top of page
EMPAIAacademy_Logo-04 (002).jpg

Hands-on Workshop ECDP2023 in Budapest

At the European Congress on Digital Pathology 2023, we held a 90-minute workshop as a hands-on introduction to computational pathology both for pathologists and computer scientists. Participants were able choose between three tracks, where they worked in small groups to analyze histological images interactively, using scripted image processing or AI techniques. The workshop material is available via this page.

Whole-Slide Image Analysis mit QuPath

In the tutorial “WSI Analysis with QuPath”‚ we explore the open source software QuPath, a free to use platform, which provides tools for analysis of scanned histopathological slides, so-called whole slide images (WSIs). You will learn: (1) How to install QuPath and use it for viewing and annotating a WSI. (2) How to train the AI-based classification algorithm, which is provided by QuPath, so that this algorithm can find tumor and non-tumor areas on a WSI. (3) How to use QuPath to determine the Ki-67 index, a well-known method for assessment of cancer cell proliferation in various tumors.
Neither programming skills nor pathology knowledge are needed to participate in this tutorial.

Patch-basierte WSI-Analyse

In “Patch-based WSI analysis: tiling strategies and result-aggregation,” we will look at how to tile a WSI for a patch-based analysis such as Convolutional Neural Networks. We will learn about different strategies for splitting WSI and aggregating the results using a trivial custom segmentation algorithm.
Python knowledge is useful but not mandatory, as the concepts of image processing are the main focus.

Introduction to Training an AI for Histological Image Classification

In this tutorial, we will teach a computer to distinguish between tissue images of two classes. We will guide you through a simple machine  learning pipeline including training an algorithm, applying it, and displaying the results. You will learn to interpret the results and how to check them for plausibility in order to avoid common pitfalls of this type of approach.
There are a few lines of Python code involved, but no programming or machine learning experience is required.

bottom of page