer examples versus 25% to 68% for common conditions (at least 10 cases), with gains tracking engagement. Retrieval ablations additionally reveal a model-dependent trade-off in which gains in diagnosis can coincide with losses in localisation, reinforcing the case for joint evaluation of diagnosis, localisation, and captioning in medical VLMs.

Gradient-manifold alignment scheduling for physics-guided diffusion

2026-05-27T00:00:00+01:00

Physics-guided diffusion models combine learned priors over valid designs with physics-based objectives for inverse design. A key challenge is scheduling the guidance strength across the iterative generation process. We propose alignment scheduling, a sample-dependent guidance rule based on the relative magnitude of the projected physics gradient and the raw physics gradient. Under the approximate manifold-projection view of diffusion denoisers, this ratio indicates how much of the physics update lies in directions the model can follow while staying near realistic designs. The method uses quantities already computed during sampling and adds minimal overhead. We evaluate on two photonic inverse design tasks, a colour router for CMOS image sensors and a waveguide bend, using differentiable FDTD simulation. In an exploratory study across three fabrication classes and two guidance strengths, alignment achieves the strongest colour-router performance at the higher setting, while constant guidance remains strongest on the waveguide task. These mixed results suggest alignment scheduling is most beneficial at higher guidance strength for the colour-router task and that task-dependent tuning remains important.

Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning

2025-09-21T00:00:00+01:00

Self-supervised deep learning has accelerated 2D natural image analysis but remains difficult to translate into 3D MRI, where data are scarce and pre-trained 2D backbones cannot capture volumetric context. We present a sequence-invariant self-supervised framework leveraging quantitative MRI (qMRI). By simulating multiple MRI contrasts from a single 3D qMRI scan and enforcing consistent representations across these contrasts, we learn anatomy-centric rather than sequence-specific features. The result is a single 3D encoder that excels across tasks and protocols. Experiments on healthy brain segmentation (IXI), stroke lesion segmentation (ARC), and MRI denoising show significant gains over baseline SSL approaches, especially in low-data settings (up to +8.3% Dice, +4.2 dB PSNR). It also generalises to unseen sites, supporting scalable clinical use.

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

2025-09-19T00:00:00+01:00

Segmenting stroke lesions in MRI is challenging due to diverse acquisition protocols that limit model generalisability. In this work, we introduce two physics-constrained approaches to generate synthetic quantitative MRI (qMRI) images that improve segmentation robustness across heterogeneous domains. Our first method, qATLAS, trains a neural network to estimate qMRI maps from standard MPRAGE images, enabling the simulation of varied MRI sequences with realistic tissue contrasts. The second method, qSynth, synthesises qMRI maps directly from tissue labels using label-conditioned Gaussian mixture models, ensuring physical plausibility. Extensive experiments on multiple out-of-domain datasets show that both methods outperform a baseline UNet, with qSynth notably surpassing previous synthetic data approaches. These results highlight the promise of integrating MRI physics into synthetic data generation for robust, generalisable stroke lesion segmentation.

DeepISLES: a clinically validated ischemic stroke segmentation model from the ISLES’22 challenge

2025-08-09T00:00:00+01:00

Diffusion-weighted MRI (DWI) is essential for stroke diagnosis, treatment decisions, and prognosis. However, image and disease variability hinder the development of generalizable AI algorithms with clinical value. We address this gap by presenting a novel ensemble algorithm derived from the 2022 Ischemic Stroke Lesion Segmentation (ISLES) challenge. Our ensemble model combines the strengths of top-performing algorithms and achieved superior ischemic lesion detection and segmentation accuracy (median Dice score: 0.82, median lesion-wise F1 score: 0.86) on our internal test set compared to individual algorithms. This accuracy generalized well across diverse image and disease variables. Validation using a real-world external dataset (N=1686) confirmed the model’s generalizability. The algorithm’s outputs also demonstrated strong correlations with clinical scores (admission NIHSS and 90-day mRS) on par with or exceeding expert-derived results, underlining its clinical relevance. Notably, in a Turing-like test, neuroradiologists consistently preferred the algorithm’s segmentations over manual expert efforts, highlighting increased comprehensiveness and precision.

Synthetic Data for Robust Stroke Segmentation

2025-01-01T00:00:00+00:00

Deep learning-based semantic segmentation in neuroimaging currently requires high-resolution scans and extensive annotated datasets, posing significant barriers to clinical applicability. We present a novel synthetic framework for the task of lesion segmentation, extending the capabilities of the established SynthSeg approach to accommodate large heterogeneous pathologies with lesion-specific augmentation strategies. Our method trains deep learning models, demonstrated here with the UNet architecture, using label maps derived from healthy and stroke datasets, facilitating the segmentation of both healthy tissue and pathological lesions without sequence-specific training data. Evaluated against in-domain and out-of-domain (OOD) datasets, our framework demonstrates robust performance, rivaling current methods within the training domain and significantly outperforming them on OOD data. This contribution holds promise for advancing medical imaging analysis in clinical settings, especially for stroke pathology, by enabling reliable segmentation across varied imaging sequences with reduced dependency on large annotated corpora.

LKA: Large-kernel Attention for Efficient and Robust Brain Lesion Segmentation

2023-12-01T00:00:00+00:00

Vision transformers are effective deep learning models for vision tasks, including medical image segmentation. However, they lack efficiency and translational invariance, unlike convolutional neural networks (CNNs). To model long-range interactions in 3D brain lesion segmentation, we propose an all-convolutional transformer block variant of the U-Net architecture. We demonstrate that our model provides the greatest compromise in three factors: performance competitive with the state-of-the-art; parameter efficiency of a CNN; and the favourable inductive biases of a transformer.

Development and evaluation of intraoperative ultrasound segmentation with negative image frames and multiple observer labels

2021-09-01T00:00:00+01:00

When developing deep neural networks for segmenting intraoperative ultrasound images, several practical issues are encountered frequently, such as the presence of ultrasound frames that do not contain regions of interest and the high variance in ground-truth labels. In this study, we evaluate the utility of a pre-screening classification network prior to the segmentation network. Experimental results demonstrate that such a classifier, minimising frame classification errors, was able to directly impact the number of false positive and false negative frames. Importantly, the segmentation accuracy on the classifier-selected frames, that would be segmented, remains comparable to or better than those from standalone segmentation networks. Interestingly, the efficacy of the pre-screening classifier was affected by the sampling methods for training labels from multiple observers, a seemingly independent problem. We show experimentally that a previously proposed approach, combining random sampling and consensus labels, may need to be adapted to perform well in our application. Furthermore, this work aims to share practical experience in developing a machine learning application that assists highly variable interventional imaging for prostate cancer patients, to present robust and reproducible open-source implementations, and to report a set of comprehensive results and analysis comparing these practical, yet important, options in a real-world clinical application.