Chun Fung Kwok
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chun Fung Kwok has a total of 63 co-authors. Here are the top co-authors by number of shared publications:
- Davis James McCarthy
Bioinformatics and Cellular Genomics, St. Vincent’s Institute of Medical Research, Fitzroy, Australia
7 shared publications - Carlos A. Peña-Solorzano
St Vincent's Institute of Medical Research, Melbourne, VIC, Australia
6 shared publications - Chong Wang
Stanford University
6 shared publications - Michael S. Elliott
St Vincent's Institute of Medical Research, Melbourne, VIC, Australia
6 shared publications - Helen Frazer
St Vincent's BreastScreen St Vincent's Hospital Melbourne Melbourne Victoria Australia
5 shared publications Yuanhong Chen
School of Computer Science, Australian Institute for Machine Learning, University of Adelaide, SA, Australia
5 shared publications
Latest Publications
- Open Access
- 3 March 2026
- Published by Elsevier in The Lancet Digital Health
- Vol. 8 (4) , 100987
AI-based BRAIx risk score for the intermediate-term prediction of breast cancer: a population cohort study
- 3 March 2026
- Published by Elsevier in The Lancet Digital Health
- Vol. 8 (4) , 100987
Open AccessAI-based BRAIx risk score for the intermediate-term prediction of breast cancer: a population cohort study
- Open Access
- 3 July 2025
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 81 (3)
Bayesian inference for copy number intra-tumoral heterogeneity from single-cell RNA-sequencing data
Copy number alterations (CNA) are important drivers and markers of clonal structures within tumors. Understanding these structures at single-cell resolution is crucial to advancing cancer treatments. The objective is to cluster single cells into clones and identify CNA events in each clone. Early attempts often sacrifice the intrinsic link between cell clustering and clonal CNA detection for simplicity and rely heavily on human input for critical parameters such as the number of clones. Here, we develop a Bayesian model to utilize single-cell RNA sequencing (scRNA-seq) data for automatic analysis of intra-tumoral clonal structure concerning CNAs, without reliance on prior knowledge. The model clusters cells into sub-tumoral clones, identifies the number of clones, and simultaneously infers the clonal CNA profiles. It synergistically incorporates input from gene expression and germline single-nucleotide polymorphisms. A Gibbs sampling algorithm has been implemented and is available as an R package Chloris. We demonstrate that our new method compares strongly against existing software tools in terms of both cell clustering and CNA profile identification accuracy. Application to human metastatic melanoma and anaplastic thyroid tumor data demonstrates accurate clustering of tumor and non-tumor cells and reveals clonal CNA profiles that highlight functional gene expression differences between clones from the same tumor.
- 3 July 2025
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 81 (3)
Open AccessBayesian inference for copy number intra-tumoral heterogeneity from single-cell RNA-sequencing data
Copy number alterations (CNA) are important drivers and markers of clonal structures within tumors. Understanding these structures at single-cell resolution is crucial to advancing cancer treatments. The objective is to cluster single cells into clones and identify CNA events in each clone. Early attempts often sacrifice the intrinsic link between cell clustering and clonal CNA detection for simplicity and rely heavily on human input for critical parameters such as the number of clones. Here, we develop a Bayesian model to utilize single-cell RNA sequencing (scRNA-seq) data for automatic analysis of intra-tumoral clonal structure concerning CNAs, without reliance on prior knowledge. The model clusters cells into sub-tumoral clones, identifies the number of clones, and simultaneously infers the clonal CNA profiles. It synergistically incorporates input from gene expression and germline single-nucleotide polymorphisms. A Gibbs sampling algorithm has been implemented and is available as an R package Chloris. We demonstrate that our new method compares strongly against existing software tools in terms of both cell clustering and CNA profile identification accuracy. Application to human metastatic melanoma and anaplastic thyroid tumor data demonstrates accurate clustering of tumor and non-tumor cells and reveals clonal CNA profiles that highlight functional gene expression differences between clones from the same tumor.
