Article
Engineering, Biomedical
Yue Li, Zilong He, Jiawei Pan, Weixiong Zeng, Jialing Liu, Zhaodong Zeng, Weimin Xu, Zeyuan Xu, Sina Wang, Chanjuan Wen, Hui Zeng, Jiefang Wu, Xiangyuan Ma, Weiguo Chen, Yao Lu
Summary: This study develops a deep learning-based CADe model with an adaptive receptive field to detect atypical architectural distortions (ADs) in digital breast tomosynthesis (DBT). The results show a significant improvement in AD detection performance, especially for atypical ADs.
PHYSICS IN MEDICINE AND BIOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Jingkun Wang, Haotian Sun, Ke Jiang, Weiwei Cao, Shuangqing Chen, Jianbing Zhu, Xiaodong Yang, Jian Zheng
Summary: This study proposes a novel method for MC detection in digital breast tomosynthesis (DBT). The method improves the feature sharing and extraction in CNN models to achieve accurate and rapid detection of small and low-contrast MCs. Experimental results on a clinical dataset demonstrate impressive performance of the method in MC detection, providing valuable diagnostic suggestions for early breast cancer screening.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2023)
Article
Engineering, Biomedical
Jiawei Pan, Zilong He, Yue Li, Weixiong Zeng, Yaya Guo, Lixuan Jia, Hai Jiang, Weiguo Chen, Yao Lu
Summary: This study proposes an anatomical-structure-based multi-view information fusion approach for computer-aided detection of breast architectural distortion. The experimental results show a significant improvement in performance for detecting atypical AD of breast cancer, which can lead to earlier diagnosis and better patient outcomes.
PHYSICS IN MEDICINE AND BIOLOGY
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Md Belayat Hossain, Robert M. Nishikawa, Juhun Lee
Summary: This study developed algorithms for detecting biopsy-proven breast lesions on digital breast tomosynthesis (DBT) using multi-depth level convolutional models and leveraging non-biopsied samples. The study found that false positive findings from non-biopsied benign lesions could improve detection algorithms, and the ensemble model with multi-depth levels showed improved lesion detection performance.
Article
Radiology, Nuclear Medicine & Medical Imaging
Bruno Barufaldi, Raymond J. Acciavatti, Emily F. Conant, Andrew D. A. Maidment
Summary: A virtual clinical trial method is proposed to determine the limit of calcification detection in tomosynthesis. The study simulated breast anatomy, focal findings, image acquisition, and interpretation using screening data and inserted calcifications into virtual breast phantoms. The results suggest that DBT acquisition geometries that use super-resolution reconstruction voxel size smaller than the detector element size and step-and-shoot motion have the potential to improve the detection of calcifications.
EUROPEAN RADIOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Yinhao Ren, Xuan Liu, Jun Ge, Zisheng Liang, Xiaoming Xu, Lars J. Grimm, Jonathan Go, Jeffrey R. Marks, Joseph Y. Lo
Summary: Computer-aided detection (CAD) frameworks for breast cancer screening have been widely researched, with early adoption of deep-learning models showing improved detection performance compared to traditional methods on single-view images. Recent studies have focused on merging information from multiple views to further improve performance, but most multi-view CAD frameworks are black-box techniques that lack the ability to analyze model behaviors and fine-tune performance. This hinders clinical adoption due to the lack of explicit reasoning for each multi-view referencing step.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Jung Hyun Yoon, Eun-Kyung Kim, Ga Ram Kim, Kyunghwa Han, Hee Jung Moon
Summary: This study assessed the impact of adding DBT or AI-CAD on recall rate and diagnostic performance in women undergoing mammographic surveillance after BCT. The results showed that adding DBT or AI-CAD reduced recall rates and improved diagnostic accuracy.
AMERICAN JOURNAL OF ROENTGENOLOGY
(2022)
Article
Oncology
Adem Polat, Raziye Kubra Kumrular
Summary: This study introduced a realistic breast phantom for determining the reconstruction method in clinical applications of digital breast tomosynthesis (DBT). By testing various parameters such as iteration numbers and projections, it was found that the SART method yields better results than FBP in DBT.
