Learning with Fewer Labels in Medical Computing
Scope, aims and topics
In medicine and healthcare, we often don’t have enough labelled data due to the great efforts required in annotation, thus requiring models that are able to (a) learn with few annotated examples, and (b) continually adapt to new data without forgetting the prior knowledge. Luckily, we can learn from just few examples (i.e., few shot learning), do long-term learning, and form abstract models of a situation and manipulate these models to achieve extreme generalization. As a result, one of the next big challenges in medical computing is to develop effective learning-based methods that are capable of addressing the important shortcomings of existing techniques in this regard. Therefore, there is pressing need for novel research methods, (1) to drastically reduce require-ments for labeled training data, (2) to significantly reduce the amount of data necessary to adapt the model to new environment, and (3) to even use as little labeled training data as we need.
This special session focuses on the development of deep learning methods with fewer labels for medical computing. The topics of interest include (but are not limited to) the following areas:
Few label learning in medical visual question and answering
Generating diagnostic reports from medical images
Incremental learning in NLP
Generating reports from clinical notes
Embedding medical knowledge in diagnosis
Knowledge transfer under various clinical environments
Image classification with MRI/CT/PET/mesh/point cloud etc
Learning robust medical image representation with noisy annotation
Un-/semi-/weakly-supervised medical data analysis
Predicting clinical outcomes from multimodal medical data
Anomaly detection in medical data
Meta-learning in medical computing
Active learning and life-long learning in medical computing
Zero/few-shot learning in medical computing
Important dates:
Submission of papers: 11:59pm (AoE), June 15, 2022
Acceptance notification: 11:59pm (AoE), August 15, 2022
Camera ready: 11:59pm (AoE), August 31, 2022
Conference date: 11:59pm (AoE), November 22-26, 2022
Organizers
Imran Razzak, University of New South Wales, Australia
Xuequan Lu, Deakin University, Australia
Xiao Liu, Deakin University, Australia
Please submit using the guidelines in ICONIP 2022.