Article
Multidisciplinary Sciences
Mohammed A. S. Ali, Kaspar Hollo, Tonis Laasfeld, Jane Torp, Maris-Johanna Tahk, Ago Rinken, Kaupo Palo, Leopold Parts, Dmytro Fishman
Summary: Brightfield cell microscopy is a fundamental tool in life sciences, but the acquired images often contain visual artifacts that can hinder downstream analysis. This study proposes a pipeline called ScoreCAM-U-Net, which can automatically segment and remove these artifacts with limited user input. The model is trained using only image-level labels, making the process significantly faster than traditional pixel-level annotation methods. The study demonstrates the existence of artifacts in different brightfield microscopy image datasets and shows that the automated artifact removal improves downstream analyses. This method has the potential to become a standard step in large scale microscopy experiments.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Artificial Intelligence
Rodrigo Rill-Garcia, Eva Dokladalova, Petr Dokladal
Summary: Recent road crack detection methods achieve high scores in crack localization but still show inaccuracies in crack width measurement. By introducing a synthetic dataset and label noise correction techniques, accuracy and robustness of crack detection can be improved to some extent.
Article
Computer Science, Artificial Intelligence
Adam Byerly, Tatiana Kalganova, Ian Dear
Summary: The study demonstrates that a simple convolutional neural network using Homogeneous Vector Capsules (HVCs) performs as well as previous capsule networks on the MNIST dataset, but with fewer parameters, fewer training epochs, and no routing mechanism required.
Article
Computer Science, Information Systems
Ziyi Huang, Yu Gan, Theresa Lye, Yanchen Liu, Haofeng Zhang, Andrew Laine, Elsa Angelini, Christine Hendon
Summary: Automatically identifying structural substrates underlying cardiac abnormalities can provide real-time guidance for interventional procedures. Our study develops a two-stage deep learning framework for cardiac adipose tissue segmentation using image-level annotations on OCT images, bridging the gap between automatic tissue analysis demand and the lack of pixel-wise annotations.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Interdisciplinary Applications
Junhao Hu, Chenyang Zhang, Kang Zhou, Shenghua Gao
Summary: Chest X-ray is an important imaging method for diagnosing chest diseases. We propose a semantic-segmentation-based solution for chest X-ray diagnostic quality assessment and create a dataset to validate it. Our experiments validate the effectiveness of our solution.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Artificial Intelligence
Jialun Pei, He Tang, Wanru Wang, Tianyang Cheng, Chuanbo Chen
Summary: In this paper, a salient instance segmentation model trained by weak supervisions is proposed, which makes use of existing salient object detection datasets and combines salient regions and bounding boxes for supervision. With the global feature refining layer and labeling updating scheme, the model can accurately locate salient instances. Extensive experiments demonstrate the competitiveness of the proposed model trained by weak labels with the existing fully-supervised state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Jinyang Liu, Shutao Li, Haibo Liu, Renwei Dian, Xiaohui Wei
Summary: In this paper, a lightweight pixel-level unified image fusion (L-PUIF) network is proposed to achieve more efficient and accurate image fusion. The experimental results show that L-PUIF achieves better fusion efficiency and has practicality in high-level computer vision tasks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Multidisciplinary Sciences
Xiujuan Zou, Youming Zhang, Ruoyu Lin, Guangxing Gong, Shuming Wang, Shining Zhu, Zhenlin Wang
Summary: High-performance metasurface-based Bayer-type color routers are realized using the inverse-design method, achieving higher color collection efficiency and energy utilization efficiency for color imaging applications.
NATURE COMMUNICATIONS
(2022)
Article
Automation & Control Systems
Xubing Yang, Run Chen, Fuquan Zhang, Li Zhang, Xijian Fan, Qiaolin Ye, Liyong Fu
Summary: The proposed automatic annotation method for forest fire images introduces supervised information through interactive convex hulls at the pixel level, allowing visually selecting irregular fire and no-fire regions. It expands the view of fire detection to multi-class classification and loosens the constraints of i.i.d hypothesis in machine learning methods. Experimental evaluations show that it achieves higher fire detection rate and lower false alarm rate compared to state-of-the-art methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2021)
Article
Optics
Zhi Zhang, Tingbiao Guo, Zijian Lin, Zhenchao Liu, Nan He, Xinan Xu, Yuxin Xing, Dong Wang, Ying Li, Yi Jin, Sailing He
Summary: Reflective structural colors offer backlight-free and energy-efficient color production, but face challenges in complex fabrication and color accuracy. A new pixelated electrothermal oxidation technique is proposed to precisely tailor reflective structural colors, enabling customized color filter arrays. By thermal engineering, a photonic-firework-like color filter array is achieved in a single step, and a computational spectrometer with a resolution of 10 nm is realized. With its compactness and mass production capability, this method has potential applications in imaging, anti-counterfeiting, printing, color display, and spectroscopy.
LASER & PHOTONICS REVIEWS
(2023)
Article
Computer Science, Artificial Intelligence
Jiaxu Miao, Yunchao Wei, Xiaohan Wang, Yi Yang
Summary: This study introduces a large-scale video scene parsing dataset, VSPW, which contains 251,633 frames from 3,536 videos with pixel-level annotations. It covers 231 scenes and 124 object categories. The study also proposes a Temporal Attention Blending (TAB) network to improve pixel-level semantic understanding of videos. Experimental results demonstrate the superiority of the TAB approach on the VSPW dataset.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Cristina Gonzalez, Nicolas Ayobi, Isabela Hernandez, Jordi Pont-Tuset, Pablo Arbelaez
Summary: This paper proposes a new framework and algorithm for the Panoptic Narrative Grounding task, achieving significant results in the experiments and demonstrating its competitiveness and practicality.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Geochemistry & Geophysics
Zhuo Chen, Lingjun Zhao, Qishan He, Gangyao Kuang
Summary: This paper proposes a pixel-level and feature-level domain adaptation approach to enhance the performance of target recognition in heterogeneous SAR situations.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Chemistry, Multidisciplinary
Haotian Li, Zhuang Yue, Jingyu Liu, Yi Wang, Huaiyu Cai, Kerang Cui, Xiaodong Chen
Summary: This paper proposed a pixel-level crack segmentation network called SCCDNet based on convolutional neural networks, which achieved the best crack segmentation performance with an F-score of 0.7763 by using techniques such as depthwise separable convolution. The network was trained and tested on a dataset containing cracks collected in different scenes, demonstrating its effectiveness in detecting cracks accurately and efficiently.
