Article
Multidisciplinary Sciences
Laszlo Makrai, Bettina Fodroczy, Sara Agnes Nagy, Peter Czeiszing, Istvan Csabai, Geza Szita, Norbert Solymosi
Summary: This article presents a method for automated counting of bacterial colonies using convolutional neural networks. By culturing 24 bacteria species of veterinary importance and manually annotating 56,865 colonies in a dataset of 369 digital images, it provides a resource for developing artificial intelligence-based approaches to count bacterial colonies.
Article
Chemistry, Multidisciplinary
Zhengru Shen, Marco Spruit
Summary: The Summary of Product Characteristics from the European Medicines Agency serves as a reference document on medicines in the EU, providing information on safe medication use. Natural language processing techniques can be utilized to automatically extract adverse drug reactions, aiding clinical experts in efficient utilization.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Information Systems
Amril Nazir, Ricky Maulana Fajri
Summary: Efficient diagnosis of COVID-19 is crucial in preventing its spread, and recent advancements in artificial intelligence, particularly deep learning techniques like the proposed DS3 framework, have shown significant improvements in data annotation for COVID-19. Experimental results across various countries' datasets demonstrate the superior performance of DS3 over other state-of-the-art models, with an average G-Mean improvement of 10%. Significance testing further confirms the effectiveness and superiority of DS3 in active learning for COVID-19 data.
Article
Ecology
Joan Gomez-Gomez, Ester Vidana-Vila, Xavier Sevillano
Summary: This paper introduces the deployment of an expert system running over a wireless acoustic sensors network that recognizes bird species from their sounds. It presents the development of the Western Mediterranean Wetland Birds (WMWB) dataset, which consists of annotated audio excerpts of 20 endemic bird species. The paper also presents the results of bird species classification experiments using deep neural networks fine-tuned on the dataset.
ECOLOGICAL INFORMATICS
(2023)
Article
Computer Science, Theory & Methods
Philipp Terhorst, Daniel Faehrmann, Jan Niklas Kolf, Naser Damer, Florian Kirchbuchner, Arjan Kuijper
Summary: Soft-biometrics are important in face biometrics and related fields, but current face databases lack in attribute annotations. This study introduces a novel annotation-transfer pipeline to accurately transfer attribute annotations, resulting in the MAAD-Face annotation database with high correctness and a large number of annotations. Human evaluation shows the superiority of MAAD-Face annotations, and insights into soft-biometrics for recognition are provided using this dataset.
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
(2021)
Article
Multidisciplinary Sciences
Hannah C. Cubaynes, Peter T. Fretwell
Summary: Monitoring whales in remote areas is crucial for conservation, but using traditional survey platforms is difficult. Very high-resolution satellite imagery shows promise, but accurate automated whale detection systems are lacking. This study presents a dataset of 633 annotated whale objects detected in satellite images, creating a valuable resource for training and testing automatic detection systems. The dataset covers four species across various regions and was captured by different high-resolution satellites.
Article
Engineering, Electrical & Electronic
Bilal Hassan, Muhammad Fiaz, Hafiz Husnain Raza Sherazi, Usman Javed Butt
Summary: To evaluate standalone soft biometrics systems, there is a need for multi-modality annotated datasets. We have designed and annotated a new dataset called "Annotated Pedestrians for the individuals", which incorporates multiple modalities of the human body. This dataset is highly diverse and suitable for short-term tracking and feature-based retrieval from the database.
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING
(2023)
Article
Multidisciplinary Sciences
Jenn-Jhy Tseng, Chien-Hsing Lu, Jun-Zhou Li, Hui-Yu Lai, Min-Hu Chen, Fu-Yuan Cheng, Chih-En Kuo
Summary: This study developed a chromosome detector using deep learning that achieved an accuracy of 98.88% in chromosomal identification. The researchers also created a large database containing chromosome images and annotations for training chromosome detectors, which can serve as a reference and facilitate the development of clinical applications.
