LEARNING OUTCOMES
Upon successful completion of the course the students will be able:
• To be familiar with the design and operation of medical image analysis systems used in
Radiology.
519
• To be aware of the methodologies for mathematical quantification of texture properties,
edge and other properties of the image (e.g. homogeneity texture- inhomogeneity in
ultrasound images).
• To know the methods of classification into categories (e.g. benign – malignant cancer) of
images based on the quantified properties of the digital radiographic image.
• To have knowledge of the methods of evaluating the quality of medical image analysis
systems in Radiology.
Course aim:
Pattern Recognition System is a Decision Support System (DSS) that gives a possible
diagnosis which is taken into account by the radiologist, in order to make the final diagnosis.
With a command in the program, a series of elements from the image are collected (texture
characteristics – a series of numbers that express the texture of the cell nucleus), on the
basis of which a possible diagnosis of a degree of malignancy is made.
Computer analysis of digital medical images produced from modern radiological systems
(e.g. CT and MRI images) is important: (a) to draw useful conclusions that illustratively
differentiate between normal – healthy tissue and pathological tissue or pathological from
pathological tissue (Grade I / Grade III) and (b) for the classification of the imaged texture
into categories such as normal or abnormal. Image analysis differs from other types of
image processing methods, such as restoration and quality optimization, as the final
outcome is usually numerical rather than virtual. Consequently, image resolution is not
concerned with improving image quality. It deals with the diagnosis, in a similar way that
the radiologist examines an image: The computer examines the image, detects and
quantifies features and properties of the image and suggests a possible diagnosis (e.g.
benign – malignant cancer). A medical image analysis system includes: Production of
features that quantify medical image properties, system design with methods of
classification and evaluation of system reliability.
Course objective:
The student can formulate with a mathematical approach the structure of the radiological
image analysis systems used.
Course field:
The subject of Pattern Recognition briefly includes the following sections:
- Medical image analysis
- Data acquisition- Samples preparation
- Data processing
- Image resolution- Feature extraction
- Pattern Classification
- Integrated system design
- Methods of evaluation and reliability of the system
Course Outline