Interpretation And Analysis Of Medical Images Using Computational And Machine Learning Techniques
University of Siena - Dipartimento di Ingegneria dell'Informazione e Scienze Matematiche
May 06 h. 09:00-11:00 Aula E
May 07 h. 15:00-17:00 Aula infor. 124
May 08 h. 09:00-11:00 Aula 20
May 09 h. 15:00-17:00 Aula infor. 147
May 10 h. 09:00-11:00 Aula 20
May 13 19h 09:00-11:00 Aula E
May 14 15:00-17:00 Aula infor. 124
May 15 h. 09:00-11:00 Aula 20
May 16 h. 15:00-17:00 Aula infor. 124
May 17 h. 09:00-11:00 Aula 20
Medical imaging is the imaging of the human body for the purposes of diagnosis, treatment planning (e.g. surgical intervention), therapy and patient monitoring. The diverse variety of imaging methods include digital x-ray, computerized tomography (CT), magnetic resonance imaging (MRI), ultrasound and are characterized by producing image data which is assessed by radiologists and clinicians. Increasingly however, powerful techniques from image analysis, computer vision and machine learning are being used to aid the tasks of radiologists, clinicians and surgeons to produce more accurate measurements and make automated and semi-automated predictions of the type and nature of the patient's pathology. These methods are becoming essential as the vast size and numbers of data makes it impossible to assess them manually, but perhaps more importantly, they also make better the accuracy and reliability of the image-based assessments, reducing heath care costs and improving patient outcomes.
This short course aims to introduce the principal imaging methods, method in image analysis and machine learning use to analyse the image data, and examples of applications of these methods in clinical diagnosis and therapy planning. After an overview of how the medical images are acquired and the mathematics of tomographic reconstruction, the course will then consider challenges faced by computational methods on medical imaging data. After a brief discussion of the data and how it can be usefully visualised, it will present several important generic methods of analysis: shape and appearance modelling for anatomy, image registration and alignment, image segmentation and classification. With real-world examples and data, we will look in more detail at the application of these methods, including some from machine learning, to solve several interesting clinical problems using radiological data from x-rays, CT and MRI.
Pre-requisites, Delivery and Assessment
The course assumes a post graduate-level background in one or more areas of: mathematics/statistics, electronic engineering, information engineering and computer science. The course will be delivered as 16 hours of lecture/seminars and lab-based practical/interactive demonstrations, conducted over 8 days. It will cover the use of appropriate programming tools for medical image computing, such as MATLAB and Python and associated libraries, e.g. SimpleITK.
The course will be assessed by coursework with an individual (practical) assignment to be conducted on some 'challenge' data to produce results and a documented laboratory report.