Tumour grading and discrimination based on class assignment and quantitative texture analysis techniques

Medical imaging represents the utilisation of technology in biology for the purpose of noninvasively revealing the internal structure of the organs of the human body. It is a way to improve the quality of the patient's life through a more precise and rapid diagnosis, and with limited side-effects, leading to an effective overall treatment procedure. The main objective of this thesis is to propose novel tumour discrimination techniques that cover both micro and macro-scale textures encountered in computed tomography (CI') and digital microscopy (DM) modalities, respectively. Image texture can provide significant information on the (ab)normality of tissue, and this thesis expands this idea to tumour texture grading and classification. The fractal dimension (FO) as a texture measure was applied to contrast enhanced CT lung tumour images in an aim to improve tumour grading accuracy from conventional CI' modality, and quantitative performance analysis showed an accuracy of 83.30% in distinguishing between advanced (aggressive) and early stage (non-aggressive) malignant tumours. A different approach was adopted for subtype discrimination of brain tumour OM images via a set of statistical and model-based texture analysis algorithms. The combined Gaussian Markov random field and run-length matrix texture measures outperformed all other combinations, achieving an overall class assignment classification accuracy of 92.50%. Also two new histopathological multi resolution approaches based on applying the FO as the best bases selection for discrete wavelet packet transform, and when fused with the Gabor filters' energy output improved the accuracy to 91.25% and 95.00%, respectively. While noise is quite common in all medical imaging modalities, the impact of noise on the applied texture measures was assessed as well. The developed lung and brain texture analysis techniques can improve the physician's ability to detect and analyse pathologies leading for a more reliable diagnosis and treatment of disease.