Steatosis quantification in Non-Alcoholic Fatty Liver Disease: A statistical approach for comparing different image processing procedures

Mindouro, Andreia Sofia Pedro

http://hdl.handle.net/10400.5/102427

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
TM_Andreia_Mindouro.pdf		30.72 MB	Adobe PDF	Download

Send Feedback

Authors

Mindouro, Andreia Sofia Pedro

Advisor(s)

Sousa, Lisete Maria Ribeiro de, 1972-

Paixão, Tiago

Abstract(s)

The liver plays a crucial role in metabolism and immunity, aiding in nutrient metabolism, detoxification, protein production, digestion, and vitamin storage. The liver is essential for overall well-being. Hepatic steatosis, or fatty liver, is primarily caused by obesity, type 2 diabetes, poor diet, and chronic alcohol consumption. A specific form, Non-alcoholic fatty liver disease (NAFLD), is linked to metabolic issues and high-fat diets. NAFLD is characterized by fat accumulation in liver cells without alcohol involvement, ranging from benign steatosis to more severe conditions like cirrhosis. Diagnosing NAFLD typically involves blood or imaging tests, though liver biopsy remains the most accurate method. However, the standard scoring system for assessing steatosis based on biopsy results is prone to inaccuracies due to variability among observers. Animal models, particularly mice, are commonly used in research to study NAFLD progression and evaluate treatments. New techniques such as machine learning and quantitative data analysis have improved the accuracy of NAFLD diagnosis. In this study, mouse liver samples will be used to compare two automated image analysis plugins, Waikato Environment for Knowledge Analysis (Weka) and Saturation, to determine their accuracy in assessing steatosis. Weka applies machine learning techniques, specifically the Random Forest classifier, to distinguish vacuoles from other cellular components based on morphological features. The study also aims to identify the optimal combinations of magnification, resolution, and the number of images for accurate analysis, and to assess whether there is a pattern in vacuole size related to the steatosis percentage. The appropriate application of statistical methods, such as generalized linear mixed models and bootstrap techniques, enables robust data interpretation and supports scientific communication and evidence-based decision-making. Weka detected more vacuoles than Saturation, especially at low magnification and high resolution. Three to four images were sufficient for reliable estimates, and vacuole size was associated with steatosis severity. Accurate classification of steatosis is vital, as NAFLD is a significant manifestation of metabolic syndrome and poses increased mortality risks, particularly from cardiovascular diseases. Early diagnosis and targeted treatment are key to improving patient outcomes and addressing this global health concern.