Overview

Machine learning (ML) systems are achieving remarkable performances at the cost of increased complexity. Deep neural networks, in particular, appear as black-box machines, and their behavior can sometimes be unpredictable. Furthermore, more complex models are less interpretable, which may cause distrust. As these systems are pervasively introduced to critical domains, such as medical image computing and computer-assisted intervention (MICCAI), developing methodologies for explaining model predictions is imperative. Such methodologies would help physicians to decide whether they should follow/trust a prediction or not and might help to identify failure cases. Additionally, it could facilitate the deployment of such systems from a legal perspective. Ultimately, interpretability is closely related to AI safety in healthcare.

However, there needs to be more work regarding interpretability of ML systems in the MICCAI research. Besides increasing trust and acceptance by physicians, the interpretability of ML systems can be helpful during method development. For instance, inspecting if the model is learning aspects coherent with domain knowledge or studying failures. Also, it may help reveal biases in the training data or identify the most relevant data (e.g., specific MRI sequences in multi-sequence acquisitions). This is critical since the rise of chronic conditions has led to a continuous growth in the usage of medical imaging, while at the same time reimbursements have been declining. Hence, interpretability can help improve image acquisition protocols' productivity by highlighting learned features and their relationships to disease patterns.

The Workshop on Interpretability of Machine Intelligence in Medical Image Computing (iMIMIC) at MICCAI 2024 aims at introducing the challenges & opportunities related to the topic of interpretability of ML systems in the context of MICCAI.

Scope

Interpretability can be defined as an explanation of the machine learning system. It can be broadly defined as global or local. The former explains the model and how it has learned, while the latter explains individual predictions. Visualization is often helpful in assisting the process of model interpretation. The model’s uncertainty may be seen as a proxy for interpreting it by identifying difficult instances. Still, more work is needed to tackle the lack of formal and clear definitions, general approaches, and regulatory frameworks. Additionally, interpretability results often rely on comparing explanations with domain knowledge. Hence, there is a need for defining objective, quantitative, and systematic evaluation methodologies.

This workshop aims to foster discussions, and presentation of ideas to tackle the many challenges and identify opportunities related to the topic of interpretability of ML systems in the context of MICCAI. Therefore, the primary purposes of this workshop are:

1. To introduce the challenges/opportunities related to the interpretability of machine learning systems in the context of MICCAI. While there have been workshops on the interpretability of machine learning systems in general machine learning and A.I. conferences (NeurIPS, ICML), to the best of our knowledge, iMIMIC is the only workshop dedicated to the medical imaging application domain.
2. To understand the state of the art of this field. This will be achieved through the submitted manuscripts and the invited keynote speakers.
3. To join researchers in this field and to discuss the issues related to it and future work.
4. To understand the implications of (or lack of) interpretability of machine learning systems in the MICCAI field.

Covered topics include but are not limited to:

• Definition of interpretability in context of medical image analysis.
• Visualization techniques useful for model interpretation in medical image analysis.
• Local explanations for model interpretability in medical image analysis.
• Interpretability methods to make use of multimodal data.
• Causal interpretability.
• Methods to improve transparency of machine learning models commonly used in medical image analysis.
• Textual explanations of model decisions in medical image analysis.
• Uncertainty quantification in context of model interpretability.
• Quantification and measurement of interpretability.
• Legal and regulatory aspects of model interpretability in medicine.

