Finished Projects
Runtime of the project: 01.05.2019 - 31.12.2021
Applicants:
Prof. Mazda Farshad (University of Zurich)
Prof. Jess Snedeker (University of Zurich/ETH Zurich)
PD Dr. Philipp Fürnstahl (University of Zurich)
Christian Pfirrmann (University of Zurich)
Prof. Mirko Meboldt (ETH Zurich)
Prof. Ender Konukoglu (ETH Zurich)
Prof. Orcun Göksel (ETH Zurich)
Prof. Luca Regli (University of Zurich)
Lennard Stieglitz (University of Zurich)
Marco Senteler (University of Zurich/ETH Zurich)
The HMZ Flagship Project SURGENT (SURGEON ENHANCING TECHNOLOGIES) aims to establish new standards of patient specific planning and execution of precision spinal and neurosurgery. We will create, integrate, optimize, and clinically validate state-of-the-art technologies for surgical skill augmentation, embedding these tools within efficient and economical clinical workflows
that benefit the patient.
The scientific foundations of the project are:
- Image data science for mining content-rich clinical imaging datasets. We aim to extract anatomical features and measures of tissue quality to feed interactive “maps” of the surgical landscape. Cutting edge analytical methods will be both developed and exploited.
- Advanced biophysical modelling and simulation frameworks for optimal planning of patientspecific surgical parameters and/or prediction of functional outcomes.
- Augmented reality (AR) systems that efficiently and effectively support intraoperative navigation – enhancing the surgeon’s perception through timely and appropriately delivered visual and auditory information.
- Artificial intelligence (AI) driven optimization of the human computer interface (HCI). Through
quantitative online monitoring and detailed analysis of surgeon behaviour during surgery, the
HCI can be dynamically adapted to ensure the effective feed of relevant information based
upon surgical context.
This highly interdisciplinary project stands at the cross-roads of surgery, clinical radiology, deep mining of 4D datasets, real-time image processing, clinical biomechanics and biophysics, computer assisted surgical planning and navigation, human behavioural analysis, human computer interfaces and augmented reality technologies. It will converge these fields for targeted development and validation of tools that revolutionize planning and execution of spine and neurosurgery.
The ambitions of the project are both visionary and feasible:
A) we will extend first-generation platforms for patient-specific modelling to incorporate high-value information that is currently hidden within standard clinical imaging datasets – information such as tissue quality and underlying disease states;
B) we will develop cutting edge augmented reality systems that lay a technical foundation for intraoperative guidance,
C) finally, we will develop frameworks for continuous improvement of the AR human computer interface, enabled through quantitative analysis of intraoperative surgeon behaviour (eye and tool tracking).
Overall this project targets achievable and meaningful advances across a range of disciplines that will converge to yield true breakthroughs in surgical planning and execution. We will demonstrate the benefits of these breakthroughs in both preclinical and clinical trials – creating value for the patient, and for Zurich as a hub of excellence in medical research and clinical care.
Runtime of the project: 01.08.2019 - 31.01.2022
Main Applicant: Stieltjes Bram (University Hospital Basel)
Co-Applicants:
Prof. Adrien Depeursinge (University of Applied Sciences Western Switzerland Sierre)
Prof. Henning Müller (University of Applied Sciences Western Switzerland)
Prof. Ender Konukoglu (ETH Zürich)
Prof. Hatem Alkadhi (University Hospital Zurich)
Clarisse Dromain (University Hospital of Lausanne)
Prof. Hendrik von Tengg-Kobligk (Inselspital Bern)
The inclusion of imaging studies and quantitative image biomarkers into clinical data analysis, as for example in radiomic studies requires the basic imaging data to be understood well, so the extracted quantitative biomarkers are comparable. Raw image data (such as CT, Computed Tomography or MRI, Magnetic Resonance Imaging) can have many variations in the imaging process, stemming from differences in machine manufacturer, internal software versions, image acquisition protocols, image reconstruction parameters and image resolutions that strongly influence the quantitative biomarkers extracted from the images. Studies have shown the strong differences of different CT machines in terms of basic texture characteristics using phantom studies in US hospitals. To our knowledge no study addressing these variations in image characteristics exists in Switzerland and for the CT machines used in Switzerland in the past and currently. Many of the challenges in image differences are expected to be same. Particularly for retrospective studies it needs to be made clear which images can actually be compared against each other in terms of visual characteristics, when important decisions on treatment are made based on the automatically extracted quantitative image biomarkers. Modalities such as CT are regularly calibrated but only with density-based phantoms that do not analyse texture characteristics, which play for example an important role as quantitative radiomic features.
The objectives of this project are an analysis of image diversity in Swiss University hospitals including phantom studies with 3D printed solid texture phantoms to analyse diversity in image acquisition and the influence of this on quantitative biomarkers. Open source software tool boxes and available phantom data sets will be produced and should help all hospitals interested in quantitative image biomarkers for research, teaching and clinical work to analyse how comparable image studies are across scanners, manufacturers and image acquisition parameters.
Runtime of the project: 01.06.2019 - 30.11.2021
Main Applicant: Prof. Christian Wolfrum (ETHZ)
Co- Applicants: Prof. Ender Konukoglu, Dr. Nicola Zamboni, Prof. Bart Deplancke (ETHZ)
Health care costs are becoming an increasing burden for societies. The three most important cost driving disease are diabetes mellitus ($ 101 billion), ischemic heart disease ($ 88 billion) and low back pain ($ 87 billion). The annual increase rate between 1996-2013 for diabetes mellitus was 6.1% and due the continuous rise in obesity in western countries (and worldwide) can be expected to increase at a similar rate or even accelerate in the coming years. The awareness of the huge impact of these complex and multifactorial disorders not only for the patient itself, but also for society has received increased interest; nevertheless, this has not led to a decline in disease progression.
