Schedule

Introduction of the Workshop

[Dr Yafei Wang, University of Alberta, Canada]

[Introduction of the Challenges in Medical Imaging Data Analysis]

[Dr Peng Liu, University of Kent/Loughborough University, UK]

[When Parkinson’s Disease Meets Artificial Intelligence]

[Professor Huiyu Zhou, Leicester University, UK]

Abstract: Parkinson’s disease (PD) is a severe condition that affects the brain. PD causes huge problems in humans such as shaking and stiffness that become worse over time. Early diagnosis and prognosis of PD results in effective and personalised treatment, reduced care costs and better quality of life. In this talk, first of all, Zhou introduces fundamental knowledge about PD and the technologies used for PD identification. This talk is divided into two streams, animal mice- and human-based PD identification. Afterwards, Zhou reports how his research group deal with immersive challenges such as single and multiple mice detection and tracking, single and multiple mice behaviour recognition, and social behaviour analysis using new video analytics technologies developed within the team. Zhou presents the machine learning techniques used to distinguish between normal and PD mice through social behaviour analysis. Zhou also shows the artificial intelligence methods developed within his team for biomarker analysis. Finally, conclusions are given to summarise the talk. 

[Modelling Skewness with Applications]

[Professor Jian Zhang, University of Kent, UK]

Abstract: Skewness, which measures the asymmetry of a distribution, is an important data feature to characterise. Change of data skewness can serve as a basis for detecting an attack upon a sensor network, for providing an early warning for abrupt climate changes, for estimating  aggregates of small domain business, for modelling equity excess returns, for characterising sensitivity of anti-cancer drugs, among others(Buttyan et al., 2006; He et al., 2013; Colacito et al., 2016; Ferrane and Pacei, 2017;  Dhar et al.,1996). However, skew models may suffer from inferential drawbacks, namely singular Fisher information when it is close to symmetry and diverging of maximum likelihood estimation. This causes a large variation of the conventional maximum likelihood estimate (MLE).  Here, we propose a cross-validated maximum penalised MLE  to improve its performance when the underlying model is close to symmetry. We develop a theory for the proposal, where an asymptotic rate for the cross-validated penalty coefficient is derived. We further show that the proposed cross-validated estimate is asymptotically efficient under certain conditions.  In simulation studies and a real data application, we demonstrate that the proposed estimator can outperform the conventional MLE.

[Towards the Internet of Medical Things (IoMT) and advanced machine learning (ML) algorithms for diabetes, cancer and healthcare analysis]

[Professor Victor Chang, Aston University, UK]

Abstract: This keynote proposes an e-diagnosis system based on machine learning (ML) algorithms to be implemented on the Internet of Medical Things (IoMT) environment, particularly for diagnosing diabetes mellitus (type 2 diabetes) and breast cancer analysis. It can also be further adapted for heart disease detection analysis. The ML applications tend to be mistrusted because of their inability to show the internal decision-making process, resulting in slow uptake by end-users within certain healthcare sectors. This research delineates the use of three interpretable supervised ML models: Naïve Bayes classifier, random forest classifier, and J48 decision tree models to be trained and tested using the Pima Indians diabetes dataset. The performance of each algorithm is analyzed to determine the one with the best accuracy, precision, sensitivity, and specificity. An assessment of the decision process is also made to improve the model. It can be concluded that a Naïve Bayes model works well with a more fine-tuned selection of features for binary classification, while random forest works better with more features. IoMT is then used to help analyzing breast cancer and other medical analysis. Other methods such as support vector machine (SVM) and its variant models using the radial basis function kernel outperformed all other models we tested and those previously developed by others, achieving an accuracy of 99%. Additional scientific research for heart disease detection. We explain our analysis, research contributions and impacts. In summary, this keynote will present the latest research outputs for using applied IoMT and machine learning algorithms for healthcare domains.

[Doubly Adaptive Spatial Quantile Regression for Neuroimaging Data]

[Professor Linglong Kong, University of Alberta, Canada]

Abstract: Numerous contemporary techniques for analyzing neuroimaging data commonly utilize least-squares estimation and Gaussian smoothing. Unfortunately, these approaches are not robust against imaging outliers and artifacts, which are generally unavoidable in practice, and fail to accommodate the sharp edges and various spatial scales of different neurological regions of interest. To address these issues and provide greater insight into the distribution of neuroimaging responses in a regression framework, we propose a doubly adaptive spatial quantile regression model (DASQRM). Our approach leverages information across both spatial locations and quantile levels in an adaptive fashion to robustly estimate model parameters and perform hypothesis testing. Furthermore, we rigorously establish important statistical properties of our proposed estimator and present an efficient method for model estimation. Through three simulation studies resembling real-world neuroimaging data together with an analysis of a dataset from the ADHD-200 Initiative, we demonstrate the benefits of our doubly adaptive approach in reducing estimator noise and bias, increasing statistical power and efficiency in hypothesis testing, and improving estimate interpretability.