Article
Oncology
Daniel Truhn, Chiara M. L. Loeffler, Gustav Mueller-Franzes, Sven Nebelung, Katherine J. Hewitt, Sebastian Brandner, Keno K. Bressem, Sebastian Foersch, Jakob Nikolas Kather
Summary: The study demonstrates that large language models, like GPT-4, can extract structured data from unstructured pathology reports with high concordance to human-generated structured data, potentially enabling routine extraction of ground truth data for machine learning.
JOURNAL OF PATHOLOGY
(2023)
Article
Biochemical Research Methods
Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon, Tie-Yan Liu
Summary: This paper introduces a domain-specific generative Transformer language model BioGPT pre-trained on large-scale biomedical literature, which performs well on biomedical natural language processing tasks, especially achieving high accuracy in relation extraction tasks.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Emrah Inan
Summary: This study introduces an entity-centric summarization method that extracts named entities using a dependency parser and generates a small graph, ranking entities using the harmonic centrality algorithm. Experimental results show that it outperforms state-of-the-art unsupervised learning baselines by more than 10% for ROUGE-1 and more than 50% for ROUGE-2 scores, achieving comparable results to recent end-to-end models.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Flor Miriam Plaza-del-Arco, M. Dolores Molina-Gonzalez, L. Alfonso Urena-Lopez, M. Teresa Martin-Valdivia
Summary: The paper discusses the task of Spanish hate speech identification on social media and the capabilities of new techniques based on machine learning. The study compares the performance of different methods, with the main contribution being the achievement of promising results in Spanish through the application of multilingual and monolingual pre-trained language models.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Information Systems
Qinglin Meng, Yan Song, Jian Mu, Yuanxu Lv, Jiachen Yang, Liang Xu, Jin Zhao, Junwei Ma, Wei Yao, Rui Wang, Maoxiang Xiao, Qingyu Meng
Summary: Electric power audit text classification is an important research problem in electric power systems, and automatic classification methods based on machine learning or deep learning have been applied. However, existing pre-training models usually use general corpus, neglecting texts related to electric power, especially electric power audit texts. This paper proposes EPAT-BERT, a BERT-based model pre-trained using two-granularity pre-training tasks to learn abundant morphology and semantics about electric power. Experimental results show that EPAT-BERT significantly outperforms existing models in various evaluation metrics, indicating its potential for electric power audit text classification.
Article
Computer Science, Information Systems
Jipeng Qiang, Feng Zhang, Yun Li, Yunhao Yuan, Yi Zhu, Xindong Wu
Summary: This paper proposes an unsupervised statistical text simplification method using pre-trained language modeling BERT for initialization. Experimental results show that the method outperforms some supervised baselines.
FRONTIERS OF COMPUTER SCIENCE
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Wilson Lau, Kevin Lybarger, Martin L. Gunn, Meliha Yetisgen
Summary: This paper presents a study on semantic annotation of radiology reports using deep learning models. The authors built a corpus of 500 annotated reports and extracted trigger words and argument entities using the advanced deep learning architecture BERT. They predicted the linkages between triggers and argument entities and demonstrated the model's generalizability through testing on an external validation set.
JOURNAL OF DIGITAL IMAGING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Amir M. Hasani, Shiva Singh, Aryan Zahergivar, Beth Ryan, Daniel Nethala, Gabriela Bravomontenegro, Neil Mendhiratta, Mark Ball, Faraz Farhadi, Ashkan Malayeri
Summary: This study compares the quality and content of radiologist-generated and GPT-4 AI-generated radiology reports. The AI-generated reports showed comparable quality to radiologist-generated reports in most categories, with significant differences observed in clarity, ease of understanding, and structure. The AI-generated reports were more concise, but had greater variability in sentence length. Content similarity between the two sets of reports was high.
