Article on Hand Prostheses Control published in Frontiers in AI

Our work on “Improving Robotic Hand Prostheses Control with Eye Tracking and Computer Vision: a Multimodal Approach based on the Visuomotor Behavior of Grasping“, by M. Cognolato, M. Atzori, R. Gassert and H. Müller, was published in Frontiers in Artificial Intelligence, section Machine Learning and Artificial Intelligence.

In this paper, we investigate a multimodal approach that exploits the use of eye-hand coordination to improve the control of myoelectric hand prostheses.

Example of a typical unimodal and multimodal analysis process flow.

Full paper available open-access online:

Congratulations to Mara Graziani for her PhD

Mara Graziani successfully defended her PhD thesis on the interpretability of deep learning for medical image analysis. This excellent work was supervised by Prof. Stephane Marchand-Maillet (UNIGE) and, Prof. Henning Müller (HES-SO, UNIGE). Prof. Mauricio Reyes (UniBE) was part of the external committee.

Mara defending her work (left) and later literally opening the black-box model with PhD gifts (right)

FishLab : An innovative video system to observe and take a census of fish populations

FishLab, led by the COREALIS and our institute of information systems at HES-SO, will analyze near real-time fish flows and migration using machine learning algorithms on videos. Preliminary experiments with three stations located in the Rhône and the Aare rivers already enabled the improvement of the embedded system to detect and count fish.

More information (in french or in german): here

Paper on Hand Prostheses Control published in Sensors

Our paper on “Questioning Domain Adaptation in Myoelectric Hand Prostheses Control: An Inter- and Intra-Subject Study“, Giulio Marano et al. has been published in Sensors as part of a Special Issue on Biomedical Sensors for Functional Mapping.

In this study, we question the benefit of domain adaptation in transfer learning techniques (using pre-trained models obtained from prior subjects) applied to machine learning algorithms for automatic myoelectric hand prostheses control.

For more information, check the paper now available online:

HECKTOR event at MICCAI 2021

On Sept. 27, we hosted the HECKTOR2021 challenge as a satellite event of MICCAI 2021. We presented the excellent results obtained by the various international teams on the tasks of tumor segmentation and outcome prediction in head and neck cancer.

We had great participation in the event with lively interaction with the participants, and a very interesting keynote by Prof. Clifton Fuller.

The Leaderboard of the challemge is available online:

Stay tuned for the LNCS proceedings reporting the methods and results.

Thank you to all the participants, sponsors and organizers! We look forward to next year’s challenge.

Valais/Wallis AI Workshop

The 7th edition of the Valais/Wallis AI Workshop will take place on Nov. 15 2021 at the IDIAP research institute with a live youtube streaming (hybrid event).

Organized by IDIAP and HES-SO Valais, it will feature various presentations and discussions on “Energy in all its artificial states”, including a keynote presentation by Pr Guglielmina Mutani.

Full program and registration are now available online:

3 papers accepted at MICCAI workshops

We will be presenting our two papers on Head and Neck tumor segmentation and prognostic prediction, and one paper on colorectal cancer detection in whole slide images at MICCAI workshops Sept. 27th and Oct. 1st. Together with our paper at the main conference and the organization of the HECKTOR challenge, we are happy to have a total of five great contributions to MICCAI this year.

3D multi-modal (PET/CT) and multi-task architecture with a common down-sampling branch (green), an up-sampling segmentation branch (blue) and a radiomics branch (red).
Example of a manual annotation of primary tumor in blue and the UNet-generated automatic annotation in red on a fused PET/CT image.
Comparison between pixel-wise annotations made by a pathologist with attention maps of MuSTMIL (ours), Hashimoto et al. (2020) MSMIL and SSMIL. Cancer (red), lgd (yellow), hyperplastic poly (blue), normal tissue (orange).

Paper published in Journal of Personalized Medicine

Our review paper on harmonization methods in radiomics, in collaboration with the School for Oncology, Maastricht University, has been published in the Journal of Personalized Medicine.

The paper is now available online: “Making Radiomics More Reproducible across Scanner and Imaging Protocol Variations: A Review of Harmonization Methods“, Shruti Atul Mali et al.

