Lab

MedGIFT

About the lab

The MedGIFT project started at the medical faculty of the University of Geneva, Switzerland in 2002 and is since 2007 located in the Institute for Business Information Systems at the HES-SO in Sierre (Valais), Switzerland. The name stems originally from the use of the GNU Image Finding Tool (GIFT) for medical applications. Over the years the GIFT has been used less frequently and a large set of tools and applications have been developed to advance the field of medical visual information retrieval. All developed tools are open source and can be requested by email. Some very old tools might not be available anymore. A very strong collaboration with medical informatics and the University Hospitals and University of Geneva, Switzerland continues to keep the group activities in medical informat

Featured projects (1)

Project
Exascale volumes of diverse data from distributed sources are continuously produced. Healthcare data stand out in the size produced (production 2020 >2000 exabytes), heterogeneity (many media, acquisition methods), included knowledge (e.g.diagnostic reports) and commercial value. The supervised nature of deep learning models requires large labeled, annotated data, which precludes models to extract knowledge and value. EXA MODE solves this by allowing easy & fast, weakly supervised knowledge discovery of exascale heterogeneous data provided by the partners, limiting human interaction. Its objectives include the development and release of extreme analytic methods and tools, that are adopted in decision making by industry and hospitals. Deep learning naturally allows building semantic representations of entities and relations in multimodal data. Knowledge discovery is performed via document-level semantic networks in text and the extraction of homogeneous features in heterogeneous images. The results are fused, aligned to medical ontologies, visualized and refined. Knowledge is then applied using a semantic middleware to compress, segment and classify images and it is exploited in decision support and semantic knowledge management prototypes. The ExaMode project is supported by the European Union through the Horizon 2020 framework.

Featured research (24)