Article
Chemistry, Multidisciplinary
Yunya Gao, Stefan Lang, Dirk Tiede, Getachew Workineh Gella, Lorenz Wendt
Summary: Refugee-dwelling footprints derived from satellite imagery are beneficial for humanitarian operations. The Object-Based Image Analysis (OBIA) approach has been widely applied to this task for the past decade. This research compares the performance of OBIA labels and Manual labels and finds that OBIA labels can produce applicable results and can be improved with a small number of Manual labels.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Interdisciplinary Applications
Krist Shingjergji, Remzi Celebi, Jan Scholtes, Michel Dumontier
Summary: The quality of data used in machine learning models for drug discovery and clinical decision support is crucial. This study combines weak supervision and deep learning methods to extract high-quality drug-disease relationship data and build a machine learning model to classify different relationships, showing significant improvement over traditional methods.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Health Care Sciences & Services
Fangyu Zhou, Shahadat Uddin
Summary: In recent years, the amount of data on drugs and their associated adverse drug reactions (ADRs) has significantly increased. A high hospitalization rate worldwide has been reported due to these ADRs. This has led to extensive research on predicting ADRs in the early stages of drug development to minimize future risks. In this study, a drug-to-drug network was constructed using non-clinical data sources to reveal the relationships between drug pairs based on their shared ADRs. The network features were combined with drug features and fed into various machine learning models, showing that logistic regression had the highest mean AUROC score (82.1%) among all the models tested, indicating the potential importance of the network-based approach in future ADR prediction.
Review
Pharmacology & Pharmacy
Kanoot Jaruthamsophon, Paul J. Thomson, Chonlaphat Sukasem, Dean J. Naisbitt, Munir Pirmohamed
Summary: This review explores the translational progress of using HLA as a key susceptibility factor for immune ADRs and highlights gaps in our knowledge. Additionally, it covers relevant findings of HLA-mediated drug-specific T cell activation.
ANNUAL REVIEW OF PHARMACOLOGY AND TOXICOLOGY
(2022)
Article
Biology
Pierre-Marie Allard, Arnaud Gaudry, Luis-Manuel Quiros-Guerrero, Adriano Rutz, Miwa Dounoue-Kubo, Tom W. N. Walker, Emmanuel Defossez, Christophe Long, Antonio Grondin, Bruno David, Jean-Luc Wolfender
Summary: As privileged structures, natural products often exhibit potent biological activities. However, the complexity of the biological matrices they are found in often hinders the discovery of novel bioactive scaffolds. Large natural extract collections are valuable for their chemical novelty potential but challenging to exploit in drug discovery projects. Establishing digital layers documenting the complexity and bioactivity of these collections can help prioritize isolation efforts. In this study, the exploration of a collection of 1,600 plant extracts for drug discovery purposes is presented, including taxonomic coverage, liquid chromatography high-resolution mass spectrometric profiling, and computational solutions for data analysis. The resulting dataset and metadata are made available for researchers interested in computational natural products exploration.
Review
Biochemistry & Molecular Biology
Fumi Miyagawa, Hideo Asada
Summary: For severe cutaneous adverse reactions (SCARs) like Stevens-Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN) and drug-induced hypersensitivity syndrome (DIHS)/drug reaction with eosinophilia and systemic symptoms (DRESS), analyzing distinct chemokine profiles can aid in their diagnosis and treatment. It is crucial to identify new therapeutic targets for SCARs, with chemokines playing a key role in the pathogenesis and adjuvant diagnosis of these conditions. Additionally, the association between human herpesvirus 6 (HHV-6) and DIHS/DRESS, as well as the possible roles of chemokine/chemokine receptor homologs encoded by HHV-6 in the pathogenesis, are areas of increasing interest.
