Saga Helgadottir defended her PhD Thesis in Physics on June 16, 2021. Congrats!

Digital video microscopy enhanced by deep learning
Saga Helgadottir defended her PhD Thesis in Physics on June 16, 2021. Congrats!

The disputation took place at 9 a.m. digitally via Zoom. A link to the Zoom meeting was published the day before dissertation on the GU website.

Title:  Deep Learning Applications – From image analysis to medical diagnosis

Deep learning is a subcategory of machine learning and artificial intelligence. Instead of using explicit rules to perform a desired task as in standard algorithmic approaches, machine-learning algorithms autonomously learn from data to determine the rules for the task at hand. The idea of deep learning has been around since the 1950s but was for a long time limited by available computational power and amount of training data. Once overcome these problems, in recent years, deep learning has made great advances in solving various problems.

In this thesis, I show how deep learning can be applied in image analysis and medical diagnosis, while outperforming standard algorithmic methods and simpler machine-learning methods. I begin with showing that a convolutional neural network trained with simulated particle images is able to track experimental single particles, even in poor illumination conditions. I then show how this inspired the development of an all-in-one software to design, train and validate deep-learning solutions for digital microscopy, from particle tracking and characterization in 2D and 3D to the segmentation, characterization and counting of biological cells and image transformation. I show that this software package can be further used to develop a generative adversarial neural network to virtually stain brightfield images of cells, replacing the traditional chemical staining for a downstream analysis of biological features. I then move on from applications in microscopy and image analysis to show the potential of deep learning in medical diagnosis. I show that dense neural networks perform better than simpler machine-learning algorithm and the clinical standard in the diagnosis of a genetic disease and in the prediction of short- and long-term morbidity in patients with congenital-heart-disease. At last, I have shown that a neural network- powered strategy for testing and isolating individuals adapts to the parameters of a disease outbreak achieves an epidemic containment.

The interdisciplinary nature of the work in this thesis has allowed the application of new technologies developed in the field of physics to solve problems in the fields of biology and biomedicine, as well as overcoming barriers for the continued revolutionization of deep learning in microscopy.


Supervisor: Giovanni Volpe
Examiner: Bernhard Mehlig
Opponent: Carolina Wählby
Committee: Marj Tonini, Maria Garcia-Parajo, Alexander Rohrbach

Screenshots from Saga Helgadottir’s PhD Thesis defense.

PhD Opponent’s presentation.
PhD Thesis presentation: Saga Helgadottir, Giovanni Volpe (Supervisor), Raimund Feifel (GU Physics), Carolina Wählby (Opponent), Marj Tonini (Committee member), Maria Garcia-Parajo (Committee member), Måns Henningson (GU Physics Department Chair), Alexander Rohrbach (Committee member).
PhD Thesis presentation.
PhD Thesis presentation front slide.
PhD Thesis presentation content slide (1).
PhD Thesis presentation content slide (2).
PhD Thesis presentation conclusion slide.
Screenshot from the discussion (1).
Screenshot from the discussion (2).
Screenshot from the discussion (3).

Olle Fager defended his Master thesis on 15 June 2021. Congrats!

Olle Fager defended his Master thesis in MPCAS at the Chalmers University of Technology on 15 June 2021. Congrats!

Title: Real-Time Multi-Object Tracking and Segmentation with Generated Data using 3D-modelling

Multi-Object Tracking and Segmentation (MOTS) is an important branch of computer vision that has applications in many different areas. In recent developments these methods have been able to reach favorable speed-accuracy trade-offs, making them interesting for real-time applications. In this work different deep learning based MOTS methods have been investigated with the purpose of extending the DeepTrack framework with real-time MOTS capabilities. Deep learning methods rely heavily on the data on which they are trained. The collection and annotation of the data can however be very time-consuming. Therefor, a pipeline is developed and investigated that automatically produces synthetic data by utilizing 3D-modelling. The most accurate tracker achieves a MOTSA score of 94 and the tracker with the best speed-accuracy trade-off achieves a MOTSA score of 88. It is also observed that satisfactory results can be achieved in most situations with a quite general data generation pipeline, indicating that the developed pipeline could be used in different scenarios.

​Name of the master programme: MPCAS – Complex Adaptive Systems
Supervisor: Giovanni Volpe
Examiner: Giovanni Volpe, Department of Physics, University of Gothenburg
Opponent: Arianit Zeqiri and Morad Mahmoudyan

Place: Online via Zoom
Time: 15 June, 2021, 16:00

Aykut Argun defended his PhD Thesis in Physics on 14 June 2021. Congrats!

