Skip to main content
Department of Information Technology

CBA Seminars in Autumn 2023

Date Time Title Speaker
2023-08-28 14:15-15:00
Theatrum Visuale

Zoom-link here
Image analysis for Antibiotic Susceptibility Testing and corresponding production inspection - using classical image analysis and/or machine learning
In this presentation, I will present some examples of image analysis tasks related to antibiotic susceptibility testing and corresponding production inspection. These examples include both classical rule-based image analysis and machine learning-based analysis. I will try to give arguments for why we have chosen to use one or the other method. I will also share some experiences from working in industrial versus academic environments.
Petter Ranefall, Sysmex-Astrego
2023-09-04 14:15-15:00
Theatrum Visuale
Dynamic Contrast-Enhanced MRI and the Tofts Model
Perfusion relates to the delivery of blood to tissue and is medically relevant, for example, for differentiation of tumors and for monitoring the effect of treatment. One way of obtaining quantitative information related to perfusion is to make use of dynamic contrast-enhanced MRI. A series of MRI volumes of a patient is imaged over a period of time, as a contrast agent is injected. Based on the changes in signal over time, caused by the contrast agent, quantitative parameters can be extracted for each voxel by fitting a model to the data. I will describe how the commonly used Tofts model can be used for this purpose, as well as outlining some image analysis related challenges relevant to dynamic contrast enhanced MRI and perfusion.
Teo Asplund, RaySearch Laboratories
2023-09-11 14:15-15:00
Theatrum Visuale
Image processing and AI in visual inspection
Abstract: Image processing and artificial intelligence (AI) have emerged as transformative technologies in visual inspection across various domains, including healthcare and industries. Here, I will present some specific use cases in medical image segmentation & classification using classical image processing & ML methods, AI based object and anomaly detection in quality control & manufacturing process, AI based peak detection & extraction from data and how InfraVis: national infrastructure for data Visualization can assist and help in several scientific domains.
Nikita Singh, InfraVis
2023-09-18 14:15-15:00
Theatrum Visuale
Document Image Processing for Handwritten Text Recognition - Deep Learning-based Transliteration of Astrid Lindgren’s Stenographic Manuscripts
(PhD Rehearsal)

Abstract: Document image processing and handwritten text recognition have been applied to a variety of materials, scripts, and languages, both modern and historic. They are crucial building blocks in the on-going digitisation efforts of archives, where they aid in preserving archival materials and foster knowledge sharing. The latter is especially facilitated by making document contents available to interested readers who may have little to no practice in, for example, reading a specific script type, and might therefore face challenges in accessing the material.
The first part of this dissertation focuses on reducing editorial artefacts, specifically in the form of struck-through words, in manuscripts. The main goal of this process is to identify struck-through words and remove as much of the strikethrough artefacts as possible in order to regain access to the original word. The second part of this dissertation is centred around applying handwritten text recognition to the stenographic manuscripts of Swedish children's book author Astrid Lindgren (1907 - 2002). Manually transliterating stenography, also known as shorthand, requires special domain knowledge of the script itself. Therefore, the main focus of this part is to reduce the required manual work, aiming to increase the accessibility of the material.
Raphaela Heil
2023-09-25 14:15-15:00
Theatrum Visuale
TissUUmaps 3: Improvements in interactive visualization, exploration, and quality assessment of large-scale spatial omics data
Abstract: Spatially resolved techniques for exploring the molecular landscape of tissue samples, such as spatial transcriptomics, often result in millions of data points and images too large to view on a regular desktop computer, limiting the possibilities in visual interactive data exploration. TissUUmaps is a free, open-source browser-based tool for GPU-accelerated visualization and interactive exploration of 107+ data points overlaying tissue samples. TissUUmaps 3 provides instant multiresolution image viewing and can be customized, shared, and also integrated into Jupyter Notebooks. TissUUmaps introduces new modules where users can visualize markers and regions, explore spatial statistics, perform quantitative analyses of tissue morphology, and assess the quality of decoding in situ transcriptomics data. Thanks to targeted optimizations the time and cost associated with interactive data exploration were reduced, TissUUmaps 3 enables to handle the scale of today's spatial transcriptomics methods. TissUUmaps 3 provides significantly improved performance for large multiplex datasets as compared to previous versions and we envision TissUUmaps to contribute to broader dissemination and flexible sharing of largescale spatial omics data.
Christophe Avenel, BIIF
2023-10-02 14:15-15:00
Theatrum Visuale
Explosion Imaging of Proteins
Abstract: Proteins are the building blocks of life, and their function can be determined by their structure. But how do we find the structure of something like a protein?
This talk will introduce the different types of protein imaging used today as well as how we got here. I will also introduce a new type of imaging which we call "Explosion imaging". I will explain the basics of how this method works, and will show some examples of how this method can be used to classify different proteins based on their "explosion footprint".
Tomas André
2023-10-09 14:15-15:00
Theatrum Visuale
Bias mitigation in oral cancer cytology data
Abstract: Early diagnostics of oral cancer is a prerequisite to successful treatment. Routine screening tests by brushing oral cavity could potentially decrease oral cancer incidence. AI-based systems could assist cytotechnologists in decision-making, by, for example, highlighting the most suspicious cells in a sample. However, there are numerous challenges before such systems are introduced into clinical practice. One of the challenges is the bias of AI-based systems, which can be caused, e.g., by the presence of site-specific signatures in data. In the case of oral cancer data, we are dealing with the overrepresentation of certain types of images with positive class labels. Such images are highlighted by AI-based methods but are not related to malignancy according to cytotechnologist. Therefore, we are interested in applying bias mitigation methods to oral cancer data to steer AI-based methods to draw less attention to the overrepresentation of image types not related to malignancy.
Nadezhda Koriakina
2023-10-16 14:15-15:00
Theatrum Visuale + https://uu-se.zoom.us/j/65294652066
Mass spectrometry imaging – hundreds of layers of spatial data to interpret
Abstract: Mass spectrometry imaging of thin tissue sections provides chemical information and abundance of hundreds of molecules in every pixel. This has the potential to reveal unique insights into biological systems to understand health, disease and processes of treatment. However, the enormous amount of data is not trivial to mine. Here, I will present the technique of mass spectrometry imaging, the data that is acquired and discuss potentials for automation of data mining with the audience.
Ingela Lanekoff
2023-10-23 14:15-15:00 No seminar
Staff meeting
2023-10-30 14:15-15:00 No seminar
Höstlov (Autumn school-break)
2023-11-06 14:15-15:00
Zoom-link