TECHNOLOGY IN CANCER RESEARCH & TREATMENT
(2022)
Review
Computer Science, Artificial Intelligence
Jun Bai, Russell Posner, Tianyu Wang, Clifford Yang, Sheida Nabavi
Summary: The recent reintroduction of deep learning has revolutionized diagnostic imaging interpretation, while the technology used for image acquisition is also undergoing a revolution. Digital breast tomosynthesis (DBT) has transformed the field of breast imaging and is rapidly replacing traditional two-dimensional mammography.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Biochemistry & Molecular Biology
William Lotter, Abdul Rahman Diab, Bryan Haslam, Jiye G. Kim, Giorgia Grisot, Eric Wu, Kevin Wu, Jorge Onieva Onieva, Yun Boyer, Jerrold L. Boxerman, Meiyun Wang, Mack Bandler, Gopal R. Vijayaraghavan, A. Gregory Sorensen
Summary: Breast cancer poses a global challenge and early detection through screening mammography is crucial. Recent advancements in applying deep learning to mammography have addressed key difficulties, enhancing accuracy and accessibility of screening mammography.
Article
Engineering, Biomedical
Yue Li, Zilong He, Yao Lu, Xiangyuan Ma, Yanhui Guo, Zheng Xie, Genggeng Qin, Weimin Xu, Zeyuan Xu, Weiguo Chen, Haibin Chen
Summary: A deep-learning-based model using mammary gland distribution as prior information was proposed to improve the performance of Computer Aided Detection (CADe) for breast lesions. The model achieved significantly better results compared to existing methods, demonstrating the effectiveness of incorporating gland distribution and deep learning techniques in CADe for Architectural Distortion (AD) detection.
PHYSICS IN MEDICINE AND BIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Yue Li, Zilong He, Xiangyuan Ma, Weixiong Zeng, Jialing Liu, Weimin Xu, Zeyuan Xu, Sina Wang, Chanjuan Wen, Hui Zeng, Jiefang Wu, Weiguo Chen, Yao Lu
Summary: Breast cancer is the most common cancer worldwide. This study aims to develop a deep-learning-based model for detecting architectural distortion (AD) in digital breast tomosynthesis (DBT). The results show that the model has high accuracy and significantly reduces false positives (FPs).
Article
Radiology, Nuclear Medicine & Medical Imaging
Si Eun Lee, Kyunghwa Han, Eun-Kyung Kim
Summary: This study compared the diagnostic agreement and performances of synthetic and conventional mammograms when artificial intelligence-based computer-assisted diagnosis (AI-CAD) is used. The results showed good agreement and comparable diagnostic performance between the two types of mammograms, indicating that AI-CAD can work well on synthetic mammograms.
EUROPEAN RADIOLOGY
(2021)
Article
Chemistry, Multidisciplinary
Alaa M. Adel El-Shazli, Sherin M. Youssef, Abdel Hamid Soliman
Summary: Digital breast tomosynthesis (DBT) improves the efficiency and accuracy of detecting abnormalities. The proposed computer-aided multi-class diagnosis system integrates DBT augmentation and color feature map with a modified deep learning architecture (Mod_AlexNet). The system outperforms traditional methods in terms of test accuracy.
APPLIED SCIENCES-BASEL
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Qi Wei, Yu-Jing Yan, Ge-Ge Wu, Xi-Rong Ye, Fan Jiang, Jie Liu, Gang Wang, Yi Wang, Juan Song, Zhi-Ping Pan, Jin-Hua Hu, Chao-Ying Jin, Xiang Wang, Christoph F. Dietrich, Xin-Wu Cui
Summary: The study found that CAD software on ultrasound has high accuracy, sensitivity, and specificity in distinguishing benign and malignant breast masses, and can reduce unnecessary biopsies. After the application of CAD software, there was a significant decrease in unnecessary biopsy rate and an increase in malignant biopsy rate in BI-RADS category 4a.
EUROPEAN RADIOLOGY
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Heang-Ping Chan, Mark A. Helvie, Katherine A. Klein, Carol McLaughlin, Colleen H. Neal, Rebecca Oudsema, W. Tania Rahman, Marilyn A. Roubidoux, Lubomir M. Hadjiiski, Chuan Zhou, Ravi K. Samala
Summary: This study compared the sensitivity, confidence level, and reading efficiency of radiologists in detecting microcalcifications in digital breast tomosynthesis (DBT) at two clinically relevant dose levels. The results showed that increasing the dose improved the conspicuity of microcalcifications, increased the detection sensitivity, confidence levels, and reading efficiency of radiologists, and reduced false positive detections.
