Quantitative Measurement of Irregular Tubules in Breast Cancer

Chai, Ji Xuan (2023) Quantitative Measurement of Irregular Tubules in Breast Cancer. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text - Chai Ji Xuan.pdf Restricted to Registered users only Download (5MB)

Abstract

Breast cancer has become one of the prominent issues in healthcare, accounting for the highest number of incident cancer cases among all other cancer types in Malaysia by 2020. Therefore, assessment of breast cancer is essential to grade and stage the manifestation of the tumour in a qualitative method. The Nottingham Histological Grading (NHG) system is one of the most widely adopted clinical variables in breast cancer that comprises information from three aspects, namely, the degree of tubule formation, nuclear pleomorphism, and mitotic counts. The traditional method of assessing tubule formation is through manual inspection of histological slides under a light microscope, which is tedious and prone to human errors. In addition to the uneven histological staining and irregular tubule structure due to tubule sectioning procedures being inevitable during the preparation of histological slides, it is essential to standardise the human evaluation process to a uniform and standardised decision based on the quantitative measurement of tubules. This research will be focusing on the detection of tubule structures based on the quantitative measurement of their morphological features. The k-means clustering method and locally adaptive thresholding will be implemented for nuclei and lumen segmentation on the clinical data from the Pathology Department, Hospital Tuanku Fauziah, Kangar, Perlis, Malaysia. The windowed analysis will be performed to segment the tubules based on the nuclei and lumen segmentations for morphological feature study of the tubules. This method is resilient to shape variation and image transformation such as scaling and translation. The Haralick texture features were selected to perform texture analysis on the tubule candidates and were used to build the mathematical model for quantification using logistic regression. The SV, SE, and IMC2 were found to be correlated to irregular tubule structures and the predicted probability of non-tubule is approximated to one, whereas the tubule structures to zero. An SVM classifier using a coarse Gaussian kernel was applied for the classification of the results. The tubule classification system achieved promising Accuracy, Precision, Recall, and F1 Score of 0.9500, 0.9000, 1.0000, and 0.9500, respectively.