Note: This seminar session will be organized given via zoom (only).
Tackling climate change using computer vision and generalisations of it
Abstract: In recent years, a number of powerful tools based on computer vision have seen the light and been increasingly applied in fields such as ecology and hydrology, important arenas for climate change adaptation. Since the last twenty-or-so years, this means mainly that deep convolutional neural networks are being applied in a number of tasks where the data are images. But the technique is more general than this, and there are reasons to discuss slight generalisations both in data that resembles images but are not captured using optical camera sensors and when the models are variations of the typical computer vision models as we know them. In this talk, I will go through some example applications that fall into this category and present some suggested solutions for them. This includes convolutional neural networks used for spatially distributed data such as satellite imagery and information from geographic information systems (GIS) in combination with data that aren't directly spatially distributed (or very sparsely so), and soundscape analysis which concerns soundwave data but this is generally transformed to spectrograms and modeled using techniques from computer vision. Finally, if time allows, I'll share some thoughts on annotation efficient machine learning with applications in computer vision.

About the speaker: Olof Mogren is the director of deep learning research at RISE, with a research interest in AI for ecology and climate change. He holds a PhD degree from 2018 in Computer Science from Chalmers University of Technology, with a thesis about deep representation learning.
Olof Mogren RISE
2023-11-13 14:15-15:00
Theatrum Visuale
Have CBA and CBIA something in common?
Abstract: CBA (Centrum för BildAnalysis) and CBIA (Centre for Biomedical Image Analysis, cbia.fi.muni.cz) are two groups interested in image analysis. I will give very short overview of CBIA teaching and research activities and introduce its industrial cooperation with Tescan company - one of the leading SEM (Scanning Electron Microscopy) manufacture - in the area of image processing and analysis.
The presentation aims to open possibilities and facilitate collaboration along identified common interests.
Pavel Matula
2023-11-20 14:15-15:00
Theatrum Visuale
Adapting Deep Learning for Microscopy: Interaction, Application, and Validation
PhD Thesis Defense Rehearsal