EUROPEAN RADIOLOGY
(2023)
Article
Computer Science, Theory & Methods
Alhanouf Alduailej, Abdulrahman Alothaim
Summary: This paper introduces AraXLNet, an Arabic language model based on XLNet, and applies it to sentiment analysis tasks in Arabic. The results demonstrate the superior performance of AraXLNet over the previous AraBERT model on multiple benchmark datasets.
JOURNAL OF BIG DATA
(2022)
Article
Computer Science, Artificial Intelligence
Pin Ni, Gangmin Li, Patrick C. K. Hung, Victor Chang
Summary: Predictive biomedical intelligence requires strong professional experience and external domain knowledge support. Transfer learning from massive biomedical text data can enhance the performance of downstream predictive and decision-making task models. This study introduces a StaResGRU-CNN combined with PLMs for biomedical text-based predictive tasks, aiming to improve predictive NLP tasks in the Chinese biomedical field through the introduction of prior knowledge with language models.
APPLIED SOFT COMPUTING
(2021)
Article
Medicine, General & Internal
Max Tigo Rietberg, Van Bach Nguyen, Jeroen Geerdink, Onno Vijlbrief, Christin Seifert
Summary: This study aims to understand the reasons for performing MRI scans on Multiple Sclerosis (MS) patients by extracting information from their free-form reports, specifically diagnosis, progression, or monitoring. By comparing general language models with domain expert opinions and using Explainable Artificial Intelligence (XAI) techniques, the reliability of the models is verified. BERTje, a Dutch Bidirectional Encoder Representations from Transformers (BERT) model, outperforms other models in terms of accuracy and reliability, showing that domain-specific models are not always superior. The validation of BERTje in a small prospective study shows promising results for its potential practical applications.
Article
Computer Science, Interdisciplinary Applications
Si Shen, Jiangfeng Liu, Litao Lin, Ying Huang, Lin Zhang, Chang Liu, Yutong Feng, Dongbo Wang
Summary: The academic literature of social sciences records human civilization and studies human social problems. With its large-scale growth, the ways to quickly find existing research on relevant issues have become an urgent demand for researchers. Previous studies, such as SciBERT, have shown that pre-training using domain-specific texts can improve the performance of natural language processing tasks. However, the pre-trained language model for social sciences is not available so far. In light of this, the present research proposes a pre-trained model based on the abstracts published in the Social Science Citation Index (SSCI) journals. The models, which are available on GitHub (), show excellent performance on discipline classification, abstract structure-function recognition, and named entity recognition tasks with the social sciences literature.
Article
Materials Science, Multidisciplinary
Meiting Zhao, Erxiao Wu, Dongyang Li, Junfei Luo, Xin Zhang, Zhuquan Wang, Qing Huang, Shiyu Du, Yiming Zhang
Summary: MAX/MXenes, with a unique combination of metallic and ceramic properties, have garnered significant attention. This study proposes a baseline model utilizing natural language processing to extract synthesis conditions for MAX/MXenes from literature. The developed model serves as an auxiliary tool for future research and also provides a pre-trained model for extracting synthesis routes of MAX/MXenes.
JOURNAL OF MATERIALS RESEARCH AND TECHNOLOGY-JMR&T
(2023)
Article
Chemistry, Multidisciplinary
Linan Zhu, Yifei Xu, Zhechao Zhu, Yinwei Bao, Xiangjie Kong
Summary: Sentiment-controlled text generation aims to generate texts according to the given sentiment. However, most of the existing studies focus only on the document- or sentence-level sentiment control, leaving a gap for finer-grained control over the content of generated results.
APPLIED SCIENCES-BASEL
(2023)
Review
Engineering, Multidisciplinary
Haifeng Wang, Jiwei Li, Hua Wu, Eduard Hovy, Yu Sun
Summary: This article provides a comprehensive review of representative work and recent progress in the field of NLP and introduces the taxonomy of pre-trained models. It also discusses the impact and challenges of pre-trained models in NLP and addresses future research directions.
