Institution profile

Institute of Cancer Research

Academic institutioneurope · gb
Official website
Research library5linked papers
Opportunities0open roles
Selected work

Representative Papers

Morphologically Intelligent Perturbation Prediction with FORM

Oct 24, 2025

Current computational models are largely restricted to two-dimensional cell morphology modeling, failing to capture realistic three-dimensional structural dynamics under perturbations—thereby limiting the predictive accuracy and functional interpretability of virtual cells. To address this, we propose the first 3D cell morphology prediction framework supporting both unconditional generation and conditional simulation. Our method employs a multi-channel VQGAN-based high-fidelity morphological encoder coupled with a diffusion model to explicitly model perturbation-induced morphological trajectories, trained on over 65,000 3D multi-fluorescence cellular volume images. The framework accurately predicts 3D morphological evolution under chemical or genetic interventions, infers signaling pathway activity, disentangles combinatorial perturbation effects, and generalizes to unseen perturbation types. We further release MorphoEval, a standardized benchmarking suite for quantitative evaluation, establishing a new paradigm and reproducible evaluation standard for 3D virtual cell modeling.

0 citationsRead paper

Generalizing imaging biomarker repeatability studies using Bayesian inference: Applications in detecting heterogeneous treatment response in whole-body diffusion-weighted MRI of metastatic prostate cancer

May 14, 2025

Current imaging biomarkers lack a robust, quantitative framework for assessing heterogeneous treatment responses across multiple metastatic lesions in prostate cancer. Method: We propose the first general Bayesian repeatability assessment framework specifically designed for imaging biomarkers—overcoming limitations of traditional normality assumptions by enabling posterior inference for heterogeneous parameters under complex distributions (e.g., Dirichlet-Multinomial), accommodating multimodal imaging (e.g., whole-body diffusion-weighted MRI) and non-Gaussian biomarkers. Efficient Bayesian inference is implemented via Hamiltonian Monte Carlo, integrated with whole-body DWI volumetric analysis and validated on an mCRPC clinical cohort. Results: In two independent studies, the framework quantitatively identified significantly differential treatment responses in ~70% of metastatic lesions, demonstrating clinical validity and substantially improving the accuracy, robustness, and interpretability of heterogeneity quantification.

0 citationsRead paper

Signal-based AI-driven software solution for automated quantification of metastatic bone disease and treatment response assessment using Whole-Body Diffusion-Weighted MRI (WB-DWI) biomarkers in Advanced Prostate Cancer

May 13, 2025

This study addresses the challenge of quantitative assessment of bone metastatic burden and treatment response in advanced prostate cancer. Methodologically, we propose a fully automated AI framework based on whole-body diffusion-weighted MRI (WB-DWI): (1) a weakly supervised Residual U-Net generates skeletal probability maps to guide lesion detection; (2) a WB-DWI intensity statistical normalization strategy is introduced; and (3) a lightweight CNN enables end-to-end lesion segmentation, followed by registration with gADC maps to extract tumor diffusion volume (TDV) and median gADC—key quantitative biomarkers. Our key contribution is enabling bone metastasis quantification without per-lesion annotation. Validation demonstrates skeletal segmentation Dice scores of 0.6 (pelvis/spine), coefficient of variation (CV) of 4.6% for log-TDV and 3.6% for median gADC, and treatment response classification accuracy of 80.5%, sensitivity of 84.3%, and specificity of 85.7%. Processing time per case is 90 seconds.

0 citationsRead paper

Image deidentification in the XNAT ecosystem: use cases and solutions

Apr 29, 2025

To address DICOM image privacy leakage risks in the XNAT platform, this work proposes an automated de-identification workflow tailored for research data management. Methodologically, it integrates XNAT’s native API, standardized DICOM parsing, and a configurable rule engine to establish, for the first time, a systematic, multi-scenario-adaptive technical pathway within the XNAT ecosystem. It innovatively incorporates a BERT-NER model fine-tuned for address recognition to enhance anonymization of unstructured textual fields and evaluates performance using the MIDI-B benchmark framework. On the MIDI-B Challenge test set, the workflow achieves an overall de-identification accuracy of 99.61% and a real false-negative rate of only 0.19%, representing a 1.7-percentage-point improvement over the baseline. This work fills a critical gap in systematic, production-ready privacy protection for medical imaging data in XNAT environments and delivers a reusable, empirically validated technical paradigm for compliant research data governance.

0 citationsRead paper

A weakly-supervised deep learning model for fast localisation and delineation of the skeleton, internal organs, and spinal canal on Whole-Body Diffusion-Weighted MRI (WB-DWI)

Mar 26, 2025

Clinical whole-body diffusion-weighted imaging (WB-DWI) requires precise anatomical segmentation for accurate ADC quantification and tumor volume (TDV) measurement. However, manual delineation of the entire skeleton, visceral organs (liver, spleen, kidneys, bladder), and spinal canal is prohibitively time-consuming and clinically infeasible. To address this, we propose the first weakly supervised segmentation framework tailored for WB-DWI: a soft-label–guided 3D residual U-Net enabling simultaneous probabilistic segmentation of multiple structures without voxel-level annotations. Leveraging multi-center data, patch-based training, and probabilistic output maps, our method achieves clinical efficiency—25 seconds per case (12× faster than conventional approaches)—while attaining mean Dice scores of 0.66 (skeleton), 0.80 (organs), and 0.85 (spinal canal), with surface distances <3 mm. Quantitative errors in ADC and volume measurements are <10% and <4%, respectively. Radiologists rated the results as “good to excellent.”

