Full Day Practicals (8 September 2023)

1. CT and MRI in radiology: what the patients need

Abstract

The project provides an overview of CT and MR image acquisition and post-processing in radiology. The students review the basic requirements for CT and MR imaging principles in patients, learn about the different imaging protocols and their indications, acquire CT and MR imaging data sets, and review image post-processing issues.

Agenda

Morning Session (9:00 – 12:00)

• Visit the CT scanner and reading room of the institute and recapitulate CT imaging fundamentals, with particular attention to patients’ and clinicians’ needs.
• Conduct a CT scan of an anthropomorphic phantom.
• Discuss radiation dose issues in CT imaging and current approaches for minimizing radiation applied to the patients.
• Review post-processing issues of several case examples using various post-processing software tools.

Lunch (12:00 – 13:00)

Afternoon Session (13:00 – 16:00)

• Conduct an MR imaging study on a healthy volunteer and acquire images of the knee/spine.
• Visit the clinical MR scanner and reading room of the department and recapitulate MR imaging fundamentals, with particular attention to patients’ and clinicians’ needs.
• Review MR images of case examples and discussing the issue of post-processing in MR imaging.
• Summary and conclusion of the practical with discussion.

Location

Meeting place: Main Entrance University Hospital Zurich, A floor, Rämistrasse 100, 8091 Zurich (To get there take tram no. 6, tram no. 9 or tram no. 10 and disembark at stop ETH/Universitätsspital; enter University Hospital building to be picked up

Morning session: CT scanner and reading room, Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, Rämistrasse 100, C floor, OPS C 11

Afternoon session: MR centre south, Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, Rämistrasse 100, V floor

2. Machine learning for medical image analysis

Abstract

In recent years, machine learning has gained a lot of attention after showing promising results in various applications from many different research fields. The fast development of machine learning has paved great potential for healthcare. Analysis of medical images provides essential information for the diagnostics, modelling, prediction of medical events, and personalised healthcare: which usually rely on the knowledge and experience of the clinicians. Advancements of the data-driven machine learning methods have great promise for computer-aided medical image analysis, making healthcare more cost-efficient, accurate and accessible.

In this workshop, students will be presented with an overview of learning methods in medical image analysis, starting from basic machine learning concepts to creating and training models with medical imaging data for different purposes. This workshop will consist of practical hands-on sessions held on python notebooks with interactive code snippets. After the introductory material on machine learning and deep learning, state-of-the-art applications in medical imaging will be presented. Finally, students will have the chance to implement and present their solution to a medical problem with readily available datasets from different imaging modalities such as magnetic resonance imaging and computer tomography.

Agenda

Morning Session (9:00 – 12:00)

• Starting from the basic principles of machine learning, students will learn about the workflows of regression and classification in an interactive python session.
• Continuing, students will go from coding simple commands to building a complete neural network architecture using Python.

Lunch (12:00 – 13:00)

Afternoon Session (13:00 – 16:00)

• Students will work in groups during the afternoon to solve a chosen problem using machine learning techniques. A list of proposed projects will be presented during the morning session. However, students can also come with a problem and data from their research.
• To wrap up, groups will present their problem and solution in a short 5 min presentation.

Location

Meeting place: ETZ D 96-1, Gloriastrasse 35, 8092 Zürich (Campus ETH Central)

3. Infrared tomography: last generation near infrared tomographic techniques for detection of ischemic lesions in preterm infants, and for tumor imaging

Abstract

The project presents state of the art photon detectors and reconstruction methods within the field of near infrared optical tomography. The experimental setup for preterm brain tomography consists of a supercontinuum laser source and a special time of flight (ToF) camera capable of detecting single photons and their arrival time. Parallel image reconstruction techniques based on GPUs and supercomputers, as well as AI based approaches, will be discussed and the students will have the opportunity to run some examples on our hardware. Time-resolved near infrared optical tomography (NIROT) system with dedicated sensor and head phantoms will allow hands on experience on a typical imaging problem.