Article
Radiology, Nuclear Medicine & Medical Imaging
Jonas Oppenheimer, Keno Kyrill Bressem, Fabian Henry Juergen Elsholtz, Bernd Hamm, Stefan Markus Niehues
Summary: This study examined a novel model-based iterative reconstruction (MBIR) technique for improved detection of low-contrast liver lesions. Objective image metrics showed promise for MBIR methods in improving detectability, but subjective image quality may be perceived as inferior.
Article
Radiology, Nuclear Medicine & Medical Imaging
Lisa C. Adams, Praveen Jayapal, Shakthi K. Ramasamy, Wipawee Morakote, Kristen Yeom, Lucia Baratto, Heike E. Daldrup-Link
Summary: Ferumoxytol is an approved ultrasmall iron oxide nanoparticle that has been increasingly used as an MRI contrast agent, particularly in pediatric patients. Unlike gadolinium-based contrast agents, it is biodegradable and has no potential risk of nephrogenic systemic fibrosis. It has unique MRI properties, including long-lasting vascular retention, making it suitable for various applications, such as vascular, cardiac, and cancer imaging. It is also being researched for its potential use in cellular and molecular imaging and as a potential cancer therapeutic agent.
AMERICAN JOURNAL OF ROENTGENOLOGY
(2023)
Article
Oncology
Sebastian Dahlmann, Keno Bressem, Behschad Bashian, Sevtap Tugce Ulas, Maximilian Rattunde, Felix Busch, Marcus R. Makowski, Katharina Ziegeler, Lisa Adams
Summary: This study examines sex-specific differences in renal cell carcinoma (RCC) and its association with abdominal fat accumulation, psoas muscle density, tumor size, pathology, and survival. It found that men had a higher visceral fat area and psoas muscle index, while women had a higher subcutaneous fat area. Higher psoas muscle index was associated with lower tumor grade and better overall survival. However, there were no associations between abdominal fat measurements and tumor characteristics or survival. There were also no sex-specific differences in tumor size or grade.
ANNALS OF SURGICAL ONCOLOGY
(2023)
Editorial Material
Oncology
Sebastian Dahlmann, Keno K. Bressem, Behschad Bashian, Sevtap Tugce Ulas, Maximilian Rattunde, Felix Busch, Marcus R. Makowski, Katharina Ziegeler, Lisa C. Adams
ANNALS OF SURGICAL ONCOLOGY
(2023)
Correction
Radiology, Nuclear Medicine & Medical Imaging
Keno K. Bressem, Lisa C. Adams, Fabian Proft, Kay Geert A. Hermann, Torsten Diekhoff, Laura Spiller, Stefan M. Niehues, Marcus R. Makowski, Bernd Hamm, Mikhail Protopopov, Valeria Rios Rodriguez, Hildurn Haibel, Judith Rademacher, Murat Torgutalp, Robert G. Lambert, Xenofon Baraliakos, Walter P. Maksymowych, Janis L. Vahldiek, Denis Poddubny
Editorial Material
Radiology, Nuclear Medicine & Medical Imaging
Lisa C. Adams, Daniel Truhn, Felix Busch, Avan Kader, Stefan M. Niehues, Marcus R. Makowski, Keno K. Bressem
Article
Multidisciplinary Sciences
Tabea Fluegge, Robert Gaudin, Antonis Sabatakakis, Daniel Troeltzsch, Max Heiland, Niels van Nistelrooij, Shankeeth Vinayahalingam
Summary: Oral squamous cell carcinoma (OSCC) is a common malignancy with a high mortality rate. Early detection is crucial for effective treatment. This study successfully developed a deep learning-based method using clinical photographs to automatically detect OSCC. The method achieved high accuracy and can assist clinicians in early detection, improving survival rates and quality of life for patients.
