Seminar Reports

On this page, we provide an overview of past MMIV seminar talks with relevant links. If you’re looking for upcoming seminars, please check out our Events page.

Our seminar organizing committee consists of Frank Riemer, Heidi Espedal, Eric Mörth, and Samaneh Mofrad.

MMIV Seminar – November 2019 – Behind the scenes – The people behind the machines and images



Sergej Stoppel –  “Firefly: Virtual Illumination Drones for Interactive Visualization”

My current research interest is focused on the field of visual data science. By analyzing the intrinsic dimensionality of the data, I work on finding a suitable visual embedding for the data and thus create visualization methods tailored to specifically for the data. Additionally I am interested in high dimensional parameter space exploration and intuitive interaction techniques that support users with complex tasks, by predicting the users intention and providing appropriate support. My latest research projects focused on automatic and time efficient determination of parameter settings for larger algorithmic systems, by exploiting parallel computing and evaluating the parameter space. Links: UiB


Julie Dybvik –   “MRI-assessed Tumor-free Distance to Serosa Predicts Deep Myometrial Invasion and Poor Prognosis in Endometrial Cancer”

The diagnostic accuracy of preoperative magnetic resonance imaging (MRI) and MRI-based tumor measurements is important for prediction of pathological deep (≥50%) myometrial invasion (pDMI) and for prognostication in endometrial carcinomas (EC). We have investigated preoperative pelvic MRI scans of 357 prospectively included patients with histologically confirmed EC. Three radiologists have reported findings suggesting deep (≥50%) myometrial invasion (iDMI) and the tumor measurements:  axial anteroposterior tumor diameter (APD), depth of myometrial tumor invasion (DOI) and tumor free distance to serosa (TFD). Receiver operating characteristic (ROC) curves for prediction of pDMI using hysterectomy specimen as reference standard, were plotted for the different tumor measurements and optimal cut-off values were determined. The predictive and prognostic value of the tumor measurements were analyzed and interobserver agreement was assessed. TFD yielded the highest area under the ROC-curve (AUC) for prediction of pDMI. Multivariate analysis (including cut-off based imaging variables and preoperative histological risk-status) for predicting pDMI yielded highest predictive value of TFD<6 mm with OR of 6.1 (p<0.001) and lower figures for DOI ≥5mm (OR=2.2; p=0.04), APD ≥17mm (OR=3.1, p<0.001) and iDMI (OR=1.1 (p=0.76). Patients with TFD<6 mm also had significantly reduced survival with hazard ratio of 1.9; p=0.01. The interobserver agreement was good for APD≥17mm (κ=0.70), moderate for TFD<6 mm (κ=0.52), but only fair for DOI ≥5mm  (κ=0.25) and iDMI (κ=0.36).  At preoperative MRI TFD<6 mm was the strongest predictor of pDMI and was associated with poor survival. TFD<6 mm outperforms iDMI for prediction of pDMI and could aid in identifying high-risk disease in endometrial carcinomas. Links: MMIV, UiB


Marek Kocinski – “Quantitative analysis of 3D MR images”

An accurate modeling of blood-vessel structures depicted in 3D raster images is a crucial issue in vascular disease diagnosis and treatment. Magnetic resonance imaging (MRI) includes several modalities allowing one to acquire 3D raster images in which blood vessels are visualized e.g. ToF, SWI, QSM. Using the image information about vasculature offers a possibility to create an accurate and comprehensive 3D geometrical model of arteries and veins for each individual patient with personalized geometry and structure of thick blood vessels. As vessel radius splits from thicker down to thinner, with range from centimeters to micrometers at capillary level, for different vessel thickness groups one should apply various quantitative algorithms. One should apply various quantitative algorithms: a 3D geometric model, texture analysis, and DCE-derived blood pharmacokinetics maps for respectively thick, medium and capillary blood vessels. Links: MMIV, UiB


Frank Riemer -“One box to fit them all: Diffusion microstructure imaging and the question of where to find water”

Dr Frank Riemer will give an introduction into microstructure diffusion imaging, what other methods exist and how that relates to the underlying acquired data. By making analogies and comparisons to general photography, the talk is designed to be accessible to a wider audience. Preliminary results of on-going research projects in diffusion imaging of Multiple Sclerosis here at Haukeland Hospital are given at the end. Links: PubMed, UiB


MMIV Seminar – October 2019 – MR & I – Magnetic resonance imaging from different viewpoints within the MMIV

Illustration by Laura Garrison

Noeska Smit – “Memento: Localized Time-Warping for Spatio-Temporal Selection in fMRI data”

Interaction techniques for temporal data are often focused on affecting the spatial aspects of the data, for instance through the use of transfer functions, camera navigation, or clipping planes. However, the temporal aspect of the data interaction is often neglected. The temporal component is either visualized as individual time steps, an animation, or a static summary over the temporal domain. We propose a novel technique that allows users to interactively specify areas of interest in the spatio-temporal domain. By employing a time-warp function, we are able to slow down time, freeze time, or even travel back in time, around spatio-temporal events of interest. The combination of such a (pre-defined) time-warp function and brushing directly in the data to select regions of interest allows for a detailed review of temporally and spatially localized events, while maintaining an overview of the global spatio-temporal data. In this talk, I will demonstrate an application of this technique to functional MRI (fMRI) data in particular. Links: MMIV, UiB


Njål Lura – “MRI-assessed tumor size parameters predict mortality in uterine cervical cancer” 