Agenda

Morning Session (9:00 – 12:00)

• Talk providing basic information about the clinical background, diffusion model for light propagation in tissue, inverse problem formulation, super continuum laser and SPAD arrays, strategies to reduce the ill-posed nature of inverse problem, time resolved measurements and large data sets.
• Hands-on session: NIROT experiment and reconstruction basis.
• Talk providing overview of AI based solutions to the reconstruction problem, considering speed and quality improvement, and artefact reduction.
• Exercise: Short conceptual questions will be presented by the speaker to the the students to be solved by teams in a contest mode. The correct answers will be explained and discussed.

Lunch (12:00 – 13:00)

Afternoon Session (13:00 – 16:00)

• Hands-on session on tissue-to-light interaction:
  • measurement of tissue oxygenation and dynamic changes of oxygen level by NIRS
  • optical principles and signal analysis of pulse-oximetry
  • Monte Carlo simulation of photon propagation through tissue
• Exercise: Short conceptual questions will be presented by the speaker to the students to be solved by teams in a contest mode. The correct answers will be explained and discussed.

Location

Meeting place: Wagistrasse 12, 8952 Schlieren: take Tram 2 (red) at 8:25 from Zurich Stauffacher, get off at tram stop Gasometerbrücke, meeting at main entrance.

4. Correlative light and electron microscopy on cryo-sections

Abstract

To fully understand the function of a protein, we need to know where it is expressed in a particular organ and subsequently in which subcellular compartment of the cell. Light microscopy methods image samples with a large field of view. Super-resolution methods allow now to determine the location of a protein at the nanoscale. However, the structures, which are not fluorescent, remain invisible to us, thus we lack context information about where the proteins are located. Electron microscopy provides a high context information but protein localization is often difficult to perform and only one or two different proteins can be identified simultaneously. Correlative light and electron microscopy combines the advantages of both techniques often at the cost of very demanding experiments. Our full day practicum will explain an easy and quick correlative light and electron micros-copy method to localize proteins at subcellular resolution, all done on the very same section [1].

Agenda

Aim: Basic knowledge of correlative light and electron microscopy. Quick introduction into the tools necessary for performing correlative light and electron microscopy.

Methodology: The theoretical part will consist of explanations about the topics of the practicum helped with images and close to the tools that the students will, afterwards, use. During the practicum the students will have a first contact with sample preparation, imaging by light and electron microscopy and alignment of the images.

Theoretical part (1 hour):

• Importance of combining light and electron microscopy
• How to combine light and electron microscopy

Practical part (7 hours):

• Cutting at the ultracryotome and ultratome
• Immunofluorescence experiments
• Imaging at light microscopes
• Imaging at electron microscopes
• Alignment of light and electron microscopy images

Location

Meeting place: Center for Microscopy and Image Analysis (ZMB), Y42, Stock H, room 92, University of Zurich (Irchel), Winterthurerstrasse 190

References

[1] Mateos, J. M., Barmettler, G., Doehner, J., Ojeda Naharros, I., Guhl, B., Neuhauss, S. C., Kaech, A., Bachmann-Gagescu, R., Ziegler, U. Correlative Super-resolution and Electron Microscopy to Resolve Protein Localization in Zebrafish Retina. J. Vis. Exp. (129), e56113 (2017).

5. Ultrasound: nothing to hear, much to see

Abstract

The workshop introduces the students to cardiac ultrasound techniques. The students get to know the basic principles of ultrasound and some advanced uses, such as assessment of ventricular function. The students study the methodological background of different echocardiographic applications, their power and limitations, as well as the scientific methods of comparison of different modalities. In a hands-on session the student can perform a basic echocardiographic examination.