SCIENTIFIC REPORTS
(2023)
Editorial Material
Education, Scientific Disciplines
Felix Busch, Lisa C. Adams, Keno K. Bressem
Summary: The integration of artificial intelligence (AI) applications into medical education raises ethical concerns and responsibilities due to the convergence of healthcare, AI, and education ethics. This commentary examines the ethical responsibilities of medical institutions in incorporating AI into medical education by identifying potential concerns and limitations, and provides recommendations to guide the ethical use of AI for medical educators and students.
MEDICAL SCIENCE EDUCATOR
(2023)
Article
Pharmacology & Pharmacy
Felix Busch, Lena Hoffmann, Daniel Truhn, Subish Palaian, Muaed Alomar, Kleva Shpati, Marcus Richard Makowski, Keno Kyrill Bressem, Lisa Christine Adams
Summary: This study aimed to investigate international undergraduate pharmacy students' views on integrating artificial intelligence (AI) into pharmacy education and practice. A survey was conducted with 387 pharmacy students from 16 faculties and 12 countries. The results showed that students had predominantly positive attitudes towards AI in medicine and expressed a strong desire for more AI education. However, they reported limited general knowledge of AI and felt inadequately prepared to use AI in their future careers.
BRITISH JOURNAL OF CLINICAL PHARMACOLOGY
(2023)
Article
Multidisciplinary Sciences
Firas Khader, Jakob Nikolas Kather, Gustav Mueller-Franzes, Tianci Wang, Tianyu Han, Soroosh Tayebi Arasteh, Karim Hamesch, Keno Bressem, Christoph Haarburger, Johannes Stegmaier, Christiane Kuhl, Sven Nebelung, Daniel Truhn
Summary: This work proposes and evaluates a transformer-based neural network as a novel AI architecture that integrates multimodal patient data, i.e., imaging data (chest radiographs) and non-imaging data (clinical data). The performance of the model is evaluated in a retrospective study with 6,125 patients in intensive care. The results show that the combined model outperforms the radiographs-only model and the clinical data-only model in predicting in-hospital survival of patients.
SCIENTIFIC REPORTS
(2023)
Article
Dentistry, Oral Surgery & Medicine
Norbert Neckel, Daniel Troeltzsch, Dario Zocholl, Steffen Koerdt, Yvonne Motzkus, Andrej Trampuz, Jan-Dirk Raguse, Max Heiland, Susanne Nahles
Summary: The purpose of this study was to compare different assessment methods for peri-implant inflammation, evaluate potential risk factors, and establish a comprehensive algorithm for clinical staging, treatment, and evaluation of success in periorbital implants. The results showed that probing depth and skin reaction to be valuable quick assessment tools, which can be complemented with sulcus fluid flow rate for improved accuracy. Patient-specific factors did not have a significant impact on peri-implant inflammation.
INTERNATIONAL JOURNAL OF ORAL & MAXILLOFACIAL IMPLANTS
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Felix Busch, Sarah Keller, Christopher Rueger, Avan Kader, Katharina Ziegeler, Keno K. Bressem, Lisa C. Adams
Summary: This study analyzed the gender and country distribution among editorial boards of leading computer science and AI journals. The results showed that women were underrepresented in all positions and there was a disproportionate relationship between the Global North and South.
ACTA RADIOLOGICA OPEN
(2023)
Article
Computer Science, Artificial Intelligence
Keno K. Bressem, Jens-Michalis Papaioannou, Paul Grundmann, Florian Borchert, Lisa C. Adams, Leonhard Liu, Felix Busch, Lina Xu, Jan P. Loyen, Stefan M. Niehues, Moritz Augustin, Lennart Grosser, Marcus R. Makowski, Hugo J. W. L. Aerts, Alexander Loeser
Summary: This paper presents medBERT.de, a pre-trained German BERT model designed specifically for the German medical domain. The model achieves state-of-the-art performance on various medical benchmarks and the analysis investigates the impact of data deduplication and tokenization methods on the model's performance.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)