(Image by Aykut Argun)
Aykut Argun defended his Ph.D. thesis on June 14, 2021, at 2 pm CEST. Congrats!

The details of the presentation can be found below. The link to the webinar is announced on the faculty website.

Title: Thermodynamics of microscopic environments: From anomalous diffusion to heat engines.

Unlike their macroscopic counterparts, microscopic systems do not evolve deterministically due to the thermal noise becoming prominent. Such systems are subject to fluctuations that can only be studied within the framework of stochastic thermodynamics. Within the last few decades, the development of stochastic thermodynamics has lead to microscopic heat engines, nonequilibrium relations and the study of anomalous diffusion and active Brownian motion. In this thesis, I experimentally show that the non-Boltzmann statistics emerge in systems that are coupled to an active bath. These non-Boltzmann statistics that result from correlated active noise also disturb the nonequilibrium relations. Nevertheless, I show that these relations can be recovered using an effective potential approach. Next, I demonstrate an experimental realization of a microscopic heat engine. This engine is referred to as the Brownian gyrator, which is coupled to two different heat baths along perpendicular directions. I show that when confined into an elliptical trap that is not aligned with the temperature anisotropy, the Brownian particle is subject to a torque due to the symmetry breaking. This torque creates an autonomous engine whose direction and amplitude can be controlled by tuning the alignment of the elliptical trap. Then, I show that the force fields acting on Brownian particles can be calibrated using a data-driven method that outperforms the existing calibration methods. More importantly, I show that this method, named DeepCalib, can calibrate non-conservative and time-varying force fields that no standard calibration methods exist. Finally, I show that a similar machine-learning-based approach can be used to characterize anomalous diffusion from single trajectories. This method, named RANDI, is very versatile and performs very well in various tasks including classification, inference and segmentation of anomalous diffusion. The work presented in this thesis presents novel experiments that advance microscopic thermodynamics as well as newly developed methods that open up new possibilities in analyzing stochastic trajectories. These findings increased the scientific knowledge at the nexus between microscopic thermodynamics, anomalous diffusion, active matter and machine learning.

Supervisor: Giovanni Volpe
Co-supervisors: Joakim Stenhammar, Mattias Goksör
Examiner: Bernhard Mehlig
Opponent: Juan M. R. Parrondo
Committee: Monika Ritsch-Marte, Sabine H. L. Klapp, Édgar Roldán

Screenshots from Aykut Argun’s PhD Thesis defense.

PhD Thesis Committee, Supervisor, Co-Supervisor, Opponent, and GU Physics Department Chair.
PhD Thesis Committee, Supervisor, Opponent, and GU Physics Department Chair.
PhD Opponent presentation.
PhD Thesis presentation starts.
PhD Thesis front slide.
PhD Thesis content slide.
PhD Thesis final acknowledgment slide.
PhD Thesis final acknowledgment slide.

Agaton Fransson defended his Master thesis on 4 June 2021. Congrats!

Agaton Fransson defended his Master thesis in MPCAS at the Chalmers University of Technology on 4 June 2021. Congrats!

Big plankton tracked by network-based software in a sample of big (Strombidium arenicola) and small plankton (Rhodomonas baltica). (Image by Agaton Fransson)
Title: Tracking plankton using neural networks trained on simulated images

Softwares to track particles often use algorithmic approaches to detect particles and to create tracks using the found positions, requiring human fine-tuning of parameters to achieve sought-for results. This can be time consuming and difficult, while also creating opportunities for human error and bias. With the developments of computational power and machine learning techniques such as deep learning, data driven approaches have made their way into many fields of science. Barriers preventing advances of such methods are the lack of available training data within a field and the level of proficiency required to create custom machine learning solutions. DeepTrack 2.0 is a software providing us with means to simulate digital microscopy images, build and train neural networks such as U-nets. In this paper DeepTrack 2.0 is utilized and built on to fit the needs of marine biologists when tracking plankton. Here I show that DeepTrack 2.0 provides us with the tools necessary to detect and track different types of plankton filmed in a variety of conditions with performance on par with and with the potential to outperform conventional tracking softwares. I also show that for plankton in a messy environment moving uniformly a network trained to detect motion rather than a shape proves more successful. These results demonstrate the versatility of deep learning methods and the potential of training networks on simulations for applications on real data, as is the case for marine biologists studying plankton. They also show the impact the structure of the training data has on the nature of the network.