- Open Access
An Analysis of Vectorised Automatic Differentiation for Statistical Applications
Automatic differentiation (AD) is a general method for computing exact derivatives in complex sensitivity analyses and optimisation tasks, particularly when closed-form solutions are unavailable and traditional analytical or numerical methods fall short. This paper introduces a vectorised formulation of AD grounded in matrix calculus. It aligns naturally with the matrix-oriented style prevalent in statistics, supports convenient implementations, and takes advantage of sparse matrix representation and other high-level optimisation techniques that are not available in the scalar counterpart. Our formulation is well-suited to high-dimensional statistical applications, where finite differences (FD) scale poorly due to the need to repeat computations for each input dimension, resulting in significant overhead, and is advantageous in simulation-intensive settings—such as Markov Chain Monte Carlo (MCMC)-based inference—where FD requires repeated sampling and multiple function evaluations, while AD can compute exact derivatives in a single pass, substantially reducing computational cost. Numerical studies are presented to demonstrate the efficacy and speed of the proposed AD method compared with FD schemes.
Open AccessAn Analysis of Vectorised Automatic Differentiation for Statistical Applications
Automatic differentiation (AD) is a general method for computing exact derivatives in complex sensitivity analyses and optimisation tasks, particularly when closed-form solutions are unavailable and traditional analytical or numerical methods fall short. This paper introduces a vectorised formulation of AD grounded in matrix calculus. It aligns naturally with the matrix-oriented style prevalent in statistics, supports convenient implementations, and takes advantage of sparse matrix representation and other high-level optimisation techniques that are not available in the scalar counterpart. Our formulation is well-suited to high-dimensional statistical applications, where finite differences (FD) scale poorly due to the need to repeat computations for each input dimension, resulting in significant overhead, and is advantageous in simulation-intensive settings—such as Markov Chain Monte Carlo (MCMC)-based inference—where FD requires repeated sampling and multiple function evaluations, while AD can compute exact derivatives in a single pass, substantially reducing computational cost. Numerical studies are presented to demonstrate the efficacy and speed of the proposed AD method compared with FD schemes.
- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (8) , 5687-5699
Progressive Mining and Dynamic Distillation of Hierarchical Prototypes for Disease Classification and Localisation
Constructing effective representation of lesions is essential for disease classification and localization in medical image analysis. Prototype-based models address this by leveraging visual prototypes to capture representative lesion patterns, yet effectively handling the complexity of diverse lesion characteristics remains a critical challenge, as they typically rely on single-level, fixedsize prototypes and suffer from prototype redundancy. In this paper, we present HierProtoPNet, a new prototypebased framework designed to handle the complexity of lesions in medical images. HierProtoPNet leverages hierarchical visual prototypes across different semantic feature granularities to effectively capture diverse lesion patterns. To prevent redundancy and increase utility of the prototypes, we devise a novel prototype mining paradigm to progressively discover semantically distinct prototypes, offering multi-level complementary analysis of lesions. Also, we introduce a dynamic knowledge distillation strategy that allows transferring essential classification information across hierarchical levels, thereby improving generalisation performance. Comprehensive experiments show that HierProtoPNet achieves state-of-the-art classification performances in three benchmarks: binary breast cancer screening, multi-class retinal disease diagnosis, and multilabel chest X-ray classification. Quantitative assessments also illustrate HierProtoPNet's significant advantages in weakly-supervised disease localisation and segmentation.
- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (8) , 5687-5699
Progressive Mining and Dynamic Distillation of Hierarchical Prototypes for Disease Classification and Localisation
Constructing effective representation of lesions is essential for disease classification and localization in medical image analysis. Prototype-based models address this by leveraging visual prototypes to capture representative lesion patterns, yet effectively handling the complexity of diverse lesion characteristics remains a critical challenge, as they typically rely on single-level, fixedsize prototypes and suffer from prototype redundancy. In this paper, we present HierProtoPNet, a new prototypebased framework designed to handle the complexity of lesions in medical images. HierProtoPNet leverages hierarchical visual prototypes across different semantic feature granularities to effectively capture diverse lesion patterns. To prevent redundancy and increase utility of the prototypes, we devise a novel prototype mining paradigm to progressively discover semantically distinct prototypes, offering multi-level complementary analysis of lesions. Also, we introduce a dynamic knowledge distillation strategy that allows transferring essential classification information across hierarchical levels, thereby improving generalisation performance. Comprehensive experiments show that HierProtoPNet achieves state-of-the-art classification performances in three benchmarks: binary breast cancer screening, multi-class retinal disease diagnosis, and multilabel chest X-ray classification. Quantitative assessments also illustrate HierProtoPNet's significant advantages in weakly-supervised disease localisation and segmentation.
- Open Access
- 30 August 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-12
Comparison of AI-integrated pathways with human-AI interaction in population mammographic screening for breast cancer
Artificial intelligence (AI) readers of mammograms compare favourably to individual radiologists in detecting breast cancer. However, AI readers cannot perform at the level of multi-reader systems used by screening programs in countries such as Australia, Sweden, and the UK. Therefore, implementation demands human-AI collaboration. Here, we use a large, high-quality retrospective mammography dataset from Victoria, Australia to conduct detailed simulations of five potential AI-integrated screening pathways, and examine human-AI interaction effects to explore automation bias. Operating an AI reader as a second reader or as a high confidence filter improves current screening outcomes by 1.9–2.5% in sensitivity and up to 0.6% in specificity, achieving 4.6–10.9% reduction in assessments and 48–80.7% reduction in human reads. Automation bias degrades performance in multi-reader settings but improves it for single-readers. This study provides insight into feasible approaches for AI-integrated screening pathways and prospective studies necessary prior to clinical adoption.
- 30 August 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-12
Open AccessComparison of AI-integrated pathways with human-AI interaction in population mammographic screening for breast cancer
Artificial intelligence (AI) readers of mammograms compare favourably to individual radiologists in detecting breast cancer. However, AI readers cannot perform at the level of multi-reader systems used by screening programs in countries such as Australia, Sweden, and the UK. Therefore, implementation demands human-AI collaboration. Here, we use a large, high-quality retrospective mammography dataset from Victoria, Australia to conduct detailed simulations of five potential AI-integrated screening pathways, and examine human-AI interaction effects to explore automation bias. Operating an AI reader as a second reader or as a high confidence filter improves current screening outcomes by 1.9–2.5% in sensitivity and up to 0.6% in specificity, achieving 4.6–10.9% reduction in assessments and 48–80.7% reduction in human reads. Automation bias degrades performance in multi-reader settings but improves it for single-readers. This study provides insight into feasible approaches for AI-integrated screening pathways and prospective studies necessary prior to clinical adoption.
- 10 May 2024
- Published by Elsevier in Medical Image Analysis
- Vol. 96, 103192
BRAIxDet: Learning to detect malignant breast lesion with incomplete annotations
- 10 May 2024
- Published by Elsevier in Medical Image Analysis
- Vol. 96, 103192
BRAIxDet: Learning to detect malignant breast lesion with incomplete annotations
- Open Access
- 22 December 2023
- Published by Walter de Gruyter GmbH in Studies in Nonlinear Dynamics & Econometrics
- Vol. 28 (2) , 403-434
Posterior Manifolds over Prior Parameter Regions: Beyond Pointwise Sensitivity Assessments for Posterior Statistics from MCMC Inference
Increases in the use of Bayesian inference in applied analysis, the complexity of estimated models, and the popularity of efficient Markov chain Monte Carlo (MCMC) inference under conjugate priors have led to more scrutiny regarding the specification of the parameters in prior distributions. Impact of prior parameter assumptions on posterior statistics is commonly investigated in terms of local or pointwise assessments, in the form of derivatives or more often multiple evaluations under a set of alternative prior parameter specifications. This paper expands upon these localized strategies and introduces a new approach based on the graph of posterior statistics over prior parameter regions (sensitivity manifolds) that offers additional measures and graphical assessments of prior parameter dependence. Estimation is based on multiple point evaluations with Gaussian processes, with efficient selection of evaluation points via active learning, and is further complemented with derivative information. The application introduces a strategy to assess prior parameter dependence in a multivariate demand model with a high dimensional prior parameter space, where complex prior-posterior dependence arises from model parameter constraints. The new measures uncover a considerable prior dependence beyond parameters suggested by theory, and reveal novel interactions between the prior parameters and the elasticities.