Program

DST time - October 6th

13:30 – 13:40: Opening Session

13:40 – 14:25 Keynote: Michael Kampffmeyer, University of Tromsø, Norway (45 mins)

14:30 – 15:30: Oral Presentations Part I (15 mins each)

15:30 – 16:30: Coffee break and poster presentations

16:30 – 17:15: Oral Presentations Part II (15 mins each)

17:15 – 17:30: Awards and Closing Session

Oral Presentations

• "DWARF: Disease-weighted network for attention map refinement" by Haozhe Luo et al. [PDF]
• "PIPNet3D: Interpretable Detection of Alzheimer in MRI Scans" by Lisa Anita De Santi et al. [PDF]
• "Detecting Unforeseen Data Properties with Diffusion Autoencoder Embeddings using Spine MRI data" by Robert Graf et al. [PDF]
• "Interpretability of Uncertainty: Exploring Cortical Lesion Segmentation in Multiple Sclerosis" by Nataliia Molchanova et al. [PDF]
• "TextCAVs: Debugging vision models using text" by Angus Nicolson et al. [PDF]
• "Evaluating Visual Explanations of Attention Maps for Transformer-based Medical Imaging" by Minjae Chung et al. [PDF]
• "Exploiting XAI maps to improve MS lesion segmentation and detection in MRI" by Federico Spagnolo et al. [PDF]

Keynote Spearker

• Michael Kampffmeyer, University of Tromsø (UiT) - The Arctic University of Norway.

  Title: Towards Self-explainable Deep Learning Models [PDF]

  

  Bio: Michael Kampffmeyer is a Professor at UiT The Arctic University of Norway. He is also a Senior Research Scientist II at the Norwegian Computing Center in Oslo. His research interests include medical image analysis, explainable AI, and learning from limited labels (e.g. clustering, few/zero-shot learning, domain adaptation and self-supervised learning). Kampffmeyer received his PhD degree from UiT in 2018. He has had long-term research stays in the Machine Learning Department at Carnegie Mellon University and the Berlin Center for Machine Learning at the Technical University of Berlin. He is a general chair of the annual Northern Lights Deep Learning Conference, NLDL.

  X-handle: @MKampffmeyer

Keynote speakers

Best paper award

Congratulations to the authors of the best paper award iMIMIC 2024!

"TextCAVs: Debugging vision models using text" by Angus Nicolson et al. [PDF]

Paper submission

Authors should prepare a manuscript of 8-10 pages, including references. The manuscript should be formatted according to the Lecture Notes in Computer Science (LNCS) style and anonymized. As per previous years, we will have preference to publish proceedings following MICCAI Springer’s publication model.

All submissions will be reviewed by 3 reviewers. Authors will be asked to disclose possible conflict of interests, such as cooperation in the previous two years. Moreover, care will be taken to avoid reviewers from the same institution as the authors. The selection of the papers will be based on their relevance for medical image analysis, significance of results, technical and experimental merit, and clear presentation. Following previous editions, we will employ Microsoft’s CMT platform to conduct the review process.

We intend to join the MICCAI Satellite Events joint proceedings, and publish the accepted papers as LNCS. We are also considering making the pre-print of the accepted papers publicly available.

The authors of the best paper of the workshop will receive a Best Paper award.

Click here to submit your paper.

Important dates

• Opening of submission system: 3 June 2024
• Paper submission due: 24 June 2024 29 June 2024
• Reviews due: 8 July 2024
• Notification of paper decisions: 15 July 2024
• Camera-ready papers due: 20 August 2024
• Workshop: 6 October 2024

Venue

The iMIMIC workshop will take place as part of MICCAI 2024 conference that will take place between the 6th and 10th of October 2024 in the Palmeraie Conference Centre in Marrakesh, Morocco.

More information regarding the venue can be found at the conference website.

Organizing Team

General Chairs

• Mauricio Reyes, University of Bern, Switzerland.
• Jaime Cardoso, INESC Porto, Universidade do Porto, Portugal
• Jayashree Kalpathy-Cramer, MGH Harvard University, USA.
• Nguyen Le Minh, Japan Advanced Institute of Science and Technology, Japan.
• Pedro Abreu, CISUC and University of Coimbra, Portugal.
• José Amorim, CISUC and University of Coimbra, Portugal.
• Wilson Silva, Utrecht University, The Netherlands.
• Mara Graziani, HES-SO Valais-Wallis, Switzerland.
• Amith Kamath, University of Bern, Switzerland.

Sponsors








Interested in participating and being a sponsor? Email us