engleski

Towards data-driven approaches for medical image analysis

The idea of computer-aided diagnosis in diagnosing and treating illnesses has been around from nineteen-eighties. It was then that scientists discovered that, by applying statistical algorithms on real-world patient data, using machine learning, one can establish useful (almost-)out-of-the-box mathematical models that perform well at describing some problems. In the last decade, significant increases occurred worldwide in the level of informatics-readiness of clinical centres, in the availability of standardised technology for data exchange and storage, and in the abundance of quality medical radiology techniques. This, in turn, resulted in an explosion in the availability of voluminous data, enabling further advances in the field of computer-aided diagnosis and treatment. In this lecture, in addition to some fundamentals related to the field, several topics concerning medical image analysis will be discussed: utilising information theory for organ segmentation, learning predictive models for diagnosing injuries, data pre-processing for reducing model complexity, transfer learning in medical radiology domain and diagnosing illnesses using hyperspectral imaging.

machine learning ; medical image analysis

nije evidentirano