Big Task Small Data, 1001-AI

About the Workshop

It is estimated that there are thousands of diseases afflicting humans. Since each disease embodies its own characteristics or domain knowledge, that is, with its specific signs, symptoms, and according medical image appearances, when these characteristics are learned through data-driven, image-based AI and ML, one likely needs more than one AI model, forming a so-called “1001-AI” research goal. In addition to the disease aspect, there are other factors that aggravate the modeling complexity, such as modality, anatomy, function, etc.

It is a daunting task to fulfill the above goal via existing data-savvy, knowledge-agnostic technology, that is, independent construction of such a gigantic number of AI models with one AI model for classifying one disease, because of the Big Task Small Data challenges, including: (i) difficulty in curating annotated data, which is time-consuming and costly, etc.; and (ii) statistical shift between training and testing distributions.

This workshop aims to investigate this new research frontier by encouraging the submissions from the following research lines, including but not limited to,

• Super-big AI model learning from less-supervised big datasets.
• Zero-shot medical image analysis (recognition, detection, segmentation, etc.)
• Novel methods that integrate domain knowledge with statistical learning.
• Annotation-efficient approaches that learn from one shot or few (<10) shots.
• Domain adaptation or generalization algorithms that are preferably label-free.
• Universal models that integrate data across different modalities, anatomies, imaging sites, diseases, etc.
• Multimodal medical image analysis that fuses knowledge from different sources.
• Medical image synthesis for generating images across different modalities, imaging sites, etc. and from text descriptions.
• Dataset construction that supports the themes of the workshop.
• Other attempt that supports the themes of the workshop.

Workshop Chairs

• Prof. S. Kevin Zhou

  School of Biomedical Engineering (BME)

  University of Science & Technology of China (USTC)

  Email: s.kevin.zhou@gmail.com

• Prof. Julia Schnabel (F)

  Professor for Computational Imaging and AI in Medicine

  Technical University of Munich

  Email: julia.schnabel@tum.de

• Prof. Rama Chellappa

  Bloomberg Distinguished Professor

  Departments of Biomedical Engineering (BME) and Electrical & Computer engineering (ECE)

  Johns Hopkins University

  Email: rchella4@jhu.edu

Join Us Online

We are excited to offer both online and offline participation options for our workshop. If you prefer to attend online, you can join us via Zoom.

Keynote Speakers

Keynote talk I
Prof. Song-Chun Zhu
Peking University

Keynote talk II
Prof. Rama Chellappa
Johns Hopkins University

Workshop Schedule (Date: 8 October, 2023) (Zoom)

Submission Details

The workshop will solicit paper submissions and manage the paper submission process using Microsoft CMT. The submission should be anonymous and follow the MICCAI conference format. A double-blind reviewing process is adhered to guarantee paper quality. A paper will have at least two reviewers. An acceptance rate less than 50% will be strictly followed.

Important Dates