ACADEMIC RADIOLOGY
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Lynda C. Ikejimba, Jesse Salad, Christian G. Graff, Mitchell Goodsitt, Heang-Ping Chan, Hailiang Huang, Wei Zhao, Bahaa Ghammraoui, Joseph Y. Lo, Stephen J. Glick
Summary: This study evaluated five commercial DBT systems using an inkjet-printed anthropomorphic phantom and a four alternative forced choice (4AFC) study scheme. The results showed that DBT had the highest overall detection for masses and microcalcifications, with significant differences between DBT and synthetic mammography in most systems. The study also found a relationship between increasing detection performance and greater gantry span.
Article
Medicine, General & Internal
Baptiste Vasey, Myura Nagendran, Bruce Campbell, David A. Clifton, Gary S. Collins, Spiros Denaxas, Alastair K. Denniston, Livia Faes, Bart Geerts, Mudathir Ibrahim, Xiaoxuan Liu, Bilal A. Mateen, Piyush Mathur, Melissa D. McCradden, Lauren Morgan, Johan Ordish, Campbell Rogers, Suchi Saria, Daniel S. W. Ting, Peter Watkinson, Wim Weber, Peter Wheatstone, Peter McCulloch
BMJ-BRITISH MEDICAL JOURNAL
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Yifan Wang, Chuan Zhou, Heang-Ping Chan, Lubomir M. Hadjiiski, Aamer Chughtai, Ella A. Kazerooni
Summary: In this study, a hybrid deep learning model was developed for accurate segmentation of lung nodules with different characteristics. The model combined two deep convolutional neural networks as encoders to improve the segmentation performance of complex lung nodules. The results showed that the hybrid model achieved segmentation accuracy comparable to radiologists' performance and outperformed the individual shallow or deep models.
Correction
Biochemistry & Molecular Biology
Baptiste Vasey, Myura Nagendran, Bruce Campbell, David A. Clifton, Gary S. Collins, Spiros Denaxas, Alastair K. Denniston, Livia Faes, Bart Geerts, Mudathir Ibrahim, Xiaoxuan Liu, Bilal A. Mateen, Piyush Mathur, Melissa D. McCradden, Lauren Morgan, Johan Ordish, Campbell Rogers, Suchi Saria, Daniel S. W. Ting, Peter Watkinson, Wim Weber, Peter Wheatstone, Peter McCulloch
Article
Radiology, Nuclear Medicine & Medical Imaging
Lubomir Hadjiiski, Kenny Cha, Heang-Ping Chan, Karen Drukker, Lia Morra, Janne J. Nappi, Berkman Sahiner, Hiroyuki Yoshida, Quan Chen, Thomas M. Deserno, Hayit Greenspan, Henkjan Huisman, Zhimin Huo, Richard Mazurchuk, Nicholas Petrick, Daniele Regge, Ravi Samala, Ronald M. Summers, Kenji Suzuki, Georgia Tourassi, Daniel Vergara, Samuel G. Armato
Summary: Advances in AI and machine learning, specifically in DL techniques, have made possible the application of these methods in healthcare. However, proper training and rigorous validation of machine learning algorithms are essential before clinical deployment.
Article
Radiology, Nuclear Medicine & Medical Imaging
Heang-Ping Chan, Mark A. Helvie, Mingjie Gao, Lubomir Hadjiiski, Chuan Zhou, Kim Garver, Katherine A. Klein, Carol McLaughlin, Rebecca Oudsema, W. Tania Rahman, Marilyn A. Roubidoux
Summary: Deep-learning-based denoising has the potential to improve the detection of microcalcifications (MCs) in noisy digital breast tomosynthesis (DBT) images and increase radiologists' confidence in differentiating noise from MCs without increasing radiation dose.
Article
Radiology, Nuclear Medicine & Medical Imaging
Aman Kushwaha, Rami F. Mourad, Kevin Heist, Humera Tariq, Heang-Ping Chan, Brian D. Ross, Thomas L. Chenevert, Dariya Malyarenko, Lubomir M. Hadjiiski
Summary: A murine model of myelofibrosis was used to evaluate segmentation methods for image-based biomarkers. The dataset was divided into training, validation, and test subsets, and the accuracy and repeatability of manual and computer segmentations were compared. The deep learning model achieved comparable accuracy and better repeatability than human annotators.
Proceedings Paper
Optics
Mingjie Gao, Jeffrey A. Fessler, Heang-Ping Chan
Summary: This study investigated the feasibility of improving the image quality of digital breast tomosynthesis (DBT) reconstruction by combining a model-based iterative reconstruction (MBIR) method and a deep convolutional neural network based DBT denoiser (DNGAN). The proposed approach, named DBCN+DNGAN, showed improved figures of merit (FOMs) and visually satisfactory soft tissue appearance with low background noise level compared to other reconstruction techniques.
