Article
Computer Science, Interdisciplinary Applications
Leibo Liu, Oscar Perez-Concha, Anthony Nguyen, Vicki Bennett, Louisa Jorm
Summary: In this study, a hierarchical label-wise attention Transformer model (HiLAT) is proposed for the explainable prediction of ICD codes from clinical documents. The model utilizes a pretrained Transformer model and a hierarchical label-wise attention mechanism to predict the assignment of specific ICD codes to clinical documents. Experimental results show that HiLAT + ClinicalplusXLNet outperforms previous state-of-the-art models for predicting ICD-9 codes, and attention weight visualizations serve as a potential tool for validating ICD code predictions.
JOURNAL OF BIOMEDICAL INFORMATICS
(2022)
Review
Computer Science, Artificial Intelligence
Fei Teng, Yiming Liu, Tianrui Li, Yi Zhang, Shuangqing Li, Yue Zhao
Summary: The International Classification of Diseases (ICD) is widely used for categorizing physical conditions. Manual ICD coding is time-consuming and prone to errors. Therefore, researchers are focusing on using deep neural networks for ICD automatic coding.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Interdisciplinary Applications
Kunying Niu, Yifan Wu, Yaohang Li, Min Li
Summary: Automated ICD coding is a multi-label prediction task in the deep learning regime. To mitigate the negative effects of large label sets and heavy imbalance distribution, a retrieve and rerank framework is proposed, which introduces Contrastive Learning (CL) for label retrieval and a Transformer variant for refining and reranking the candidate set. Experiments show that this framework provides more accurate results by preselecting a small subset of candidates before fine-level reranking.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Review
Computer Science, Artificial Intelligence
Rajvir Kaur, Jeewani Anupama Ginige, Oliver Obst
Summary: The manual process of assigning clinical codes to free-text clinical narratives is expensive, time-consuming, and error-prone. Researchers have explored the use of Natural Language Processing (NLP), machine learning, and deep learning methods to automate this process and improve accuracy and efficiency. This systematic literature review analyzes automated clinical coding systems that utilize NLP, machine learning, and deep learning methods to assign International Classification of Diseases (ICD) codes to discharge summaries. The review identifies datasets, techniques, and trends in performance evaluation metrics. Efforts are needed to improve code prediction accuracy and access to large-scale de-identified clinical corpora.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Allergy
Luciana Kase Tanno, Yann Briand, Melissa Mary, David A. Khan, James L. Sublett, Mark L. Corbett, Ruby Pawankar, Stefano Del Giacco, Maria Jose Torres, Ignacio J. Ansontegui, Motohiro Ebisawa, Bryan Martin, Pascal Demoly
Summary: This study presents the selection process of allergens for fitting the structure of WHO ICD-11 and provides a classification of allergens based on real-life relevance. A total of 1109 allergens (76.8%) were selected from a database of 1444, with an additional 297 relevant allergens chosen based on real-life data. The stepwise approach used in this study is essential for building a classification of allergens for the WHO ICD-11, and it has significant implications for clinical practice.
JOURNAL OF ALLERGY AND CLINICAL IMMUNOLOGY
(2023)
Article
Anesthesiology
Beatrice Korwisi, Antonia Barke, Winfried Rief, Rolf-Detlef Treede, Maria Kleinstaeuber
Summary: For the first time, the upcoming ICD-11 will include a comprehensive classification of chronic pain, based on the biopsychosocial definition. This publication provides answers to frequently asked questions about the ICD-11 chronic pain classification, aiming to guide future users and facilitate its implementation.
Article
Allergy
Luciana Kase Tanno, Pascal Demoly
Summary: The International Classification of Diseases (ICD) serves as a common diagnostic and classification tool worldwide, aiming to improve the accuracy of mortality and morbidity statistics for allergic and hypersensitivity conditions. The recent change in hierarchy in ICD-11 has allowed for a dedicated section addressing these conditions, leading to more accurate classification and definitions. The implementation of ICD-11 was adopted at the 72nd World Health Assembly in May 2019.
PEDIATRIC ALLERGY AND IMMUNOLOGY
(2022)
Article
Medical Informatics
Yicong Xu, Jingya Zhou, Yi Wang
Summary: This paper analyzes the differences and relationships between ICD-11 and ICD-O in terms of coding structure, compatibility, and intelligence level, pointing out the advantages and characteristics of ICD-11 in neoplasm coding, making it more powerful in statistics, multiaxial coding, coding granularity, compatibility, and intelligence.
BMC MEDICAL INFORMATICS AND DECISION MAKING
(2022)
Article
Computer Science, Artificial Intelligence
Leibo Liu, Oscar Perez-Concha, Anthony Nguyen, Vicki Bennett, Louisa Jorm
Summary: Encouraged by the success of pretrained Transformer models, this study explores their use for ICD coding tasks. The study investigates existing models (PLM-ICD and XR-Transformer) and proposes a novel model (XR-LAT) to address challenges in extreme label set and long text classification. The optimized PLM-ICD models outperform previous SOTA models, while XR-LAT models perform competitively.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Pengli Lu, Jingjin Xue
Summary: Automatic International Classification of Diseases (ICD) coding is a method of classifying diseases through a computer program based on etiology and clinical presentation rules. This paper proposes an approach called TF-GCN to improve the accuracy of automatic ICD coding by feature extraction and relationship analysis.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Nuria Lebena, Alberto Blanco, Alicia Perez, Arantza Casillas
Summary: In this study, we focused on classifying Spanish Electronic Health Records (EHR) based on the International Classification of Diseases (ICD) using Topic Models. We found that Topic Models offer a suitable alternative approach for Spanish clinical text mining when there are limited resources available. Specifically, we explored two methods, Latent Dirichlet Allocation (LDA) and Partially Labelled Latent Dirichlet Allocation (PLDA), and found that PLDA is able to discover topics associated with the ICD, making it a versatile representation for EHRs. Compared to supervised categorization approaches, LDA and PLDA provide an interpretable approach that can be associated with ICDs.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Owen Trigueros, Alberto Blanco, Nuria Lebena, Arantza Casillas, Alicia Perez
Summary: This study focuses on obtaining explainable predictions of diseases and procedures in EHRs using CNNs with attention mechanisms, achieving challenging results in a Spanish corpus. It highlights the helpful information stored in attention mechanisms for assisting medical experts in accurate medical code prediction.
INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS
(2022)
Article
Education & Educational Research
Farkhondeh Asadi, Shokoofeh Afkhami, Farideh Asadi
Summary: This study aimed to evaluate the impact of a training course on poisoning coding rules based on ICD-10 on clinical coders.
BMC MEDICAL EDUCATION
(2023)
Article
Public, Environmental & Occupational Health
Katie Labgold, Kaitlyn K. Stanhope, Naima T. Joseph, Marissa Platner, Denise J. Jamieson, Sheree L. Boulet
Summary: The study assessed the validity of using ICD-10 codes for identifying hypertensive disorders in pregnancy research, and found that while the codes performed well in identifying overall hypertensive disorders, they showed lower sensitivity for certain subdiagnoses. The findings provide important parameters for future studies on hypertensive outcomes during pregnancy in high-burden populations using hospital ICD-10 codes.
Review
Virology
Joseph W. Schaefer, Joshua M. Riley, Michael Li, Dianna R. Cheney-Peters, Chantel M. Venkataraman, Chris J. Li, Christa M. Smaltz, Conor G. Bradley, Crystal Y. Lee, Danielle M. Fitzpatrick, David B. Ney, Dina S. Zaret, Divya M. Chalikonda, Joshua D. Mairose, Kashyap Chauhan, Margaret Szot, Robert B. Jones, Rukaiya Bashir-Hamidu, Shuji Mitsuhashi, Alan A. Kubey
Summary: The study aims to compare the accuracy of manual-chart-review and ICD-10-based comorbidity data, revealing variations in sensitivity and opportunities for improvement in ICD-10 data accuracy.
JOURNAL OF MEDICAL VIROLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Shurong Sheng, Katrien Laenen, Luc Van Gool, Marie-Francine Moens
Summary: This paper presents a weakly supervised alignment model for fine-grained image-text alignment and cross-modal retrieval in the cultural heritage domain, which utilizes attention mechanisms and a common semantic space to achieve superior performance compared to two state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Liesbeth Allein, Isabelle Augenstein, Marie-Francine Moens
Summary: The study reveals that considering the temporal information of evidence can improve the veracity predictions of time-sensitive claims. In fact-checking, time-aware evidence ranking surpasses assumptions based purely on semantic similarity or position in a search results list.
JOURNAL OF WEB SEMANTICS
(2021)
Article
Computer Science, Interdisciplinary Applications
Katrien Laenen, Marie-Francine Moens
Summary: Understanding multimedia content in e-commerce is challenging, and disentangled representation learning is a promising approach. In this study, an explainable variational autoencoder framework (E-VAE) is proposed to obtain disentangled item representations by jointly learning visual and textual data. With the automatic interpretation mechanism, E-VAE provides insight into the quality of the disentanglement. Experimental results demonstrate the effectiveness of the proposed framework in outfit recommendation and cross-modal search tasks.
Article
Computer Science, Information Systems
Liesbeth Allein, Marie-Francine Moens, Domenico Perrotta
Summary: The online audience of a news article provides valuable insights about its identity, but using this information for fake news classification may result in reliance on profiling. To address the increasing demand for ethical AI, a profiling-avoiding algorithm is proposed that leverages Twitter users for model optimization while excluding them during the evaluation of article veracity. This algorithm incorporates objective functions inspired by the social sciences to maximize correlation between the article and its spreaders, as well as among the spreaders. Experimental results demonstrate the positive impact of this approach in improving prediction performance and discriminatory capability between fake and true news.
INFORMATION PROCESSING & MANAGEMENT
(2023)
Proceedings Paper
Computer Science, Information Systems
Liesbeth Allein, Marie-Francine Moens
DISINFORMATION IN OPEN ONLINE MEDIA, MISDOOM 2020
(2020)
Article
Computer Science, Interdisciplinary Applications
Graham Spinks, Marie-Francine Moens
Article
Acoustics
Artuur Leeuwenberg, Marie-Francine Moens
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2020)
Article
Computer Science, Interdisciplinary Applications
Vincent Vandeghinste, Tom Vanallemeersch, Liesbeth Augustinus, Bram Bulte, Frank Van Eynde, Joris Pelemans, Lyan Verwimp, Patrick Wambacq, Geert Heyman, Marie-Francine Moens, Iulianna Van der Lek-Ciudin, Frieda Steurs, Ayla Rigouts Terryn, Els Lefever, Arda Tezcan, Lieve Macken, Veronique Hoste, Joke Daems, Joost Buysschaert, Sven Coppers, Jan Van den Bergh, Kris Luyten