Article
Radiology, Nuclear Medicine & Medical Imaging
Minki Chung, Seo Taek Kong, Beomhee Park, Younjoon Chung, Kyu-Hwan Jung, Joon Beom Seo
Summary: Automated algorithms that identify nodular patterns in chest X-ray images can help reduce reading time and improve accuracy for radiologists. This study proposes a framework to generate realistic nodules and shows how they can be used to train a deep neural network for accurate detection and localization of nodular patterns in CXR images. The proposed method enhances the recall of the detection model while maintaining a low level of false positives.
JOURNAL OF DIGITAL IMAGING
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Tim Gossye, Dimitri Buytaert, Peter V. Smeets, Lieve Morbee, Charlotte De Wilde, Karen Vermeiren, Eric Achten, Klaus Bacher
Summary: This study evaluated the effects of different virtual grid software ratios on gridless chest radiographs using visual grading analysis. Two image quality assessment algorithms were also investigated.
INVESTIGATIVE RADIOLOGY
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Jong Hyuk Lee, Hyunsook Hong, Gunhee Nam, Eui Jin Hwang, Chang Min Park
Summary: This study assessed the impact of AI diagnostic performance and reader characteristics on the detection of malignant lung nodules during AI-assisted reading of chest radiographs. The use of a high accuracy AI model improved readers' detection performance to a greater extent than a low accuracy AI model, and readers using the high accuracy AI showed a higher susceptibility to changing their diagnosis based on AI suggestions.
Article
Medicine, General & Internal
Parisa Kaviani, Mannudeep K. Kalra, Subba R. Digumarthy, Reya Gupta, Giridhar Dasegowda, Ammar Jagirdar, Salil Gupta, Preetham Putha, Vidur Mahajan, Bhargava Reddy, Vasanth K. Venugopal, Manoj Tadepalli, Bernardo C. Bizzo, Keith J. Dreyer
Summary: This study found that there are significant missed findings in chest X-ray interpretation, but the AI model can help identify and reduce the frequency of these missed findings in a generalizable manner.
Article
Computer Science, Information Systems
Wenjian Sun, Dongsheng Wu, Yang Luo, Lu Liu, Hongjing Zhang, Shuang Wu, Yan Zhang, Chenglong Wang, Houjun Zheng, Jiang Shen, Chunbo Luo
Summary: In this article, a fully deep learning pneumoconiosis staging paradigm is proposed, which effectively solves the problem of model overfitting caused by stage ambiguity and noisy labels of pneumoconiosis.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Giridhar Dasegowda, Bernardo C. Bizzo, Reya V. Gupta, Parisa Kaviani, Shadi Ebrahimian, Debra Ricciardelli, Faezeh Abedi-Tari, Nir Neumark, Subba R. Digumarthy, Mannudeep K. Kalra, Keith J. Dreyer
Summary: This study evaluated radiologist-trained AI models for differentiating between suboptimal and optimal chest radiographs. The results showed that these AI models can accurately identify issues in chest radiographs, providing guidance for repeat imaging when needed.
ACADEMIC RADIOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Kyungjin Cho, Jiyeon Seo, Sunggu Kyung, Mingyu Kim, Gil-Sun Hong, Namkug Kim
Summary: In this study, a novel method for bone suppression in pediatric chest radiographs (CXRs) was developed. A model trained with digitally reconstructed radiographs (DRRs) of adults was used to generate pseudo-pediatric CXRs, and a U-Net was trained with paired data to suppress bone in pediatric CXRs. The results showed that the method effectively removed bones while preserving pixel intensity in soft-tissue regions, making it useful for detecting early pulmonary disease in pediatric CXRs.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Gil-Sun Hong, Kyung-Hyun Do, A-Yeon Son, Kyung-Wook Jo, Kwang Pyo Kim, Jihye Yun, Choong Wook Lee
Summary: This study compared image quality and radiation dose between dual-energy subtraction and software-based bone suppression techniques in chest radiography. The results showed that software-based bone suppression images had higher SSIM values in bone areas and lower radiation dose in soft tissue areas, with both readers rating the visual quality of software-based images higher.
EUROPEAN RADIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Ju Gang Nam, Hyun Jin Kim, Eun Hee Lee, Wonju Hong, Jongsoo Park, Eui Jin Hwang, Chang Min Park, Jin Mo Goo
Summary: The study found that using deep learning-based algorithm can significantly improve radiologists' sensitivity in detecting lung nodules on chest radiographs. However, the difference in detection rate was only 0.24%, requiring a large sample size for a randomized controlled trial.
EUROPEAN RADIOLOGY
(2022)
Article
Chemistry, Analytical
Lawrence Y. Deng, Xiang-Yann Lim, Tang-Yun Luo, Ming-Hsun Lee, Tzu-Ching Lin
Summary: With the rapid development of AI and ML, image recognition technology has been widely used in medical imaging diagnosis, which can alleviate the heavy workload of doctors. This study proposed a reliable system for detecting pneumothorax diseases, which showed high accuracy and low misdiagnosis rate in real hospital tests, demonstrating the feasibility of the research.
Article
Computer Science, Interdisciplinary Applications
Bowen Wang, Toshihiro Takeda, Kento Sugimoto, Jiahao Zhang, Shoya Wada, Shozo Konishi, Shirou Manabe, Katsuki Okada, Yasushi Matsumura
Summary: By utilizing positional information extracted from CXR reports, this study aimed to generate bounding boxes with disease lesions on CXR images. Through semantic segmentation and classification models, object detection on the generated attention bounding boxes improved the precision of nodule detection.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Computer Science, Information Systems
Navdeep Kaur, Ajay Mittal, Gurprem Singh
Summary: Generating accurate and coherent linguistic descriptions of visual patterns in medical images is a challenging task, as many radiologists struggle due to workload, time constraints, and fatigue. Research has been conducted in recent years to develop methods for automated report generation to tackle this issue.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Xiyue Wang, Sen Yang, Jun Lan, Yuqi Fang, Jianhui He, Minghui Wang, Jing Zhang, Xiao Han
Summary: This study proposed a two-stage deep learning method for pneumothorax segmentation, which achieved good results in pneumothorax diagnosis through ensemble of multiple models and multitask learning strategy.
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Hyunsuk Yoo, Sang Hyup Lee, Chiara Daniela Arru, Ruhani Doda Khera, Ramandeep Singh, Sean Siebert, Dohoon Kim, Yuna Lee, Ju Hyun Park, Hye Joung Eom, Subba R. Digumarthy, Mannudeep K. Kalra
Summary: The use of AI algorithm as a second reader can improve the performance of readers in lung cancer detection on chest X-rays, particularly benefiting radiology residents.
EUROPEAN RADIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Midori Ueno, Kotaro Yoshida, Atsushi Takamatsu, Takeshi Kobayashi, Takatoshi Aoki, Toshifumi Gabata
Summary: This study evaluated the independent performance of commercially available deep learning-based automatic detection (DLAD) software in the presence of background pulmonary abnormalities. The DLAD software exhibited a sensitivity of 0.689 and a false-positive rate of 0.168. The detectability of the DLAD software was lower for small and subsolid nodules and those with overlapping structures.
EUROPEAN JOURNAL OF RADIOLOGY
(2023)