​Name of the master programme: MPCAS – Complex Adaptive Systems
Supervisor: Giovanni Volpe, Daniel Midtvedt
Examiner: Giovanni Volpe, Department of Physics, University of Gothenburg
Opponent: Kevin Rylander

Place: Online via Zoom
Time: 4 June, 2021, 15:00

Kevin Andersson and Eric Lindgren defended their Master thesis on 2 June 2021. Congrats!

Kevin Andersson and Eric Lindgren defended their Master thesis on 2 June 2021. Congrats!

Title: Saliency mapping of RS-fMRI data in GCNs for sex and brain age prediction
Subtitle: Identifying important functional brain networks using explainability in Graph Convolutional Networks

Insights into how biological sex and healthy ageing affects the human brain are important for an increased understanding of the brain. Healthy ageing insights are also useful for clinical applications, for instance in identifying unhealthy ageing due to neurodegenerative disease. To this end, several studies in the last few years have used machine learning methods on neuroscientific data to predict subject sex and brain age. One particularly interesting approach has been to represent functionally connected networks in the brain as graphs, and apply Graph Convolutional Networks (GCNs). To investigate which functional brain networks are connected with sex and age, we develop and analyse GCN-based models that predicts sex and age from resting-state fMRI data. The analysis of the models is done using saliency mapping techniques which gives insight into what functional brain networks in the data are relevant for the predictions. With this approach, we obtain a sex prediction accuracy of up to 79% and an age prediction MAE of 5.9 years. Furthermore, we find indications that the Sensory Motor Network and the cerebellum are among the more important functional brain networks for predicting sex and brain age.

​Master programme: Physics
Supervisor: Alice Neimante Diemante (Syntronic AB)
Examiner: Giovanni Volpe, Department of Physics, University of Gothenburg
Opponent: Rasmus Svensson

Place: Online via Zoom
Time: 2 June, 2021, 16:00

Pernilla Huynh, Hannes Johansson, Olof Lind Stefansson, Oskar More Arvidsson, William Olsson, Filip Sterner defended their Bachelor Thesis at Chalmers University of Technology on 27 May 2021. Congrats!

Pernilla Huynh, Hannes Johansson, Olof Lind Stefansson, Oskar More Arvidsson, William Olsson, Filip Sterner defended their Bachelor Thesis at Chalmers University of Technology on 27 May 2021. Congrats!

Title: Kollektiva beteenden hos aktiva agenter i komplexa system
Topologiska interaktioner och skillnader med tidsfördröjning

Studierna av aktiva agenters beteende i komplexa system har rönt stort intresse den senaste tiden. Dels då det är en passande modell för att beskriva många biologiska system, men också för deras för potentiella tillämpningar. Syftet med studien är att undersöka skillnader mellan beteendet för agenter som interagerar genom metriska interaktioner (det vill säga beroende av det metriska avståndet mellan agenter) och agenter som interagerar genom topologiska interaktioner (beroende på ett bestämt antal närliggande agenter). Framförallt studeras hur de båda modellerna förändras då en tidsfördröjning mellan agenternas uppfattning och reaktion introduceras. Undersökningen utförs genom analys av tre olika komplexa system: (1) agenter som rör sig bland periodiska hinder; (2) agenter som följer en ledare genom en labyrint; och (3) aktiva agenters interaktioner med passiva agenter. Utifrån de resultat som erhålls kan det framgångsrikt observeras skillnader i interaktionerna hos den topologiska modellen gentemot den metriska modellen: de aktiva agenterna kan i den topologiska modellen interagera mer med varandra trots periodiska hinder, en större andel agenter tar sig genom labyrinten och klusterbildningen är oftast lägre i systemet med ett lågt antal passiva agenter. Resultaten tyder också på att den topologiska interaktionen i många fall är mindre känslig för tidsfördröjning.

Studies of the behavior of active agents in complex systems have received a lot of interest recently. This interest derives from the fact that active agents provide an ideal model to describe many biological systems and also because of their potential applications. The purpose of this study is to explore which differences there are between the behaviour of agents that interact through metric interactions (i.e. depending on the metric distance between agents) and those of agents that interact through topological interactions (i.e. depending on a certain number of surrounding agents). This report also discusses how the interaction models for active agents change when a time delay between sensing and acting is introduced. These investigations were made by analyzing three different complex environments: (1) agents moving in the presence of periodic obstacles; (2) agents following a leader in a maze; and (3) active agents interacting with passive agents. Based on the results obtained from this study, we could successfully observe differences in the topological interaction compared to the metric model: the active agents interacted more frequently with each other when using the topological model despite periodic obstacles, a larger proportion of agents managed to pass through the maze and the cluster formation was usually smaller in the system with a low number of passive agents. The result also shows that the topological interaction is less sensitive to time delay.

