讲座编号:jz-yjsb-2018-y058
讲座问题:Spatial modelling of intratumoral activity from Positron Emission Tomography imaging data
主 讲 人:Eric Wolsztynski 爱尔兰科克大学数学学院教授
讲座时间:2018年10月17日(星期三)下昼16:00
讲座所在:阜成路东校区一号楼241 (理学院聚会室)
加入工具:理学院西席和相关学科的研究生、本科生
主理单位:研究生院
承办单位:理学院
主讲人简介:
Eric Wolsztynski,coordinator of the MSc in Data Science and Analytics (jointly with Computer Science) and assistant coordinator of the School's Higher Diploma in Statistics. His main research activity in medical diagnostic imaging analysis (developing novel statistical techniques for the analysis of medical imaging data, across PET, CT and MR imaging modalities). This research has a specific focus on innovative non-invasive pseudobiomarkers (i.e. image-based assessment of a patient’s biology and metabolic activity that would provide diagnostic and prognostic value in the treatment of cancer and other medical conditions, in view to complement or replace biopsy-based assessment).
This research work involves a substantial amount of components taught within the actuarial programmes taught at the school of mathematical sciences, including survival analysis, time series analysis, regularization and graduation techniques, and importance sampling for statistical learning. For example, within the research group we develop statistical models for the analysis of temporal PET data that rely on nonparametric survival curve estimation. An important part of my research also consists in designing significant (Cox proportional hazard) prognostic models and risk stratification strategies using machine learning techniques. Implementations for this research are done mainly in R. I also have an involvement in research on computational and inferential statistics with applications to actuarial science problems. In particular I have recently supervised a couple
of postgraduate research projects on late-life mortality modeling, which were presented at the national Conference on Applied Statistics in Ireland.
主讲内容:
This research work focuses on the development of statistical methods for the analysis of cancer imaging data. We consider in particular problems related to the assessment of prognosis, staging and chances of disease recurrence, mainly from PET imaging data but also from MRI and CT information. Spatial heterogeneity of the 18F-fluorodeoxyglucose uptake pattern in particular has been established as a strong prognostic indicator for sarcoma, lung, breast and other cancers. Our approach consists in developing new quantification methodologies for characterization of tumour metabolism and structure. Spatial models of the volumetric distribution of PET tracer uptake within the volume of interest are used to extract relevant metabolic and structural descriptors of the tumour, for the assessment of prognosis and therapeutic response. This work involves a number of technical aspects including computational statistics, adaptive estimation and statistical learning. The main application of this research is cancer patient-adaptive treatment, but it also links with problems found in other biomedical and actuarial applications.