The translation of routinely collected medical data into knowledge that drives the continual improvement of medical care poses grand technical challenges: 1) the extraction, organisation and assembly of large amounts of structured and free-text data embedded with electronic patient records, 2) near real-time analytics and knowledge discovery from the large, temporal, unevenly-sampled and uncertainty-ridden healthcare data and 3) overcoming data biases, problem and domain heterogeneity to design trustworthy prognostic and decision support models that are robust and align with clinical guidelines and workflow. Therefore, the successful design and implementation of tools that convert the data generated as a by-product of patient care into useful insight that can improve efficiency encompasses research in prominent areas of Artificial Intelligence including language engineering, data mining, knowledge representation and reasoning, learning and autonomous systems. This workshop is centred around novel AI methodologies that aim to solve some of the grand challenges associated with medical data. Held in conjunction with the International Joint Conference on Artificial Intelligence (IJCAI 2023), this year's workshop continues the successful KDH series from 2016 to 2022 and builds on this year's IJCAI theme of 'AI for Good.' The workshop received 20 submissions, all of which underwent peer review by members of our program committee. Following the review phase, 5 long papers and 2 short papers were selected for presentation during the workshop. Within the accepted papers, the workshop addresses topics in AI ethics, robustness, interpretability, and fairness. These themes are covered across various domains, including diseasespecific prognosis, natural language processing, data mining and time-series analysis.
Introduction
Biography: Dr Neil Hurley is an Associate Professor in Computer Science, the head of the School of Computer Science at University College Dublin and a principal investigator at the Insight SFI Centre for Data Analytics, Ireland. His research spans data analytics, social network analysis, recommender systems, data hiding, digital watermarking, fingerprinting and high-performance computing. He has won over €1 million euro in research funding from Enterprise Ireland, Science Foundation Ireland, the European Union and industrial partners.
The following full papers presenting original research works were accepted. In Natural Language Processing, Moscato et al. detail a data augmentation framework for named entity recognition and an associated refinement that allows the selection of the most informative examples in an augmented data pool by minimizing noise. Zhang and Roberts present a framework for entity recognition and relation extraction from biomedical text. The framework is a generative model that bypasses the issues that arise when considering the two tasks in a pipeline manner, which ignores the interactions between the tasks.
Two contributions were centred around improving algorithmic performance in biomedical settings. In time-series analysis, Qian, Ibrahim and Dobson build on the state-of-the-art deep learning imputation models via a transformer-based architecture that scales to domains where the data distribution is highly skewed. Ali, Chourasia and Patterson present an evaluation of the effectiveness of Anderson Acceleration in aiding the convergence of machine learning algorithms using bioinformatics settings.
In condition-specific contributions, Kok et al. present an end-to-end machine-learning pipeline for detecting changes in breathing patterns in COPD patients, while Wu et al develop a multi-headed Transformer-based framework to predict viral mutation in SARS-CoV-2.
Finally, our workshop addressed implementation issues, especially pertaining to the fairness and trust of AI systems in sensitive domains such as healthcare. Yogarajan et al. propose an AI-based healthcare framework that comprises a feedback loop that facilitates quality improvements via continuous input provision.
We very much appreciate the support of the workshop chairs, Hadi Hosseini (Penn State University, US) and Viviana Mascardi (the University of Genova, IT). We sincerely hope that the participants enjoyed this year’s workshop program and that this collection of papers will inspire and encourage more AI-related research for and within medicine in the future. Organisation § § § § § § § § § §
Zina Ibrahim, King’s College London (UK) Honghan Wu, University College London (UK) Nirmalie Wiratunga, Robert Gordon University, Aberdeen (UK)
Kerstin Bach, Norwegian University of Science and Technology (Norway) Sadid Hasan, Philips Research North America (USA) Zina Ibrahim, King's College London (UK) Jonathan Rubin, Philips Research North America (USA) Nirmalie Wiratunga, The Robert Gordon University (UK) Honghan Wu, University of Edinburgh (UK)
§ § § § § § § § § §