Article
Health Care Sciences & Services
Mohammad Ali Khaleel, Amer Hayat Khan, Siti Maisharah Sheikh Ghadzi, Azreen Syazril Adnan, Qasem M. Abdallah
Summary: FAERS is one of the largest spontaneous adverse events reporting databases in the world, but analyzing its data presents many challenges. This study provides a processed and deduplicated dataset from the FAERS database, with standardized and normalized drug names. Additionally, pre-calculated disproportionate analysis is included for each drug-adverse event pair in the database.
Review
Computer Science, Interdisciplinary Applications
Yuqi Si, Jingcheng Du, Zhao Li, Xiaoqian Jiang, Timothy Miller, Fei Wang, W. Jim Zheng, Kirk Roberts
Summary: Patient representation learning involves developing dense mathematical representations of patients from Electronic Health Records (EHRs) using advanced deep learning methods. Studies from 2015 to 2019 saw a doubling in publications on this topic, with structured EHR data, recurrent neural networks, and supervised learning being commonly used approaches. Disease prediction was the most common application, while privacy concerns and lack of benchmark datasets were challenges faced by researchers in this field.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Computer Science, Interdisciplinary Applications
Yuqi Si, Elmer Bernstam, Kirk Roberts
Summary: The study proposes a multi-task pre-training and fine-tuning approach to learn generalized and transferable patient representations, focusing on the impact on low-prevalence phenotypes. By validating the pre-training representations and fine-tuning the models, it is found that multi-task pre-training can increase learning efficiency and performance.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Multidisciplinary Sciences
Nicholas L. Rider, Gina Cahill, Tina Motazedi, Lei Wei, Ashok Kurian, Lenora M. Noroski, Filiz O. Seeborg, Ivan K. Chinn, Kirk Roberts
Summary: The study constructed a Bayesian network model for real-time risk assessment of primary immunodeficiency and guidance for appropriate diagnostic work. The model showed high accuracy in classifying immunodeficiency patients from controls and categorizing validation cohort members. Compared to other machine learning models, the network performed better, providing higher transparency with a prescriptive output element.
Article
Computer Science, Interdisciplinary Applications
Kirk Roberts, Tasmeer Alam, Steven Bedrick, Dina Demner-Fushman, Kyle Lo, Ian Soboroff, Ellen Voorhees, Lucy Lu Wang, William R. Hersh
Summary: The TREC-COVID Challenge is an IR shared task aimed at evaluating search on scientific literature related to COVID-19, conducted over five rounds with participation from 92 unique teams and 556 individual submissions. A total of 50 topics were used in the evaluation, starting at 30 topics for Round 1 and adding 5 new topics per round.
JOURNAL OF BIOMEDICAL INFORMATICS
(2021)
Article
Computer Science, Interdisciplinary Applications
Tavleen Singh, Kirk Roberts, Trevor Cohen, Nathan Cobb, Amy Franklin, Sahiti Myneni
Summary: This study developed a method called PRISM to analyze peer interactions in an online health community, revealing the interplay of multilevel characteristics of online communication and their association with individual health behaviors. The analysis showed that social support and behavioral progress were common communication themes, feedback and monitoring and comparison of behavior were the most used behavior change techniques, and expressive and emotional speech acts were commonly used to express intentions. Social network analysis also revealed associations between users' engagement or abstinence status with various categories of behavior change techniques and speech acts.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Article
Computer Science, Information Systems
Sarvesh Soni, Surabhi Datta, Kirk Roberts
Summary: This article proposes a system called quEHRy, which retrieves precise and interpretable answers to natural language questions from structured data in electronic health records (EHRs). The performance of quEHRy is evaluated on 2 clinical question answering (QA) datasets, and the results show high precision and accuracy. However, errors in medical concept extraction affect the downstream generation of correct logical structures, indicating the need for QA-specific clinical concept normalizers.
JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION
(2023)
Article
Health Care Sciences & Services
Surabhi Datta, Kirk Roberts
Summary: The objective of this study is to improve clinical natural language processing by utilizing weak supervision methods that leverage domain resources and expertise. The authors proposed a weak supervision approach based on data programming to extract spatial information from radiology reports. Their weakly supervised BERT model achieved satisfactory results in extracting spatial relations without the need for manual annotations, and outperformed state-of-the-art models when annotated data was available.
Article
Health Care Sciences & Services
Andrew Wen, Huan He, Sunyang Fu, Sijia Liu, Kurt Miller, Liwei Wang, Kirk E. Roberts, Steven D. Bedrick, William R. Hersh, Hongfang Liu
Summary: Clinical phenotyping is essential for developing digital health applications, but the traditional manual abstraction method is time-consuming and costly. Therefore, there is significant interest in using in-silico methods to accomplish this task. However, current in-silico phenotyping development lacks reusable tools for cross-task generalization and is inaccessible for many potential users. This article highlights the barriers to using in-silico phenotyping and proposes a framework with potential solutions. An example implementation of this framework is introduced as a software application, focusing on ease of adoption, cross-task reusability, and facilitating clinical phenotyping algorithm development.
NPJ DIGITAL MEDICINE
(2023)
Article
Medicine, Research & Experimental
Ana Szarfman, Jonathan G. Levine, Joseph M. Tonning, Frank Weichold, John C. Bloom, Janice M. Soreth, Mark Geanacopoulos, Lawrence Callahan, Matthew Spotnitz, Qin Ryan, Meg Pease-Fye, John S. Brownstein, W. Ed Hammond, Christian Reich, Russ B. Altman
Summary: Efficient and easy access to large quantities of health data is crucial for improving medical care and advancing medical research. This article discusses the importance of standardized data collection and transmission systems, universal identifiers, and modernized working practices in achieving this goal.
COMMUNICATIONS MEDICINE
(2022)
Article
Health Care Sciences & Services
Enshuo Hsu, Ioannis Malagaris, Yong-Fang Kuo, Rizwana Sultana, Kirk Roberts
Summary: This paper focuses on extracting information from scanned documents, particularly the sleep apnea indicators in sleep study reports. The study demonstrates that the proper use of image preprocessing and document layout can be beneficial for scanned document processing.
Article
Geriatrics & Gerontology
Meghana Gudala, Mary Ellen Trail Ross, Sunitha Mogalla, Mandi Lyons, Padmavathy Ramaswamy, Kirk Roberts
Summary: This study aimed to assess the benefits, barriers, and information needs that can be provided by an artificial intelligence-powered medication information voice chatbot for older adults. Results showed that the technology would help older adults overcome vision and dexterity hurdles, increase medication knowledge and adherence, and support overall health. However, technology familiarity and cost were identified as major barriers.
Article
Health Care Sciences & Services
Tiffany Champagne-Langabeer, Michael Swank, Shruthi Manas, Yuqi Si, Kirk Roberts
Summary: The study found that the COVID-19 pandemic led to a rapid expansion of telehealth services, and most users have a positive or neutral attitude towards telehealth.
Article
Computer Science, Information Systems
Sarvesh Soni, Kirk Roberts
Summary: The COVID-19 pandemic has increased demand for scientific information, leading to the emergence of commercial search engines, which underperformed academic alternatives in a comparative study. This has implications for the trust in popular health search engines and the development of biomedical search engines for future health crises.
JOURNAL OF THE AMERICAN MEDICAL INFORMATICS ASSOCIATION
(2021)
Proceedings Paper
Biochemical Research Methods
Muhammad Amith, Jing Wang, Grace Xiong, Kirk Roberts, Cui Tao
2020 IEEE INTERNATIONAL CONFERENCE ON BIOINFORMATICS AND BIOMEDICINE
(2020)
Proceedings Paper
Health Care Sciences & Services
Kate Fultz Hollis, Kirk Roberts, Steven Bedrick, William R. Hersh
DIGITAL PERSONALIZED HEALTH AND MEDICINE
(2020)