From Tubules to Clusters: Building the Foundations for Automatic Tubular Classification
Anne Mari and Martine, third-year Medical Technology students at the University of Bergen, are interning at PiV this semester. Their project focuses on developing pre-processing methods for an AI tool that may help identify pathological features in tubular lesions from kidney biopsies.
Hi! We are Anne Mari and Martine, and we are delighted to have our internship here at PiV this semester. Currently we both are in our third year of an engineering degree in Medical Technology at the University of Bergen and have chosen the chemistry specialization in our study program.
Here at PiV we are working with digital slides from biopsies of patients with chronic kidney disease. Our project deals with advancing the understanding of the landscape of tubular lesions. To date, little research has been conducted on tubular classification, and we are honored to contribute to this field. We have recently completed the first step of our project, which has been to create a code to extract small image patches centered on individual tubules from whole slide images from kidney biopsies. Next, we will work with extracting relevant AI-derived features for all tubular patches. We will then investigate different clustering strategies based on the collected features. The aim is to evaluate the clustering outcomes and assess whether they correspond to pathological classification.
One of the main challenges in our project is that tubules are highly variable in size and morphology. Therefore, we have experimented with different patch extraction strategies, such as scaling the patches or extracting them based on the largest tubule. We have also masked the background in some tubular patches, and we will evaluate which patching approach performs best when testing the different clustering strategies.
So far, we are enjoying our intern here! Working with digital pathology has been both fun and educational, allowing us to improve our programming and problem-solving abilities. Although we initially had limited knowledge of kidney pathology, it has been exciting to learn more about it and develop an understanding of its significance. The work can be challenging at times, but we get good help form Borghild. We have weekly meetings with her, where she guides and assists us in solving our problems.
It is also very interesting to see how the topics we study are implemented in real-world work, and to observe how well the team collaborates. We are grateful for all the experience we are gaining here and are looking forward to continuing our internship.
Find out more about our practice at PIV and our experiences on our blog.