A video on our 3D printed prosthetic hand on Swiss national television

The recording “In buone mani” (In good hands) was presented at a scientific show of the Swiss Radio-television in Italian language (RSI). The show described the projects ProHand and MeganePro , targeting the development of robotic hands with advanced technologies, such as 3D scanning, additive manufacturing and machine learning.


Paper accepted in Computers in Biology and Medicine

Our journal paper entitled “A lung graph model for the radiological assessment of chronic thromboembolic pulmonary hypertension in CT” by O. Jimenez del Toro, Y. Dicente Cid, A. Platon, A-L. Hachulla Lemaire, F. Lador, P-A. Poletti and H. Müller has been accepted for publication in Computers in Biology and Medicine.

The paper is now available online: https://www.sciencedirect.com/science/article/abs/pii/S001048252030295X

Paper accepted at iMIMIC, workshop at MICCAI

Our paper entitled “Interpretable CNN Pruning for Preserving Scale-Covariant Features in Medical Imaging” by M. Graziani, T. Lompech, H. Müller, A. Depeursinge and V. Andrearczyk has been accepted for presentation at the Workshop on Interpretability of Machine Intelligence in Medical Image Computing (iMIMIC) at MICCAI 2020.

In this paper, we analyze the invariance to scale in standard pre-trained CNNs and the effects of this invariance in a medical imaging task, where scale often carries crucial information.

New paper accepted at MICCAI LABELS 2020

Our paper entitled “Semi-Weakly Supervised Learning for Prostate Cancer Image Classification with Teacher-Student Deep Convolutional Networks“, by S. Otálora, N. Marini, M. Atzori and H. Müller has been accepted for publication at the MICCAI workshop LABELS 2020: Large-scale Annotation of Biomedical data and Expert Label Synthesis.

In this paper, we propose a teacher-student approach for training data-hungry models using images automatically annotated by a deep CNN model. Our results in the challenging task of automatic grading of prostate cancer images show that using this approach is significantly better than training the CNN with the small set of ground-truth annotations. It opens the possibility to use vast amounts of non-annotated data. 

Two papers published in the Ebook: Digital Personalized Health and Medicine

Two of our papers have been published in the EBook: Digital Personalized Health and Medicine.

“Automating Quality Control for Structured Standardized Radiology Reports Using Text Analysis”, by Anjani Dhrangadhariya, S. Millius, C. Thouly, B. Rizk, D. Fournier, H. Müller, H. Brat.

Available online: https://pubmed.ncbi.nlm.nih.gov/32570346/

“Machine Learning Assisted Citation Screening for Systematic Reviews” by Anjani Dhrangadhariya, R. Hilfiker, R. Schaer, H. Müller.

Available online:  http://ebooks.iospress.nl/publication/54173

Paper accepted at MICCAI 2020

Our paper entitled “Multimodal Latent Semantic Alignment for Automated Prostate Tissue Classification and Retrieval” has been accepted for presentation at MICCAI 2020.

In this work, our group collaborated with the MindLAB group (Universidad Nacional de Colombia) to develop an information fusion method for the automatic analysis of prostate cancer images.

More details can be found in the paper.

Paper accepted in Computers in Biology and Medicine

Our new paper entitled “Concept attribution: Explaining CNN decisions to physicians” by Mara Graziani, Vincent Andrearczyk, Stephane Marchand-Maillet and Henning Müller has been accepted for publication in Computers in Biology and Medicine.

Feature attribution techniques explain Convolution Neural Networks (CNNs) in terms of the input pixels. The abstraction to higher level impacting factors, however, can be difficult when only a pixel-based analysis is performed. In this paper, we generate explanations in terms of prognostic factors for breast cancer, such as nuclei pleomorphism.

Official start of the HECKTOR challenge

We are happy to announce the official start of the HECKTOR challenge (at MICCAI 2020) with the release of the training data. It comprises 201 cases of Head and Neck tumors with manual annotations on PET-CT images for segmentation of primary Gross Tumor Volumes.

The data can be downloaded from the aicrowd page.

A code for preprocessing the data, a baseline CNN and an example evaluation can be found on our github repository.