Radiomics aims to extract quantitative features from medical images to improve decision support. In this work, we discuss various harmonization solutions to make the radiomic features more reproducible across various scanners and protocol settings. The different harmonization solutions are divided into two main categories: image domain (including image acquisition, post-processing of raw sensor-level image data, data augmentation, and style transfer) and feature domain (including statistical normalization, intensity harmonization, ComBat and deep learning harmonization).

Overview of harmonization methods at different stages of medical imaging.

Paper accepted in Frontiers in Computer Science

Our paper on “Multi_Scale_Tools: A Python Library to Exploit Multi-Scale Whole Slide Images” by Niccolò Marini et al. has been published in Frontiers in Computer Science, section Digital Public Health.

This paper describes our python library called Multi_Scale_Tools. It includes four components:

  • a pre-processing component with two methods to split the WSIs in patches from arbitrary magnification levels,
  • a CNN-based scale regressor to predict the magnification in a continuous range between 5-40x,
  • two multi-scale CNN architectures to make predictions at patch-level and
  • a multi-scale CNN architecture called HookNet to segment WSIs.

The code is available on GitHub:

An example of WSI format including multiple magnification levels.
An example of tissue represented at multiple magnification level (5x, 10x, 20x, 40x).
An example of the multicenter extraction method.
Overview of a multi-scale CNN architecture.

HECKTOR test data release

The emulation of the HECKTOR challenge at MICCAI2021 is on with many active participants! The test data has just been released and the submission website will be open from Sep. 01 to Sep. 10. We are very much looking forward to your participation !

More info here:

We are very grateful to the HECKTOR 2021 sponsors:

Task 1 (segmentation) is sponsored by Siemens Healthineers Switzerland with a prize of 500 €.
Task 2 (prediction of PFS) is sponsored by Aquilab with a prize of 500 €.
Task 3 (prediction of PFS using ground-truth delineation of tumors) is sponsored by Biomemtech with a prize of 500 €.

Paper accepted at MICCAI 2021

Our paper “Sharpening Local Interpretable Model-agnostic Explanations for Histopathology: Improved Understandability and Reliability” by Mara Graziani, Iam Palatnik De Sousa et. al will be presented at MICCAI 2021, Sept. 27 to Oct 1st, Strasbourg, France.

In this work, we improve the application of LIME to histopathology images by leveraging nuclei annotations. The obtained visualizations reveal the sharp, neat and high attention of the deep classifier to the neoplastic nuclei in the dataset, an observation in line with clinical decision making.
Compared to standard LIME, our explanations show improved understandability for domain-experts, report higher stability and pass the sanity checks of consistency to data or initialization changes and sensitivity to network parameters.

Paper accepted for presentation at MIDL 2021

Our work on “Common limitations of performance metrics in biomedical image analysis” has been accepted for publication in MIDL 2021. Conducted by a large international consortium including the BIAS initiative, the MICCAI Society’s challenge working group and the benchmarking working group of the MONAI framework, this initiative aims at generating best practice recommendations with respect to metrics in medical image analysis.

Full paper available on arxiv:

Paper accepted in IEEE Transaction on Image Processing

Check out our steerable detectors if you lost your (biological) patterns in noisy backgrounds 🙂

A single-shot template learning will allow finding all occurrences no matter their orientation are and with high recall and specificity ! 

And it is obviously all packaged in an open-source ImageJ plugin!

Plus open source code on GitHub:

Joint work with Julien Fageot, Virginie Uhlmann, Zsuzsanna Püspöki, Benjamin Beck and Michael Unser.

HECKTOR 2021 at MICCAI starting soon

The second edition of the HECKTOR challenge will take place at MICCAI 2021 tackling the Head & Neck tumor analysis in PET-CT images. Visit the challenge page for full description:

In addition to last year’s challenge, we will include more data for the segmentation task as well as a new task for the automatic prediction of patient outcome.

Evaluate your cutting edge algorithm on 300 cases with high-quality annotations from six centers.


  • Release date of the training cases: June 1 2021
  • Release date of the test cases: Aug. 1 2021
  • Submission date(s): opens Sept. 1 2021 closes Sept. 10 2021
  • Release date of the results: Sept. 15 2021
  • Associated workshop days: Sept. 27 2021 or Oct. 1 2021