Institutional Repository of Vilnius University
Inkstų kanalėlių skaitmeninės mikroskopijos nuotraukų klasifikavimas pasitelkiant mašininį mokymąsi /
Abstract
dc:descriptionKidney health is crucial for maintaining the body's internal balance, and timely detection and evaluation are critical, especially in cases of transplant or overall kidney failure. In digital pathology, whole slide images, have become indispensable in the diagnosis of kidney-related injuries. However, inter-observer concordance and concerns about the reliability of classification systems pose challenges. Artificial intelligence models offer promising solutions, providing repeatable and efficient results for automatic workflows. This thesis aims to assess the feasibility of classifying kidney tubules from whole slide images using image analysis, computer vision techniques to create a feature extraction pipeline, then evaluating a dataset of extracted features for classification feasibility, and developing a machine learning model for predicting kidney tubule classes. The implemented pipeline successfully addressed challenges associated with large-scale image processing, providing enhanced quality and computation speed. The workflow opens possibilities for large-scale studies and potential discoveries of new biomarkers through quantitative feature comparisons between pathological groups. Although the dataset lacked informative features for reasonable clustering. The classification model development highlighted the importance of meaningful features and the need for further improvements. Overall, the feature extraction pipeline represents a significant step towards creating an end-to-end system for digital renal pathology. Further integration with deep learning solutions could help enhance the analytical capabilities in kidney pathology.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Jokūbaitis, Daumantas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192845104&prefLang=en_US