Agenda

Morning Session 1 (9:00 – 11:00)

• Introductory talk I: “Ultrasound principle”
  • Ultrasound properties, Reflection, scattering, absorption, attenuation, Mechanik Index and power. Temporal and lateral resolution, depth, gain.
  • B-Mode, speckle
  • Doppler effects including CW/PW/Color/Tissue, Artefacts, Nyquist
• Introductory Talk II: “Characteristics of the right ventricle”
  • Structure, anatomy, and physiology of the right ventricle
  • Significance in clinical practice
  • Functional imaging of the right ventricle with basic and advanced echocardiographic techniques

• The students deepen the topic in groups.
  • Group A. Learn the principles of ultrasound applications for assessing right ventricular function in the context of multimodality imaging (Discussion within the group)
  • Group B. How to compare different echocardiographic modalities in a clinical setting (Discussion within the group)

Coffee Break (11:00 - 11:30)

Morning Session 2 (11:30 – 13:00)

• Group presentations
• Wrap-up: Discussion, Questions and Answers

Lunch (13:00 - 14:00)

Hands-on Session (14.00 -16.00)

• Students perform scanning to get to know ultrasound probe handling
• Demonstration of a complete echocardiographic examination
• Students focus on ventricular function and exercise the parameters learned in the morning

References

6. Multi-color fluorescence immunolabelling of cultured cells

Abstract

The project’s aim is to demonstrate a typical fluorescence immunostaining workflow. Labelling cellular proteins and structures with fluorescence tagged antibodies belongs to the most common sample preparations in fluorescence microscopy. The hands-on cover fixation and permealization of osteosarcoma cells (U2OS) as well as subsequent blocking and fluorescence labelling steps. Finally, the double labelled cells (labelled structures: golgi, microtubules) will be analyzed via confocal microscopy.

Agenda

Morning Session (9:00 – 12:00)

• Introduction into the staining protocol
• Fixation, permeabilization, blocking, 1st antibody incubation

Lunch (12:00 - 13:00)

Afternoon Session (13:00 – 17:00)

• 2nd antibody incubation, mounting
• Staining quality check (confocal microscopy)

Location

Meeting place: HPM2 building (ETH Hönggerberg Campus), C floor, room: C43

7. MRI and image analysis of human blood flow

Abstract

MRI allows for a quantitative assessment of blood flow in the circulatory system. Besides estimation of blood velocities, flow and stroke volume it offers the ability to analyze flow patterns in e.g. the aorta or the heart’s chambers and to derive sheer stresses in vessel walls.

During this daylong project we will recapitulate the basics of MRI flow measurement and illustrate pitfalls and limitations. The lecture is followed by a hands-on imaging session at the hospital’s MR center where we will collect quantitative flow data from the participants (voluntary). We will spend the second half of the day analyzing the acquired data, counter act imaging artifacts and visualize blood flow within the aorta.

Agenda

Morning Session (9:00 – 12:00)

• Lecture (approx. 1h)
• Imaging in the USZ MR-center

Lunch (12:00 - 13:00)

Afternoon Session (13:00 – 16:00)

• Data Analysis

Location

Meeting point: ETZ (ETH main campus), room: F91

8. Paleoradiology - Impact and challenges of diagnostic imaging of ancient human remains

Abstract

The exploration of human health and disease evolution gains significant depth through the study of ancient human remains. Paleoradiology, with its focus on the radiological examination of such remains - ranging from skeletal material to mummified soft tissue - provides unparalleled insights into the disease patterns of past civilizations.

In paleoradiology, taphonomic processes present recurring challenges, especially evident in mummies. These processes often result in alterations in shape, tissue density, and the manifestation of cracks and cavities. The selection of an imaging modality for paleoradiological studies is heavily influenced by both the specimen's preservation state and its transport feasibility. Consequently, researchers frequently resort to conventional X-rays or portable cone-beam X-ray tomography. However, the preferred choices are clinical computed tomography (CT) and magnetic resonance imaging (MRI) when circumstances allow.

