Special Issue of Information Sciences (Elsevier)
Special Issue on Special Issue on Open-world Multi-modal Machine Learning for Uncertain
Medicine and Healthcare Big Data Analysis

The computerization of medical charts enables the recording of patients'medical histories over time and this generates large-scale datasets that pose various challenges for data analytics, including high dimensionality, large heterogeneity, class imbalance, and, in some cases, low numbers of samples. The medicine and healthcare big data can be available in different formats, such as numeric, textual, time series, and images. The data also can originate from diverse sources. More importantly, there are many uncertainties in the medical decision-making due to incomplete, imprecise, or contradictory data. This can include limited understanding of biological mechanisms, imprecise test measurements, highly subjective and imprecise medical history, inconsistent information from different sources, and missing information in some cases. Although the current research has shown promising results, there is an urgent need to explore and develop advanced intelligent medicine and healthcare decision models that can deal with randomness, imprecision, vagueness, incompleteness, and missing values. Additionally, they must efficiently handle the variety, velocity and volume of medicine and healthcare data, especially the models that can be applied for epidemic monitoring, virus tracking, prevention, control and treatment, and resource allocation.

Multi-modal machine learning (MML) is the practice of training AI models using data from different modalities such as text, images, audio, and video, with the goal of leveraging the complementary information across these modalities for improved performance and deeper insights. Unlike conventional machine learning, which often deals with single-modal data, MML is essential for real-world scenarios where multiple information sources are available. This involves developing algorithms and models capable of effectively handling and integrating data from different modalities through techniques like feature extraction, representation learning, alignment, interpretation, and modality fusion. The objective is to create models that can effectively exploit the synergies between modalities to excel in tasks like classification, regression, clustering, and generation. As its applications span domains like computer vision, natural language processing, healthcare, and autonomous driving, MML continues to evolve alongside advancements in deep learning and reinforcement learning, driving the development of increasingly sophisticated and effective modeling techniques.

The trajectory of current research in MML is shifting from close-set MML to the more expansive and practical domain of open-world MML. Unlike traditional close-set MML models, which are trained and tested on fixed datasets with predetermined modalities, open-world MML encompasses a field where models are trained to accommodate data from multiple modalities in dynamic and evolving environments. In these settings, new modalities may emerge over time, existing modalities may be missing, data may be corrupted or poisoned, or the distribution of data may shift. Consequently, the objective of open-world MML is to cultivate robust and reliable models capable of adapting and generalizing to arbitrary new or missing modalities, uncertain semantics, adversarial perturbations, and changing environments, all while effectively harnessing the complementary information across modalities to enhance performance.

The benefits of investigating open-world multi-modal machine learning for uncertain medicine and healthcare big data analysis have potential to apply in multiple research disciplines and medicine and healthcare applications. Thus, designing an efficient and effective MML model, algorithm, system to handle uncertain medicine and healthcare big data is an emerging and promising topic to improve reasoning and intelligent epidemic monitoring, control and treatment of medicine and healthcare data.

Scope of the Special Issue
This special issue is dedicated to exploring novel learning theories, techniques, and experiments applied to the realm of trustworthy open-world Multi-Modal Machine Learning for Uncertain Medicine and Healthcare Big Data Analysis. We invite submissions encompassing a wide array of topics within the domain of reliable open-world MML. These topics include, but are not limited to:
1) Development of Open-world MML models accommodating new modalities, categories, and distribution shifts
2) MML models designed to handle incomplete multimodal medicine and healthcare data
3) Transfer learning strategies tailored for open-world MML scenarios
4) Implementation of Open-world MML models in the presence of partially-observed data
5) Robust MML architectures resilient to data poisoning, adversarial attacks, and backdoor attacks
6) Open-world MML approaches addressing noise data, semantic noise, and correspondence noise
7) Integration of privacy protection and sensitive information handling in Open-world MML frameworks
8) Efficient compression techniques for large-scale MML systems
9) Value alignment and autonomous evolution strategies for large-scale MML models.
10) Utilization of large-scale pretrained MML models on low-quality medicine and healthcare big data
11) Interpretability-focused approaches in MML
12) Exploration of fundamental theories underpinning large-scale MML
13) Surveys or reviews documenting the current state of research in Open-world for uncertain medicine and healthcare big data analysis
14) Strategies and methodologies for medicine and healthcare big data collection in Open-world MML research endeavors.

We highly recommend the submission of multimedia associated with each article as it significantly increases the visibility, downloads, and citations of articles.

Submission format
Papers will be evaluated based on their originality, presentation, relevance and contribution to current trends of open-world multi-modal machine learning for medicine and healthcare big data analysis as well as their suitability and the quality in terms of both technical contribution and writing. The submitted papers must be written in English and describe original research which has not been published nor currently under review by other journals or conferences. Previously published conference papers should be clearly identified by the authors (at the submission stage) and an explanation should be provided about how the papers have been extended to be considered for this special issue.

Guest Editors will make an initial judgment of the suitability of submissions to this special issue. Papers that either lack originality, clarity in presentation or fall outside the scope of the special issue will not be sent for review and the authors will be promptly informed in such cases.
Author guidelines for preparation of manuscript can be found at www.elsevier.com/locate/ins

Submission Guidelines
All manuscripts and any supplementary material should be submitted through Elsevier Editorial System (EES). The authors must select “SI: Open-world MML” when they identify the “Article Type” step in the submission process. The EES website is located at http://ees.elsevier.com/ins/

Guide for Authors
This site will guide you stepwise through the creation and uploading of your article. The guide for authors can be found on the journal homepage (www.elsevier.com/ins).

Important Dates:
Submission Open Date: June 15, 2024
Final Manuscript Submission Deadline: November 30, 2024
Editorial Acceptance Deadline: April 30, 2025