- 22 December 2023
- Published by Walter de Gruyter GmbH in Studies in Nonlinear Dynamics & Econometrics
- Vol. 28 (2) , 403-434
Open AccessPosterior Manifolds over Prior Parameter Regions: Beyond Pointwise Sensitivity Assessments for Posterior Statistics from MCMC Inference
Increases in the use of Bayesian inference in applied analysis, the complexity of estimated models, and the popularity of efficient Markov chain Monte Carlo (MCMC) inference under conjugate priors have led to more scrutiny regarding the specification of the parameters in prior distributions. Impact of prior parameter assumptions on posterior statistics is commonly investigated in terms of local or pointwise assessments, in the form of derivatives or more often multiple evaluations under a set of alternative prior parameter specifications. This paper expands upon these localized strategies and introduces a new approach based on the graph of posterior statistics over prior parameter regions (sensitivity manifolds) that offers additional measures and graphical assessments of prior parameter dependence. Estimation is based on multiple point evaluations with Gaussian processes, with efficient selection of evaluation points via active learning, and is further complemented with derivative information. The application introduces a strategy to assess prior parameter dependence in a multivariate demand model with a high dimensional prior parameter space, where complex prior-posterior dependence arises from model parameter constraints. The new measures uncover a considerable prior dependence beyond parameters suggested by theory, and reveal novel interactions between the prior parameters and the elasticities.
- Open Access
Bayesian inference for copy number intra-tumoral heterogeneity from single-cell RNA-sequencing data
High-resolution molecular characterization of intra-tumoral clonal structure defined by genomic and epigenomic alterations is crucial in understanding the natural history of tumors and advancing cancer treatment strategies. Copy number alterations (CNA) are of notable importance as both drivers and markers of clonal structure that can now be assayed at individual cell resolution. However, specific computational methods are needed for accurate inference of clonal profiles and cell states from sparse and noisy single-cell ’omics data. Here, we develop a new Bayesian model to utilize single-cell RNA sequencing (scRNA-seq) data for automatic analysis of intra-tumoral clonal structure with respect to CNAs, without reliance on prior knowledge. The model clusters cells into sub-tumoral clones while simultaneously identifying CNA events in each clone, jointly modelling input from gene expression and germline single-nucleotide polymorphisms. Unlike previous methods, our approach automatically infers the number of clones present in the tumor. In detailed simulation studies our model frequently achieves very high (>90%) cell clustering accuracy and high (>80%) CN state inference accuracy, even in settings of high variance and sparsity. Overall, our method compares strongly against existing software tools. Application to human metastatic melanoma tumor data demonstrates accurate clustering of tumor and non-tumor cells, and reveals clonal CNA profiles that highlight functional gene expression differences between clones from the same tumor. Our method is implemented in a publicly-available, open-source R package, Chloris.