MEDICAL IMAGING 2022: PHYSICS OF MEDICAL IMAGING
(2022)
Proceedings Paper
Engineering, Biomedical
Di Sun, Lubomir M. Hadjiiski, Rohan Garje, Yousef Zakharia, Lauren Pomerantz, Monika Joshi, Ajjai Alva, Heang-Ping Chan, Richard Cohan, Elaine Caoili, Kenny H. Cha, Galina Kirova-Nedyalkova, Matthew S. Davenport, Prasad R. Shankar, Isaac R. Francis, Kimberly Shampain, Nathaniel Meyer, Daniel Barkmeier, Sean Woolen, Phillip L. Palmbos, Alon Z. Weizer, Ravi K. Samala, Chuan Zhou, Martha Matuszak
Summary: This study evaluated the impact of a computerized decision support system (CDSS-T) on bladder cancer treatment response assessment. The results showed that CDSS-T significantly improved the accuracy and consistency of the assessment.
MEDICAL IMAGING 2022: COMPUTER-AIDED DIAGNOSIS
(2022)
Proceedings Paper
Engineering, Biomedical
Yifan Wang, Chuan Zhou, Heang-Ping Chan, Lubomir M. Hadjiiski, Aamer Chughtai
Summary: In this study, we developed DCNN-based methods for lung nodule segmentation. Fusion methods combining different DCNNs and deep learning models were used to improve segmentation accuracy. The results showed that fusion methods utilizing SEAB and CBAM achieved better performance in segmentation.
MEDICAL IMAGING 2022: COMPUTER-AIDED DIAGNOSIS
(2022)
Article
Computer Science, Information Systems
Chuan Zhou, Heang-Ping Chan, Lubomir M. Hadjiiski, Aamer Chughtai
Summary: This study developed a recursive training strategy to improve the detection and segmentation performance of a deep learning model for nuclei using incomplete annotation. The proposed method achieved significant improvement in sensitivity and quality metrics compared to traditional training methods.
Article
Radiology, Nuclear Medicine & Medical Imaging
Di Sun, Lubomir Hadjiiski, Ajjai Alva, Yousef Zakharia, Monika Joshi, Heang-Ping Chan, Rohan Garje, Lauren Pomerantz, Dean Elhag, Richard H. Cohan, Elaine M. Caoili, Wesley T. Kerr, Kenny H. Cha, Galina Kirova-Nedyalkova, Matthew S. Davenport, Prasad R. Shankar, Isaac R. Francis, Kimberly Shampain, Nathaniel Meyer, Daniel Barkmeier, Sean Woolen, Phillip L. Palmbos, Alon Z. Weizer, Ravi K. Samala, Chuan Zhou, Martha Matuszak
Summary: This observer study investigates the impact of a computerized artificial intelligence decision support system on physicians' diagnostic accuracy in assessing bladder cancer treatment response. The study found that the system significantly improved the performance of the physicians and that specific factors of the cancer cases also influenced the diagnostic results.
Proceedings Paper
Engineering, Biomedical
Ravi K. Samala, Lubomir Hadjiiski, Heang-Ping Chan, Chuan Zhou, Jadranka Stojanovska, Prachi Agarwal, Christopher Fung
Summary: The study demonstrates the potential of deep learning in quantitatively assessing the severity of COVID-19 pneumonia, showing high accuracy in level classification and extraction of global descriptors.
MEDICAL IMAGING 2021: COMPUTER-AIDED DIAGNOSIS
(2021)
Proceedings Paper
Engineering, Biomedical
Yifan Wang, Chuan Zhou, Lei Ying, Heang-Ping Chan, Lubomir M. Hadjiiski, Aamer Chughtai, Ella A. Kazerooni
Summary: A radiomic-based reinforcement learning model was developed for early diagnosis of lung cancer, utilizing a Markov decision process to classify lung nodules as malignant or benign. The model learned a policy mapping between patients' clinical conditions and decisions based on expected rewards associated with lung cancer risk. This model, trained with multi-year CT scans, showed potential to improve early diagnosis of lung nodules, reducing unnecessary follow-up exams and costs.
MEDICAL IMAGING 2021: COMPUTER-AIDED DIAGNOSIS
(2021)