This course invites you on an intriguing journey into Paleoradiology, offering a comprehensive theoretical introduction enriched with enlightening case studies. Participants will also gain practical experience processing paleoradiological imaging data using Horos software (with RadiAnt as an alternative for PC users).

Embark on an expedition into our shared past and unveil the hidden secrets in ancient human remains. Are you ready to join us?

Agenda

Morning Session 1 (9:00 – 10:30)

• Introduction and Case Studies in Paleoradiology.

Coffee Break (10:30 – 11:00)

Morning Session 2 (11:00 – 12:00)

• Demonstration: CT-Scan to 3D Model (image processing and segmentation in Horos).

Lunch (12:00 – 13:00)

Afternoon Session (13:00 – 16:00)

• Hands-on training: Image segmentation in Horos, on the example of funerary amulets contained in an ancient Egyptian mummy.

Location

Meeting place: University Zürich, Irchel Campus, Building Y42, room: G66 (secretary of IEM)

9. Optoacoustic imaging

Abstract

Optoacoustic (photoacoustic) imaging emerged in the early 2000s as a new modality for non-invasive visualization of living organisms. It has experienced massive technological developments in the last two decades and is growingly being used as a biomedical research tool. Current efforts are further directed towards its clinical translation, with initial trials offering promising results. State-of-the-art optoacoustic systems provide a unique five-dimensional (real-time multi-spectral three-dimensional) imaging performance. Also, microscopic to macroscopic spatial scales as well as the dynamics of fast to longitudinal processes can be covered. Overall, these unique features open exciting research opportunities in both physical and life sciences.

In the project day, we will introduce the basic principles of the optoacoustic technology. We will cover theoretical aspects as well as well as a practical session with state-of-the-art systems. Students will have the opportunity to acquire hands-on experience in this new technology by designing and performing phantom experiments.

Agenda

Morning Session 1 (9:00 – 10:00)

• Presentation of the project day and basic principles of the optoacoustic technology

Morning Session 2 (10:00 – 12:00)

• Introduction to optoacoustic experiments covering:
  • Laser and ultrasound sensor technology
  • Acquisition of optoacoustic signals
  • Reconstruction of optoacoustic images

Lunch (12:00 – 13:00)

Afternoon Session 1 (13:00 – 14:30)

• Real-time optoacoustic imaging covering:
  • Manufacturing of phantoms
  • Three-dimensional visualization of phantoms

Afternoon Session 2 (14:30 – 16:00)

• Hands-on training

Location

Meeting point: Irchel Campus, building Y17, lab H13

10. In vivo calcium imaging in the mouse brain

Abstract

The aim of this lab course is to get first hands-on experience with in vivo calcium imaging. Following an introduction, and a presentation by a PhD student on a current project, participants will visit the labs, where the microscope components will be briefly explained.

Then participants will have the opportunity to observe techniques of measuring neuronal activity in anesthetized mice, as well as in awake, behaving animals. The participants will also have the opportunity to do hands on experiments with two-photon lasers on an optical bench, demonstrating the operating principles of two-photon microscopy.

Agenda

Morning Session 1 (9:00 – 10:00)

• Welcome + Introduction in vivo calcium imaging

Morning Session 2 (10:00 – 10:30)

• PhD Student Presentation

Coffee Break (10:30 – 10:45)

Morning Session 3 (10:45 – 11:30)

• Lab tour

Lunch Break (11:30 – 12:45)

Afternoon Session 1 (12:45 – 13:45)

• Rotation Part 1

Afternoon Session 2 (13:45 – 14:45)

• Rotation Part 2

Afternoon Session 3 (14:45 – 15:45)

• Rotation Part 3

Afternoon Session 4 (15:45 – 16:00)

• Wrap up

Location

Meeting place: University of Zurich, Brain Research Institute, Irchel Campus, Building Y55, room H12, Winterthurerstr. 190, 8057 Zürich

11. From visual activation to lie detection? An introduction to functional MRI

Abstract

The project follows up on the popular idea of using functional Magnetic Resonance Imaging (fMRI) as a lie detector [1-4]. In order to get an understanding of the feasibility and problems faced by fMRI lie detection students will first get an introduction into the fundamentals of fMRI. The students will then conduct a simple fMRI data analysis of a participant that we scan in the morning of the practical day. Based on this experience, they will then read papers and discuss the relevant literature on fMRI lie detection. Finally, students will apply what they learnt by developing an fMRI lie detection experiment.