Muscle synergy analysis is commonly used for investigating the neurophysiological mechanisms that the central nervous system employs to control muscle activations. In the last two decades, several models have been developed to decompose EMG signals into spatial, temporal or spatiotemporal synergies. However, the presence of different approaches complicates the comparison and interpretation of results. Spatial synergies represent invariant activation weights in muscle groups modulated with variant temporal coefficients, while temporal synergies are based on invariant temporal profiles that coordinate variant muscle weights. While non-negative matrix factorization (NMF) allows to extract both spatial and temporal synergies, temporal synergies and the comparison between the two approaches have been barely investigated and so far no study targeted a large set of multi-joint upper limb movements. Here we present several analyses that highlight the duality of spatial and temporal synergies as a characterization of low-dimensional and intermittent motor coordination in the upper limb, allowing high flexibility and dexterity. First, spatial and temporal synergies were extracted from two datasets representing a comprehensive mapping of proximal (REACH PLUS) and distal (NINAPRO) upper limb movements, focusing on their differences in reconstruction accuracy and inter-individual variability. For both models, we extracted synergies achieving a given level of the goodness of reconstruction (R2), and we compared the similarity of the invariant components across participants. The two models provide a compact characterization of motor coordination at spatial or temporal level, respectively. However, a lower number of temporal synergies are needed to achieve the same R2 with a higher inter-subject similarity. Spatial and temporal synergies may thus capture different levels of motor control. Second, we showed the existence of both spatial and temporal structure in the EMG data, extracting spatial and temporal synergies from a surrogate dataset in which the phases were shuffled preserving the same frequency content of the original data. Last, a detailed characterization of the structure of the temporal synergies suggested that they can be related to an intermittent control of the movement. These results may be useful to improve muscle synergy analysis in several fields such as rehabilitation, prosthesis control and motor control studies.
PICO recognition is an information extraction task for identifying participant, intervention, comparator, and outcome information from clinical literature. Manually identifying PICO information is the most time-consuming step for conducting systematic reviews (SR), which is already labor-intensive. A lack of diversified and large, annotated corpora restricts innovation and adoption of automated PICO recognition systems. The largest-available PICO entity/span corpus is manually annotated which is too expensive for a majority of the scientific community. To break through the bottleneck, we propose DISTANT-CTO, a novel distantly supervised PICO entity extraction approach using the clinical trials literature, to generate a massive weakly-labeled dataset with more than a million ``Intervention'' and ``Comparator'' entity annotations. We train distant NER (named-entity recognition) models using this weakly-labeled dataset and demonstrate that it outperforms even the sophisticated models trained on the manually annotated dataset with a 2% F1 improvement over the Intervention entity of the PICO benchmark and more than 5% improvement when combined with the manually annotated dataset. We investigate the generalizability of our approach and gain an impressive F1 score on another domain-specific PICO benchmark. The approach is not only zero-cost but is also scalable for a constant stream of PICO entity annotations.
ImageCLEF s part of the Conference and Labs of the Evaluation Forum (CLEF) since 2003. CLEF 2022 will take place in Bologna, Italy. ImageCLEF is an ongoing evaluation initiative which promotes the evaluation of technologies for annotation, indexing, and retrieval of visual data with the aim of providing information access to large collections of images in various usage scenarios and domains. In its 20th edition, ImageCLEF will have four main tasks: (i) a Medical task addressing concept annotation, caption prediction, and tuberculosis detection; (ii) a Coral task addressing the annotation and localisation of substrates in coral reef images; (iii) an Aware task addressing the prediction of real-life consequences of online photo sharing; and (iv) a new Fusion task addressing late fusion techniques based on the expertise of the pool of classifiers. In 2021, over 100 research groups registered at ImageCLEF with 42 groups submitting more than 250 runs. These numbers show that, despite the COVID-19 pandemic, there is strong interest in the evaluation campaign.KeywordsUser awarenessMedical image classificationMedical image understandingCoral image annotation and classificationFusionImageCLEF benchmarkingAnnotated data
Computational pathology is a domain that aims to develop algorithms to automatically analyze large digitized histopathology images, called whole slide images (WSI). WSIs are produced scanning thin tissue samples that are stained to make specific structures visible. They show stain colour heterogeneity due to different preparation and scanning settings applied across medical centers. Stain colour heterogeneity is a problem to train convolutional neural networks (CNN), the state-of-the-art algorithms for most computational pathology tasks, since CNNs usually underper-form when tested on images including different stain variations than those within data used to train the CNN. Despite several methods that were developed, stain colour hetero-geneity is still an unsolved challenge that limits the development of CNNs that can generalize on data from several medical centers. This paper aims to present a novel method to train CNNs that better generalize on data including several colour variations. The method, called H&E-adversarial CNN, exploits H&E matrix information to learn stain-invariant features during the training. The method is evaluated on the classification of colon and prostate histopathol-ogy images, involving eleven heterogeneous datasets, and compared with five other techniques used to handle stain colour heterogeneity. H&E-adversarial CNNs show an improvement in performance compared to the other algorithms , demonstrating that it can help to better deal with stain colour heterogeneous images. * niccolo.marini@hevs.ch
Building accurate knowledge of the identity, the geographic distribution and the evolution of species is essential for the sustainable development of humanity, as well as for biodiversity conservation. However, the difficulty of identifying plants and animals is hindering the aggregation of new data and knowledge. Identifying and naming living plants or animals is almost impossible for the general public and is often difficult even for professionals and naturalists. Bridging this gap is a key step towards enabling effective biodiversity monitoring systems. The LifeCLEF campaign, presented in this paper, has been promoting and evaluating advances in this domain since 2011. The 2021 edition proposes four data-oriented challenges related to the identification and prediction of biodiversity: (i) PlantCLEF: cross-domain plant identification based on herbarium sheets, (ii) BirdCLEF: bird species recognition in audio soundscapes, (iii) GeoLifeCLEF: remote sensing based prediction of species, and (iv) SnakeCLEF: Automatic Snake Species Identification with Country-Level Focus.

Lab head

Henning Müller
Department
  • Information systems
About Henning Müller
  • Henning Müller currently works at the Information systems Institute, HES-SO Valais-Wallis in Sierre, Switzerland. Henning does research in Artificial Intelligence, Medical Image analysis and information retrieval.

Members (9)

Manfredo Atzori
  • HES-SO Valais-Wallis
Cristina Simon-Martinez
  • HES-SO Valais-Wallis
Vincent Andrearczyk
  • HES-SO Valais-Wallis
Roger Schaer
  • HES-SO Valais-Wallis
Ivan Eggel
  • HES-SO Valais-Wallis
Mara Graziani
  • HES-SO Valais-Wallis
Niccolò Marini
  • University of Applied Sciences and Arts Western Switzerland
Valentin Oreiller
  • Lausanne University Hospital
Vincent Andrearczyk
Vincent Andrearczyk
  • Not confirmed yet

Alumni (6)

Alba García Seco de Herrera
  • University of Essex
Yashin Dicente Cid
  • Roche Diagnostics
Dimitrios Markonis
  • HES-SO Valais-Wallis