Nordic IEEE SPS workshop on Advanced Signal/Image Processing: Larger Than Memory Images to Biomedical Sensing

#Advanced #Signal #Processing #Image #Larger #Than #Memory #Images #Biomedical #Sensing #processing #signal
Share

Advanced Signal/Image Processing: Larger Than Memory Images to Biomedical Sensing

We are pleased to inform you that on January 26th, 13:15–15:00, at the University of Stavanger, we will host a workshop on “Advanced Signal/Image Processing”, featuring: larger than memory image processing and bio sensing. The tentative program is as follows: 

13:15–13:30 – Gathering and welcome
13:30–14:00 – Larger-than-Memory Image Processing, Prof. Jon Sporring
(Department of Computer Science, University of Copenhagen, Denmark)
14:00–14:15 – Coffee break
14:15–14:45 – Non-Contact and Contact SpO₂ Measurement Methods for Newborn Resuscitation,
Associate Prof. Øyvind Meinich-Bache
(Laerdal Medical and Department of Electrical Engineering and Computer Science, University of Stavanger)

Place: University of Stavanger, Kjølv Egelands Hus, room: Dataverkstedet

The event will be held in a hybrid format, and you can join online using Teams: 

Microsoft Teams meeting
Join: https://teams.microsoft.com/meet/31631175440466?p=1GFeKqdcBaQSM22j5T
Meeting ID: 316 311 754 404 66
Passcode: Gv2fh6Jt
 

More info on the talks and speakers:

First Talk: Larger than memory image processing

Authors: Jon Sporring, University of Copenhagen, and  David Stansby, University College London

Abstract: This report addresses larger-than-memory image analysis for petascale datasets such as 1.4 PB electron-microscopy volumes and 150 TB human-organ atlases. We argue that performance is fundamentally I/O-bound rather than compute-bound. We show that structuring analysis as streaming passes over data is crucial: for 3D volumes, representing data as 2D slice stacks (e.g., directories or multipage TIFF) outperforms 3D chunked layouts (e.g., Zarr/HDF5) because most algorithms can be organized to read each voxel once or only a few times, avoiding repeated neighbor reads and halo amplification.  We formalize this with sweep-based execution (natural 2D/3D orders), windowed operations, and overlap-aware tiling to minimize redundant access.  Building on these principles, we introduce a domain-specific language (DSL) that encodes algorithms with intrinsic knowledge of their optimal streaming and memory use; the DSL performs compile-time and run-time pipeline analyses to automatically select window sizes, fuse stages, tee and zip streams, and schedule passes for limited-RAM machines, yielding near-linear I/O scans and predictable memory footprints.  The approach integrates with existing tooling for segmentation and morphology but reframes pre/post-processing as pipelines that privilege sequential read/write patterns, delivering substantial throughput gains for extremely large images without requiring full-volume residency in memory.

Bio: Jon Sporring received his Master and Ph.D. degree from the Department of Computer Science (DIKU), University of Copenhagen, Denmark in 1995 and 1998, respectively. Part of his Ph.D. program was carried out at IBM Research Center, Almaden, California, USA. Following his Ph.D, he worked as a visiting researcher at the Computer Vision and Robotics Lab at Foundation for Research & Technology - Hellas, Greece, and as an assistant research professor at 3D-Lab, School of Dentistry, University of Copenhagen. During 2003-2018 he was an associate professor at DIKU. From 2008-2009 he was a part-time Senior Researcher at Nordic Bioscience a/s. In the period 2012-13, he is a visiting professor at the School of Computer Science, McGill University, Montreal, Canada. Jon Sporring also co-founded DigiCorpus Aps in 2012 and served as Chief research officer of the company from 2012-16 developing computer vision-based systems for automatic feedback for physiotherapeutic rehabilitation. In 2007-2012, 2015-2019, and 2021 he was Deputy head for Research at DIKU. Since 2018, he is a full professor at DIKU, and since 2019 he is the Deputy head of the Center for Quantifying Images for MAXIV (QIM).

 

Second Talk: Non-Contact and Contact SpO₂ Measurement Methods for Newborn Resuscitation
 
Abstract: Laerdal Medical is developing a sensor/product to assess SpO₂ during the critical first minutes after birth. Currently, the standard of care—measurement on the newborn wrist—fails to provide healthcare providers with reliable SpO₂ readings during resuscitation. This limitation may be attributed to poor peripheral perfusion in the wrists of non-breathing infants, necessitating exploration of alternative measurement sites. The measurement principle relies on differential light transmission and absorption by tissue across various wavelengths, which varies with oxygen saturation levels. Both non-contact approaches (using multispectral camera technology) and contact-based methods (utilizing smartwatch-style sensors) are being investigated to identify optimal solutions for this clinical challenge.
 
Bio: Øyvind Meinich-Bache received his PhD in image processing and deep learning (2020). Senior Data Scientist at Laerdal Medical and Associate Professor II at University of Stavanger (2020-current). Specializes in automatic activity recognition from signals, particularly images and video, with focus on emergency response applications.
 

  Date and Time

  Location

  Hosts

  Registration


  • Add_To_Calendar_icon Add Event to Calendar
  • University of Stavanger
  • Stavanger, Rogaland
  • Norway
  • Building: Kjølv Egelands Hus
  • Room Number: DataVerkstedet
  • Contact Event Hosts
  • Starts 05 January 2026 11:00 PM UTC
  • Ends 26 January 2026 11:00 AM UTC
  • No Admission Charge