Supervisor: Giovanni Volpe, Department of Physics, University of Gothenburg
Examiner: Lena Falk, Department of Physics, Chalmers University of Technology
Time: 27 May, 2021

Gustaf Sjösten defended his Master thesis on 17 May 2021. Congrats!

Gustaf Sjösten defended his Master thesis in MPCAS at the Chalmers University of Technology on 17 May 2021. Congrats!

Artistic rendering of a light source illumination a single biomolecule and scattered photons. (Image by Gustaf Sjösten, inspired by an image created by Barbora Spackova)
Title: Deep Learning for Nanofluidic Scattering Microscopy

A novel technique for label-free, real-time characterization of single biomolecules called Nanofluidic Scatter Microscopy (NSM) has recently been developed by the Langhammer research group at Chalmers. We have created a machine learning (ML) framework consisting of deep convolutional neural networks such as U-nets, ResNets and YOLO in order to characterize single biomolecules through kymographs collected through NSM, as an alternative approach to a standard data analysis method (SA). As a laser irradiates visible light onto single biomolecules freely diffusing in solution inside nanofluidic channels, the biomolecule and the nanochannel scatter light coherently into the collection optics, such that the nanochannels improve the optical contrast of the imaged biomolecules by several orders of magnitude. A video of the total scattering intensity is then recorded with a high frame rate camera (capturing 200 fps) in order to capture the movement of the molecules as well as the optical contrast of the biomolecules with respect to the nanochannel. From the movement of one single biomolecule, it is possible to predict its diffusion constant, which can then be used to infer the hydrodynamic radius of the biomolecule. Additionally, the predicted optical contrast of one single biomolecule can in turn be used to infer its molecular weight. From the combination of hydrodynamic radius and molecular weight, information about the conformal state of single biomolecules can be inferred. In this thesis, we show that the ML approach yields results comparable to the SA which was developed independently of the ML technique for biomolecules in the weight span 66-669 kDa, and we also show that the ML technique is superior to the SA in other regards, such as computational speed and potential to characterize smaller molecules. The results of the data analysis performed with the ML framework will also make an appearance in the first paper on the NSM technique which has been submitted for publication and is currently under review.

​Name of the master programme: MPCAS – Complex Adaptive Systems
Supervisor: Giovanni Volpe, Daniel Midtvedt
Examiner: Giovanni Volpe, Department of Physics, University of Gothenburg
Opponent: Anton Jansson

Place: Online via Zoom
Time: 17 May, 2021, 16:00

Nils Jacobson defended his Master thesis on 16 February 2021. Congrats!

Nils Jacobson defended his Master thesis in MPCAS at the Chalmers University of Technology on 16 February 2021. Congrats!

Screenshot of Nils Jacobson’s Master Thesis defence.
Title: Vascular Bifurcation Detection in Cerebral CT Angiography Using CNN and Frangi Filters

Segmentation and feature extraction are important tools for analysing and visualizing information in medical image data, particularly in vascular image data which relates to widely spread vascular diseases. Vessel segmentation is extensively featured in research, recently adapting trends in deep learning image processing. This paper aims to develop a vessel bifurcation detection method to support a seed point based segmentation approach. The suggested approach is a combination of classification, with a convolutional neural network (DenseNet), local vessel segmentation, with Frangi filters, and 3D morphological skeletonization. A small data set is produced for network training and evaluation. Results indicate a high classification accuracy which filters problematic samples for the Frangi filter. Thus the combination is able to suggest quality branch seed points under most circumstances. Next step would be to expand the data set to enable further optimization and more rigid evaluation. In any case a combination of a high performance classifier followed by qualitative assessment of local samples show potential.​

​Name of the master programme: MPCAS – Complex Adaptive Systems
Supervisor: Jonna Hellström and Giovanni Volpe
Examiner: Giovanni Volpe, Department of Physics, University of Gothenburg
Opponent: Eva Škvor

Place: Online via Zoom
Time: 16 February, 2021, 16:00

Link: Master thesis presentation Nils Jacobson

Falko Schmidt defended his PhD Thesis in Physics on 15 January 2021. Congrats!