Open AccessBayesian inference for copy number intra-tumoral heterogeneity from single-cell RNA-sequencing data
High-resolution molecular characterization of intra-tumoral clonal structure defined by genomic and epigenomic alterations is crucial in understanding the natural history of tumors and advancing cancer treatment strategies. Copy number alterations (CNA) are of notable importance as both drivers and markers of clonal structure that can now be assayed at individual cell resolution. However, specific computational methods are needed for accurate inference of clonal profiles and cell states from sparse and noisy single-cell ’omics data. Here, we develop a new Bayesian model to utilize single-cell RNA sequencing (scRNA-seq) data for automatic analysis of intra-tumoral clonal structure with respect to CNAs, without reliance on prior knowledge. The model clusters cells into sub-tumoral clones while simultaneously identifying CNA events in each clone, jointly modelling input from gene expression and germline single-nucleotide polymorphisms. Unlike previous methods, our approach automatically infers the number of clones present in the tumor. In detailed simulation studies our model frequently achieves very high (>90%) cell clustering accuracy and high (>80%) CN state inference accuracy, even in settings of high variance and sparsity. Overall, our method compares strongly against existing software tools. Application to human metastatic melanoma tumor data demonstrates accurate clustering of tumor and non-tumor cells, and reveals clonal CNA profiles that highlight functional gene expression differences between clones from the same tumor. Our method is implemented in a publicly-available, open-source R package, Chloris.
- 21 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 43 (1) , 392-404
An Interpretable and Accurate Deep-Learning Diagnosis Framework Modeled With Fully and Semi-Supervised Reciprocal Learning
The deployment of automated deep-learning classifiers in clinical practice has the potential to streamline the diagnosis process and improve the diagnosis accuracy, but the acceptance of those classifiers relies on both their accuracy and interpretability. In general, accurate deep-learning classifiers provide little model interpretability, while interpretable models do not have competitive classification accuracy. In this paper, we introduce a new deep-learning diagnosis framework, called InterNRL, that is designed to be highly accurate and interpretable. InterNRL consists of a student-teacher framework, where the student model is an interpretable prototype-based classifier (ProtoPNet) and the teacher is an accurate global image classifier (GlobalNet). The two classifiers are mutually optimised with a novel reciprocal learning paradigm in which the student ProtoPNet learns from optimal pseudo labels produced by the teacher GlobalNet, while GlobalNet learns from ProtoPNet’s classification performance and pseudo labels. This reciprocal learning paradigm enables InterNRL to be flexibly optimised under both fully- and semi-supervised learning scenarios, reaching state-of-the-art classification performance in both scenarios for the tasks of breast cancer and retinal disease diagnosis. Moreover, relying on weakly-labelled training images, InterNRL also achieves superior breast cancer localisation and brain tumour segmentation results than other competing methods.
- 21 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 43 (1) , 392-404
An Interpretable and Accurate Deep-Learning Diagnosis Framework Modeled With Fully and Semi-Supervised Reciprocal Learning
The deployment of automated deep-learning classifiers in clinical practice has the potential to streamline the diagnosis process and improve the diagnosis accuracy, but the acceptance of those classifiers relies on both their accuracy and interpretability. In general, accurate deep-learning classifiers provide little model interpretability, while interpretable models do not have competitive classification accuracy. In this paper, we introduce a new deep-learning diagnosis framework, called InterNRL, that is designed to be highly accurate and interpretable. InterNRL consists of a student-teacher framework, where the student model is an interpretable prototype-based classifier (ProtoPNet) and the teacher is an accurate global image classifier (GlobalNet). The two classifiers are mutually optimised with a novel reciprocal learning paradigm in which the student ProtoPNet learns from optimal pseudo labels produced by the teacher GlobalNet, while GlobalNet learns from ProtoPNet’s classification performance and pseudo labels. This reciprocal learning paradigm enables InterNRL to be flexibly optimised under both fully- and semi-supervised learning scenarios, reaching state-of-the-art classification performance in both scenarios for the tasks of breast cancer and retinal disease diagnosis. Moreover, relying on weakly-labelled training images, InterNRL also achieves superior breast cancer localisation and brain tumour segmentation results than other competing methods.
- Open Access
- 1 March 2023
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 5 (2)
ADMANI: Annotated Digital Mammograms and Associated Non-Image Datasets
“Just Accepted” papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review ...
- 1 March 2023
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 5 (2)
Open AccessADMANI: Annotated Digital Mammograms and Associated Non-Image Datasets
“Just Accepted” papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review ...