Personalized medicine is commonly associated with the field of oncology. However, it has already become clear from pre-clinical and clinical studies that metabolic diseases are diverse and complex, making them ideal targets for a personalized treatment approach, especially given the fact that obesity and type 2 diabetes have complex genetic backgrounds, which so far are only incompletely understood. Moreover, the response to modern drugs such as GLP-1 agonists for the treatment of diabetes or to ACE-inhibitors for the treatment of hypertension is significantly different among different ethnic groups, pointing towards clinically relevant consequences of the genetic background. While the field is evolving rapidly, concepts to target metabolic disease on a personalized level are still uncommon.
In contrast to white adipose tissue, BAT is a thermogenic organ, which promotes energy dissipation through uncoupling of the mitochondrial proton gradient in response to mild cold exposure. This can lead to an increase in systemic metabolic fluxes and energy expenditure, which affects glucose and lipid metabolism. Multiple human trials, which have been performed since the first description of BAT in adult humans suggest that BAT activation by cold or β3-adrenergic agonists can promote moderate weight loss and improve glucose as well as lipid homeostasis making this tissue an interesting target for obesity and its associated co-morbidities. The response of individuals to cold exposure differs considerably even in relatively homogeneous groups such as healthy volunteers. This speaks in favor of a strong genetic or epigenetic component that influences differentiation and activation of BAT and is in line with pre-clinical data from rodents. In addition, this observation offers a unique opportunity for a personalized approach with β3-agonists, which are already used clinically and which could in some patients (which have a certain amount of BAT) elicit metabolic improvements and protect from cardiovascular incidents. Thus, the primary objective of this application is to lay the foundation for the development of personalized strategies to form and activate brown adipose tissue to increase systemic energy expenditure and improve metabolic control in humans. This goal will be achieved through leveraging existing cohorts of patients and volunteers, which will drive the development of three Swiss Platforms for the advancement of Personalized Medicine. Building on this infrastructure, the consortium will develop technologies, which allow the stratification of patients with varying amounts of BAT using CTbased image analysis, metabolomics profiling of blood as well as transcriptional profiling of adipose tissue biopsies. Given the large variability and the strong genetic components associated with the activation of BAT, the consortium will further integrate the different layers of data to build predictive models applicable to larger cohorts. In line with this, one main goal of the proposal is to perform gene association studies with clinical, physiological parameters using the large AT biobank cohort from Leipzig. Furthermore we will utilize the wealth of data on human BAT that will be generated during the course of this project, to elucidate the mechanism underlying BAT formation in humans and identify new strategies to induce BAT mass and function.
The field of human BAT research is still in its early days. However, it can be expected that induction of BAT will open unique opportunities for personalized therapeutic approaches to reduce and prevent metabolic disbalance including prevalent diseases such as obesity, type 2 diabetes and cardiovascular disease. Our preliminary data on transcriptional classifiers supports this notion and demonstrates that such an approach is feasible in large human trials.
Runtime of the project: 01.04.2017 - 31.03.2021
The need for high throughput radiological assessment of Magnetic Resonance Imaging (MRI) data is increasing. On one hand, health-care relies more and more on patient images and on the other hand, the number of individuals at risk who should be screened regularly for early detection increases with the ageing population. Both of these developments lead to increasing use of MRI - due to its resolution, contrast and non-ionizing nature – and put pressure on radiologists to increase assessment throughput and pay higher attention to subtle early disease effects. The main limitation of the current radiological practice in meeting the demand is the need for extensive human interaction in analyzing images. Arguably, the most viable solution is to develop algorithms that can automatically prescreen images to accelerate assessment.
Runtime of the project: 01.07.2017 - 30.06.2020
Principal Investigator: Prof. Dominik Obrist (University of Bern)
CO-PIs:
Prof. Sebastian Kozerke (ETH Zurich)
Prof. Ender Konukoglu (ETH Zurich)
The project “HPC‐PREDICT – High‐Performance Computing for the Prognosis of Adverse Aortic Events” aims at developing software toward a novel prognostic tool for diseases of the ascending aorta. For this purpose, modern 4D PC of the ascending aorta shall be combined with high‐performance forward modelling blood flow using data assimilation techniques to enhance detail and accuracy of the imaging data. The 4D PC‐MRI principle will be extended to not only encode mean velocities per voxel but also fluctuating velocities allowing to estimate the turbulent kinetic energy and the Reynolds stress tensor. In order to accommodate the added encoding dimensions within clinically feasible scan times, highly accelerated imaging concepts will be deployed in conjunction with iterative non‐linear MR image reconstruction approaches. The inherent complexity of the resulting 4D data (mean turbulent flow field in space and time) shall be reduced through deep learning algorithms which identify landmark features in the flow field to support the prognosis of adverse events of the ascending aorta.
ETH/SDSC Data Science Project
Runtime of the project: 01.12.2017 - 28.02.2020
In scientific studies and clinical applications, microscopic images provide essential observations from 2D to 4D. To extract meaningful information from these images, image analysis algorithms are critical: they facilitate quantification and enable high-throughput analysis by
automating laborious tasks for extracting measurements from the image content, such as detecting, counting and segmenting cells. Machine learning (ML) technologies are key ingredients in these algorithms. They reduce user interaction and provide robustness. Deep learning has revolutionised many computer vision tasks but its use for microscopic imaging still remains limited. In this project, we focus on, arguably the most widely-applied analysis step, segmentation of objects in microscopic images, and propose novel technologies to widen use of deep learning technologies for this task. To this end, we identify three aspects in the current state of art that need improvement:
Aim 1. Reducing the need for labelled data
Aim 2. Improving robustness to artefacts
Aim 3. Scaling up and distributing developed algorithms