Uterine cervical cancer represents a major threat to female health worldwide; it is the fourth most common female cancer and one of the leading causes of cancer-related death in low-income countries. Important limitations in cervical cancer treatment are due to: 1) insufficient diagnostic tools with which to identify high-risk disease and 2) insufficient diagnostic tools with which to guide more individualized treatment. This project aims to address these limitations by focusing on the value of preoperative advanced imaging to provide functional and morphological tumor characteristics relevant for treatment and prognosis in uterine cervical cancer. Links: MMIV


Saruar Alam – “Analyzing atlas-based MRI features on Alzheimer’s disease detection”

Alzheimer’s disease (AD) can be distinguished using the features obtained from Magnetic resonance imaging (MRI), and a supervised classifier. Multi-atlas-based-likelihood fusion (MALF) algorithms extract the volumes features of subcortical regions of interest (ROI). The correlation among these ROI features from different brain regions may provide additional valuable information. Subsequently, these ROI-correlative features may affect the classification performance of a supervised classifier. We have classified AD and Mild Cognitive Impairment from cognitively normal subjects using these features and a Support Vector Machine classifier. This article investigates the difference in classification performance between the ROI and ROI-correlative features. Our work also reports the ranks of ROI and ROI-correlative regions. We have observed marginal differences in classification performance and ranking of the most effective regions. Links: MMIV, UiB


Rune Eikeland – “The Human Connectome Project data processing pipeline and data visualization toolbox”

The Human Connectome Project (HCP) is a five-year project sponsored by sixteen components of the National Institutes of Health, split between two consortia of research institutions. The project was launched in July 2009 as the first of three Grand Challenges of the NIH’s Blueprint for Neuroscience Research. The goal of the Human Connectome Project is to build a “network map” (connectome) that will shed light on the anatomical and functional connectivity within the healthy human brain, as well as to produce a body of data that will facilitate research into brain disorders such as dyslexia, autism, Alzheimer’s disease, and schizophrenia. The data, processing and visualization tools are freely available for the HCP and Rune will present them in this talk with an in-depth focus on EPI data pre-processing to reduce susceptibility and motion artefacts with FSL’s eddy tool (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/eddy). Links: MMIV, UiB


MMIV Seminar – September 2019 – From mouse to man – Spatial and temporal medical imaging in different species and development phases

https://lh6.googleusercontent.com/KtES41ojqwd9dSg1pc1lnZwx8SytePcLtm5KcW_kVO1VYwEzlPbfBplnw2JJpcBmJlsbUwjSrJzbz3DY9k-Gzvm0Js6T-CSwyyvfJTACSoDEhwsEfwyAJwlC8ICcITkEGVOsRf1_
Illustration by Laura Garrison

Heidi Espedal – “Imaging of preclinical gynecologic cancer models” 

Endometrial cancer is the most common type of cancer of the female reproductive tract. Although prognosis is generally good for patients with low-grade and early-stage diseases, the outcomes for high-grade and metastatic/recurrent cases remain poor, since traditional therapy have limited effects. No targeted agents have been approved so far, although several new drugs have been tested without striking results in clinical trials. Patient-derived tumor xenograft (PDX) mouse models represent useful tools for preclinical evaluation of new therapies and biomarker identification. Preclinical imaging by PET-CT and MRI during disease progression enables visualization and quantification of functional tumor characteristics, which may serve as imaging biomarkers guiding targeted therapies. The primary objective for this presentation is to give an introduction of current and novel preclinical imaging methods relevant for endometrial cancer mouse models. Links: MMIV, UiB


Eric Mörth – “The Vitruvian Baby: Interactive reformation of 3D ultrasound data to a T-pose”

Three-dimensional (3D) ultrasound imaging and visualization is often used in medical diagnostics, especially in prenatal screening. Screening the development of the fetus is important to assess possible complications early on. Performing the analysis in a 3D view would enable the viewer to better discriminate between artefacts and representative information. Additionally making data comparable between different investigations and patients is a goal in medical imaging techniques and is often achieved by standardization. “The Vitruvian Baby” incorporates a complete pipeline for standardized measuring in fetal 3D ultrasound. The input of the method is a 3D ultrasound screening of a fetus and the output is the fetus in a standardized T-pose. In this pose, taking measurements is easier and comparison of different fetuses is possible. In addition to the transformation of the 3D ultrasound data, we create an abstract representation of the fetus based on accurate measurements. Links: MMIV, UiB 


Hauke Bartsch – “The Brain Imaging Data of the ABCD Study – an Introduction protocols and tasks”

The Adolescent Brain Cognitive Development (ABCD) Study is the largest long-term study of brain development and child health in the United States. The ABCD Research Consortium have invited 11,878 children ages 9-10 to join the study. Researchers will track their biological and behavioral development through adolescence into young adulthood. Using cutting-edge technology, scientists will determine how childhood experiences (such as sports, video-games, social media, unhealthy sleep patterns, and smoking) interact with each other and with a child’s changing biology to affect brain development and social, behavioral, academic, health, and other outcomes. As part of a research agreement, MMIV will obtain access to the raw data that includes an extensive MRI protocol as well as behavioural and environmental data. Links: MMIV, Github


Sathiesh Kaliyugarasan Artificial intelligence in image diagnostics – transfer learning and active learning for efficient use of data and radiologist’s expertise”

A common stumbling block for supervised learning methods based on deep neural networks is the large number of labeled examples required for training. This is particularly troublesome when trying to use deep learning methods for segmentation in 3D medical image data. As creating labeled data for medical images is often a time-consuming, difficult and unreliable process, the amount of training data available is in general very small. To mitigate this problem we are looking into using design methodologies such as transfer learning and active learning for efficient use of data and radiologist’s expertise. Links: MMIV, UiB