Agenda

Morning Session (9:00 – 12:30)

• Introduction and fMRI scan of a healthy volunteer [MR Centre]
• Review of the fundamentals of fMRI and introduction to the Blood Oxygenation Level Dependent (BOLD) contrast [1]
• Introduction to fMRI data analysis
• Preparation for data analysis and download of data
• Split into groups. Run an fMRI analysis of the data measured in the morning

Lunch (12:30-13:30)

Afternoon Session (13:30 – 16:00)

• Each group prepares a brief summary of one of the articles [2-4] and presents it to the other groups. Check the internet for popular news on fMRI lie detection
• Discuss the limitations, potential shortcomings and possible solutions
• Design a deception experiment involving motor and visual block paradigms
• Present and discuss a summary of your experimental design

Location

Meeting place: MR Centre, University Hospital, Raemistrasse 100, V floor, Imaging room: TBA, tutorial room: ETZ G91

Notes

Students should have Matlab installed on their personal computers to run the analysis with SPM.

References

[1] What we can do and what we cannot do with fMRI. NK Logothetis. Nature (2008)

[2] Lying in the scanner: Covert countermeasures disrupt deception detection by functional magnetic resonance imaging. G Ganis et al. NeuroImage (2011)

[3] Prospects of functional magnetic resonance imaging as lie detector. E Rusconi et al. Front Hum Neurosci. (2013)

[4] Functional MRI-based lie detection: scientific and societal challenges. MJ Farah et al. Nat. Rev. Neurosci. (2014)

12. Postmortem imaging technology in forensic medicine

Abstract

Over the last 20 years, postmortem computed tomography and magnetic resonance imaging, as well as optical imaging for three-dimensional (3D) surface documentation, have found their way into forensic medicine under the name Virtopsy® or virtual autopsy. These imaging techniques are used to support and complement a postmortem examination and allows for non-invasive documentation of forensically relevant findings prior to autopsy.

The postmortem imaging technology research group at the Zurich Institute of Forensic Medicine has focused on quantitative computed tomography (qCT), postmortem magnetic resonance spectroscopy (pmMRS), and 3D whole body multispectral imaging (MSI), as well as radiological wound ballistics (RWB). Bone mineral density is measured by qCT and serves as the basis for finite element analyses, which are planned to be included in the assessment of blunt fracture injuries in the future. The pmMRS measurements are used to detect metabolite concentrations in brain, which allows identification of fatal metabolic disorders. Furthermore, it is sensitive to elevated alcohol concentrations. 3D whole body MSI includes not only the documentation of the body surface within the visible part of the electromagnetic spectrum, but also the detection of findings and latent traces beyond that. Near-infrared imaging is used to highlight bleedings in or under the skin, while ultraviolet illumination is used for fluorescence imaging to detect latent traces on the body surface. In RWB, imaging techniques are used not only to image gunshot wounds, but also to classify lodged projectiles based on shape and material. In addition, ballistic simulants of gelatin or soap are studied in relation to the expansion of the temporary wound cavity.

Agenda

Morning Session 1 (8:30 – 9:00)

• Welcome with coffee and “Gipfeli”

Morning Session 2 (9:00 – 10:30)

• Quantitative computed tomography
  • Starting from the basic principles of qCT, students will attend the CT scan and then learn how to convert the water-calibrated CT attenuation values to bone mineral density as part of a hands-on session.