0 citationsRead paper
Recent publications

Latest Papers

Morphologically Intelligent Perturbation Prediction with FORM

Oct 24, 2025

Current computational models are largely restricted to two-dimensional cell morphology modeling, failing to capture realistic three-dimensional structural dynamics under perturbations—thereby limiting the predictive accuracy and functional interpretability of virtual cells. To address this, we propose the first 3D cell morphology prediction framework supporting both unconditional generation and conditional simulation. Our method employs a multi-channel VQGAN-based high-fidelity morphological encoder coupled with a diffusion model to explicitly model perturbation-induced morphological trajectories, trained on over 65,000 3D multi-fluorescence cellular volume images. The framework accurately predicts 3D morphological evolution under chemical or genetic interventions, infers signaling pathway activity, disentangles combinatorial perturbation effects, and generalizes to unseen perturbation types. We further release MorphoEval, a standardized benchmarking suite for quantitative evaluation, establishing a new paradigm and reproducible evaluation standard for 3D virtual cell modeling.

0 citationsRead paper

Generalizing imaging biomarker repeatability studies using Bayesian inference: Applications in detecting heterogeneous treatment response in whole-body diffusion-weighted MRI of metastatic prostate cancer

May 14, 2025

Current imaging biomarkers lack a robust, quantitative framework for assessing heterogeneous treatment responses across multiple metastatic lesions in prostate cancer. Method: We propose the first general Bayesian repeatability assessment framework specifically designed for imaging biomarkers—overcoming limitations of traditional normality assumptions by enabling posterior inference for heterogeneous parameters under complex distributions (e.g., Dirichlet-Multinomial), accommodating multimodal imaging (e.g., whole-body diffusion-weighted MRI) and non-Gaussian biomarkers. Efficient Bayesian inference is implemented via Hamiltonian Monte Carlo, integrated with whole-body DWI volumetric analysis and validated on an mCRPC clinical cohort. Results: In two independent studies, the framework quantitatively identified significantly differential treatment responses in ~70% of metastatic lesions, demonstrating clinical validity and substantially improving the accuracy, robustness, and interpretability of heterogeneity quantification.

0 citationsRead paper

Signal-based AI-driven software solution for automated quantification of metastatic bone disease and treatment response assessment using Whole-Body Diffusion-Weighted MRI (WB-DWI) biomarkers in Advanced Prostate Cancer

May 13, 2025

This study addresses the challenge of quantitative assessment of bone metastatic burden and treatment response in advanced prostate cancer. Methodologically, we propose a fully automated AI framework based on whole-body diffusion-weighted MRI (WB-DWI): (1) a weakly supervised Residual U-Net generates skeletal probability maps to guide lesion detection; (2) a WB-DWI intensity statistical normalization strategy is introduced; and (3) a lightweight CNN enables end-to-end lesion segmentation, followed by registration with gADC maps to extract tumor diffusion volume (TDV) and median gADC—key quantitative biomarkers. Our key contribution is enabling bone metastasis quantification without per-lesion annotation. Validation demonstrates skeletal segmentation Dice scores of 0.6 (pelvis/spine), coefficient of variation (CV) of 4.6% for log-TDV and 3.6% for median gADC, and treatment response classification accuracy of 80.5%, sensitivity of 84.3%, and specificity of 85.7%. Processing time per case is 90 seconds.

0 citationsRead paper

Image deidentification in the XNAT ecosystem: use cases and solutions

Apr 29, 2025

To address DICOM image privacy leakage risks in the XNAT platform, this work proposes an automated de-identification workflow tailored for research data management. Methodologically, it integrates XNAT’s native API, standardized DICOM parsing, and a configurable rule engine to establish, for the first time, a systematic, multi-scenario-adaptive technical pathway within the XNAT ecosystem. It innovatively incorporates a BERT-NER model fine-tuned for address recognition to enhance anonymization of unstructured textual fields and evaluates performance using the MIDI-B benchmark framework. On the MIDI-B Challenge test set, the workflow achieves an overall de-identification accuracy of 99.61% and a real false-negative rate of only 0.19%, representing a 1.7-percentage-point improvement over the baseline. This work fills a critical gap in systematic, production-ready privacy protection for medical imaging data in XNAT environments and delivers a reusable, empirically validated technical paradigm for compliant research data governance.

0 citationsRead paper

A weakly-supervised deep learning model for fast localisation and delineation of the skeleton, internal organs, and spinal canal on Whole-Body Diffusion-Weighted MRI (WB-DWI)

Mar 26, 2025

Clinical whole-body diffusion-weighted imaging (WB-DWI) requires precise anatomical segmentation for accurate ADC quantification and tumor volume (TDV) measurement. However, manual delineation of the entire skeleton, visceral organs (liver, spleen, kidneys, bladder), and spinal canal is prohibitively time-consuming and clinically infeasible. To address this, we propose the first weakly supervised segmentation framework tailored for WB-DWI: a soft-label–guided 3D residual U-Net enabling simultaneous probabilistic segmentation of multiple structures without voxel-level annotations. Leveraging multi-center data, patch-based training, and probabilistic output maps, our method achieves clinical efficiency—25 seconds per case (12× faster than conventional approaches)—while attaining mean Dice scores of 0.66 (skeleton), 0.80 (organs), and 0.85 (spinal canal), with surface distances <3 mm. Quantitative errors in ADC and volume measurements are <10% and <4%, respectively. Radiologists rated the results as “good to excellent.”

0 citationsRead paper