Falko Schmidt defended his PhD Thesis in Physics on Friday, 15 January 2021. Congrats!

The disputation took place at 9 a.m., in PJ salen, Fysikgården.
Falko Schmidt’s opponent, Peer Fischer, gave an introductory presentation with title “Microswimmers and motile active matter”.


From Falko Schmidt’s PhD Thesis.
Title: Active Matter in a Critical State: From passive building blocks to active molecules, engines and droplets

The motion of microscopic objects is strongly affected by their surrounding environment. In quiescent liquids, motion is reduced to random fluctuations known as Brownian motion. Nevertheless, microorganisms have been able to develop mechanisms to generate active motion. This has inspired researchers to understand and artificially replicate active motion. Now, the field of active matter has developed into a multi-disciplinary field, with researchers developing artificial microswimmers, producing miniaturized versions of heat engines and showing that individual colloids self-assemble into larger microstructures. This thesis taps into the development of artificial microscopic and nanoscopic systems and demonstrates that passive building blocks such as colloids are transformed into active molecules, engines and active droplets that display a rich set of motions. This is achieved by combining optical manipulation with a phase-separating environment consisting of a critical binary mixture. I first show how simple absorbing particles are transformed into fast rotating microengines using optical tweezers, and how this principle can be scaled down to nanoscopic particles. Transitioning then from single particles to self-assembled modular swimmers, such colloidal molecules exhibit diverse behaviour such as propulsion, orbital rotation and spinning, and whose formation process I can control with periodic illumination. To characterize the molecules dynamics better, I introduce a machine-learning algorithm to determine the anomalous exponent of trajectories and to identify changes in a trajectory’s behaviour. Towards understanding the behaviour of larger microstructures, I then investigate the interaction of colloidal molecules with their phase-separating environment and observe a two-fold coupling between the induced liquid droplets and their immersed colloids. With the help of simulations I gain a better physical picture and can further analyse the molecules’ and droplets’ emergence and growth dynamics. At last, I show that fluctuation-induced forces can solve current limitations in microfabrication due to stiction, enabling a further development of the field towards smaller and more stable nanostructures required for nowadays adaptive functional materials. The insights gained from this research mark the path towards a new generation of design principles, e.g., for the construction of flexible micromotors, tunable micromembranes and drug delivery in health care applications.

Benjamin Midtvedt defended his Master Thesis on June 15, 2020. Congrats!

Benjamin Midtvedt defended his Master Thesis in Engineering Mathematics and Computer Science at Chalmers University of Technology on 15 June 2020. Congrats!

Screenshot of Benjamin Midtvedt’s Master Thesis defence.
Title: DeepTrack: A comprehensive deep learning framework for digital microscopy

Despite the rapid advancement of deep-learning methods for image analysis, they remain underutilized for the analysis of microscopy images. State of the art methods require expertise in deep-learning to implement, disconnecting the development of new methods from end-users. The packages that are available are typically highly specialized, challenging to reappropriate, and almost impossible to interface with other methods. Finally, training deep-learning models often requires large datasets of manually annotated images, making it prohibitively difficult to procure training data that accurately represents the problem.

DeepTrack is a deep-learning framework targeting optical microscopy, designed to account for each of these issues. Firstly, it is packaged with an easy-to-use graphical user interface, solving standard microscopy problems with no required programming experience. Secondly, it bypasses the need for manually annotated experimental data by providing a comprehensive programming API for creating representative synthetic data, designed to exactly suit the problem. DeepTrack creates physical simulations of samples described by refractive index or fluorophore distributions, using fully customizable optical systems. To accurately represent the data to be analyzed, DeepTrack supports arbitrary optical aberration and experimental noise. Thirdly, many standard deep-learning methods are packaged with DeepTrack, including architectures such as U-NET, and regularization techniques such as augmentations, decreasing the barrier to entry. Finally, the framework is fully modular and easily extendable to implement new methods, providing both longevity and a centralized foundation to deploy new deep-learning solutions.

We demonstrate the versatility of DeepTrack by training networks to solve a broad range of common microscopy problems, including particle tracking, cell-counting in dense biological samples, multi-particle 3-dimensional tracking, and cell segmentation and classification.

Master Programme: Engineering Mathematics and Computer Science
Supervisor: Giovanni Volpe
Examiner: Giovanni Volpe
Opponents: Aykut Argun and Saga Helgadóttir

Time: 15 June 2020, 16:00
Place: Online via Zoom