Morning Session 3 (10:30 – 12:00)

• Postmortem magnetic resonance spectroscopy
  • After a brief introduction to the basic principles of magnetic resonance spectroscopy, students will see how an pmMRS measurement is performed and interpret the metabolite spectrum in a hands-on session.

Lunch break with sandwiches and beverages (12:00 - 13:00)

Afternoon Session 1 (13:00 – 14:30)

• Multispectral imaging
  • Students are introduced to taking images using light sources with different wavelengths as well as appropriate camera filters for near infrared and ultra violet induced fluorescence imaging with a full spectrum camera.

Afternoon Session 2 (14:30 – 16:00)

• Radiologic wound ballistics
  • After an introduction to the topic, students will perform a multi-energy CT examination of projectiles in a tissue phantom and subsequently learn to distinguish them using extended CT scale reconstructions and the dual-energy index as part of a hands-on session.

Location

Meeting place: Main Entrance - Institut für Rechstmedizin Zürich (Building: Y52), University of Zurich, Campus Irchel

13. Medical image classification and segmentation with deep learning

Abstract

Analysis of medical images provides crucial visual information for the modeling and prediction of patients’ events and personalized treatment: which usually rely on clinicians’ knowledge and expertise. Advances in data-driven machine learning methods, especially deep learning, have shown great promise for computer-aided medical image analysis, improving healthcare to be more accurate and accessible.

In this workshop, students will be presented with an overview of deep learning methods, starting from basic machine learning concepts to creating and training models with medical imaging data for different purposes. In light of these advances, we have designed the workshop to introduce the principal building blocks driving the deep learning revolution in health care. The workshop is intended to be interactive, hands-on and operates in a flipped classroom setup. The participants will be introduced to key concepts such as loss functions, regularizations, model architectures etc., and apply them to build their own models for seminal tasks such as image classification and segmentation. The participants would get an opportunity to explore readily available, open-source medical imaging datasets. They will build and train their own models on these datasets. Furthermore, we will try to explore model interpretability to investigate the predictions made by these models.

Requirement: Basic knowledge of Python.

Goals: Hands-on; learning fundamentals of deep learning; building deep learning models; visualization and interpretation of results

Agenda

Morning Session 1 (9:00 – 9:45)

• Introduction to machine learning
  • Starting from the basic principles of machine learning, students will learn about the workflows of regression and classification in an interactive python session

Morning Session 2 (9:45 – 10:15)

• Hands-on session
  • Students will go from coding simple commands to building a regression/classification model using Python

Morning Session 3 (10:30 – 11:00)

• A basic introduction to deep learning
  • An intuitive and basic introduction to convolutional neural networks, including convolution, optimizer and loss function

Morning Session 4 (11:30 – 12:00)

• Hands-on session
  • Continuing, students will go from coding simple commands to building a complete neural network architecture for medical image classification using Python, playing around with the learning rate, early stopping, dropout, etc.

Lunch Break (12:00-13:00)

Afternoon Session 1 (13:00 – 13:45)

• Introduction to image segmentation
  • An intuitive and basic introduction to U-shape convolutional neural networks for image segmentation, including image pre-processing, loss function, etc.

Afternoon Session 2 (13:45 – 14:15)

• Hands-on session
  • Students will go from coding simple commands to build a U-Net architecture for medical image segmentation using Python

Afternoon Session 3 (14:30 – 15:00)

• Further Introduction to image segmentation
  • Introduction to data augmentation, evaluation metric, interpreting the results, and potential issues (such as domain shift)

Afternoon Session 4 (15:00 – 16:00)

• Hands-on session
  • Students will explore how to modify architecture, change various train/val splits, perform data augmentation, and check the results

Location

Meeting place: ETZ/D/61-1 Gloriastrasse 35, 8092 Zürich (ETH Centre Campus)