Abstract: Digital microscopy and automation have enabled biologists to conduct faster and larger-scale experiments with a sharp increase in the data generated. Image analysis is a crucial tool for biologists in the objective interpretation and extraction of quantitative measurements from microscopy data. Recently, deep learning techniques have shown superior performance in various image analysis tasks. The models learn feature representations from the data by optimizing for a task. However, the techniques require a significant amount of annotated data to perform well. Domain experts are required to annotate microscopy data, making it expensive and time-consuming. The models offer no insight into their prediction, and the learned features are not directly interpretable. This poses challenges to the reliable utilization of the technique in high-trust applications such as drug discovery or disease detection. The work in this thesis presents frameworks and methods to solve the practical challenges of applying deep learning in microscopy. The application-specific evaluation approaches were presented to validate the approaches, aiming to increase trust in the system. This thesis is aimed towards better utilization and adaptation of the DL methods and techniques to the microscopy data. We show that the annotation burden for the user can be significantly reduced by intuitive annotation frameworks and using contemporary deep-learning paradigms. We further propose architectural modifications in the models to adapt to the requirements and demonstrate the utility of application-specific analysis in microscopy.
Ankit Gupta
2023-11-27 14:15-15:00
Theatrum Visuale
DEPICTER: Deep rEPresentatIon ClusTERing
Presentation of DEPICTER, a tool for annotating histopathology images interactively, currently under review.
Automatic segmentation of histopathology whole-slide images (WSI) usually involves supervised training of deep learning models with pixel-level labels to classify each pixel of the WSI into tissue regions such as benign or cancerous. However, fully supervised segmentation requires large-scale data manually annotated by experts, which can be expensive and time-consuming to obtain. Non-fully supervised methods, ranging from semi-supervised to unsupervised, have been proposed to address this issue and have been successful in WSI segmentation tasks. But these methods have mainly been focused on technical advancements in algorithmic performance rather than on the development of practical tools that could be used by pathologists or researchers in real-world scenarios. In contrast, we present DEPICTER (Deep rEPresentatIon ClusTERing), an interactive segmentation tool for histopathology annotation that produces a patch-wise dense segmentation map at WSI level.
Eduard Chelebian Kocharyan
2023-12-04 14:15-15:00
Theatrum Visuale
CANCELLED!!
Advancing Molecular Insights: Report on the EMBL-SLL Internship in Rome

This seminar encapsulates an internship experience at the European Molecular Biology Laboratory (EMBL) in Rome, conducted in September 2023. My primary project, supervised by Alvaro Crevenna in collaboration with the Boulard group, aimed to profile the expression patterns of 200 murine olfactory receptor genes within the nasal epithelium of wild-type and mutated mice using a microscopy-based method and image analysis pipeline. The seminar will include the interdisciplinary nature of the research, hands-on laboratory work with cutting-edge in situ sequencing techniques, and the invaluable skills and knowledge acquired during the internship. My journey of learning, growth, and scientific exploration emphasizes the potential for significant contributions to the field of molecular biology.
Andrea Behanova
2023-12-11 14:15-15:00
Theatrum Visuale
Text recognition, Large Language Models and societal challenges
The first part of the seminar presents a simple, modular, and reproducible approach towards text recognition (OCR/HTR) based on an attention encoder-decoder network that leverages transfer learning to enable text recognition in cases where training data is scarce. OCR/HTR technology provides Large Language Models (LLMs) with textual data in a variety of formats, and this data can be used to train LLMs on a variety of tasks, such as text classification, summarizations, and Q&A. In the second part of the seminar, I will present LLMs, the societal challenges and a discussion on potential future prospects.
Ekta Vats
2023-12-14 15:15-16:00
Ångström room 104150
Extra seminar session
Bioimage analysis to quantify interactions from molecules to networks

We develop and apply computational methods for automated image and network analysis, aiming towards a quantitative understanding of cellular interactions in tissues. In the first part of the talk, examples will be shown where deep learning applied to high-resolution microscopy helps to interpret protein structure and synaptic organization in electron microscopy, improves resolution and image acquisition for single-molecule localization, and automatically correlates light- and electron microscopic images. On a larger scale, we are interested in imaging and analyzing network structures and cell-matrix dynamics in 3D tissue culture as well as in-vivo. Based on the quantitative information from such large-scale image data, we develop theoretical models and generate new hypotheses on the structure and function of complex biological networks on all scales.
The seminar is organized in connection with Ankit Gupta’s Ph.D. thesis defense (Friday, Dec 15), when Prof. Dr. Philip Kollmannsberger will act as the faculty opponent.
Prof. Dr. Philip Kollmannsberger, Heinrich-Heine-Universität Düsseldorf
2023-12-18 14:15-15:00
Theatrum Visuale
Participating at SciFest - why and how?
Cancelled (Postponed)
Ida-Maria Sintorn
Updated  2024-02-22 10:48:36 by Natasa Sladoje.