APPLIED SCIENCES-BASEL
(2021)
Article
Multidisciplinary Sciences
Yinqi Wang, Kun Huang, Jianan Fang, Ming Yan, E. Wu, Heping Zeng
Summary: Single-pixel cameras offer a promising alternative to multi-pixel sensors for mid-infrared (MIR) imaging. In this study, a MIR single-photon computational imaging method with a single-element silicon detector is demonstrated, employing nonlinear structured detection and upconversion detection. Advanced algorithms enable MIR image reconstruction under sub-Nyquist sampling and photon-starving illumination. This single-pixel upconversion imaging paradigm provides simplicity, single-photon sensitivity, and room-temperature operation, opening up new possibilities for sensitive imaging at longer infrared wavelengths or terahertz frequencies.
NATURE COMMUNICATIONS
(2023)
Review
Pathology
Joep M. A. Bogaerts, Miranda P. Steenbeek, Majke H. D. van Bommel, Johan Bulten, Jeroen A. W. M. van der Laak, Joanne A. de Hullu, Michiel Simons
Summary: The understanding of oncogenesis of high-grade serous cancer of the ovary and its precursor lesions has increased significantly over the last few decades, with a focus on diagnosing serous tubal intraepithelial carcinoma (STIC). Pathologists face challenges in diagnosing STIC, and the study provides insights on diagnostic features and a framework for arriving at an adequate diagnosis, emphasizing the importance of grossing protocols, subspecialized evaluation, immunohistochemical staining, and seeking a second opinion.
Article
Multidisciplinary Sciences
Michela Antonelli, Annika Reinke, Spyridon Bakas, Keyvan Farahani, Annette Kopp-Schneider, Bennett A. Landman, Geert Litjens, Bjoern Menze, Olaf Ronneberger, Ronald M. Summers, Bram van Ginneken, Michel Bilello, Patrick Bilic, Patrick F. Christ, Richard K. G. Do, Marc J. Gollub, Stephan H. Heckers, Henkjan Huisman, William R. Jarnagin, Maureen K. McHugo, Sandy Napel, Jennifer S. Golia Pernicka, Kawal Rhode, Catalina Tobon-Gomez, Eugene Vorontsov, James A. Meakin, Sebastien Ourselin, Manuel Wiesenfarth, Pablo Arbelaez, Byeonguk Bae, Sihong Chen, Laura Daza, Jianjiang Feng, Baochun He, Fabian Isensee, Yuanfeng Ji, Fucang Jia, Ildoo Kim, Klaus Maier-Hein, Dorit Merhof, Akshay Pai, Beomhee Park, Mathias Perslev, Ramin Rezaiifar, Oliver Rippel, Ignacio Sarasua, Wei Shen, Jaemin Son, Christian Wachinger, Liansheng Wang, Yan Wang, Yingda Xia, Daguang Xu, Zhanwei Xu, Yefeng Zheng, Amber L. Simpson, Lena Maier-Hein, M. Jorge Cardoso
Summary: This article presents the results of a biomedical image segmentation challenge and suggests that a method capable of performing well on multiple tasks will also generalize well to previously unseen tasks. By organizing the Medical Segmentation Decathlon (MSD), the study confirms that state-of-the-art image segmentation algorithms can generalize well when retrained on unseen tasks, and that consistent performance across multiple tasks is a strong indicator of algorithmic generalizability.
NATURE COMMUNICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Jasper Linmans, Stefan Elfwing, Jeroen van der Laak, Geert Litjens
Summary: This study evaluates prevalent methods on large-scale digital pathology datasets and provides a benchmark. The results show that different methods perform differently in detecting different out-of-distribution data in the medical imaging domain. The study also demonstrates the harmful impact of unknown data on the performance of machine learning models.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Computer Science, Artificial Intelligence
Marc Aubreville, Nikolas Stathonikos, Christof A. Bertram, Robert Klopfleisch, Natalie ter Hoeve, Francesco Ciompi, Frauke Wilm, Christian Marzahl, Taryn A. Donovan, Andreas Maier, Jack Breen, Nishant Ravikumar, Youjin Chung, Jinah Park, Ramin Nateghi, Fattaneh Pourakpour, Rutger H. J. Fick, Saima Ben Hadj, Mostafa Jahanifar, Adam Shephard, Jakob Dexl, Thomas Wittenberg, Satoshi Kondo, Maxime W. Lafarge, Viktor H. Koelzer, Jingtang Liang, Yubo Wang, Xi Long, Jingxin Liu, Salar Razavi, April Khademi, Sen Yang, Xiyue Wang, Ramona Erber, Andrea Klang, Karoline Lipnik, Pompei Bolfa, Michael J. Dark, Gabriel Wasinger, Mitko Veta, Katharina Breininger
Summary: The density of mitotic figures (MF) in tumor tissue is an important marker for tumor grading, but its recognition by pathologists is biased and limited. Deep learning methods can support the recognition, but their performance deteriorates in different clinical environments due to variability caused by using different whole slide scanners. The MICCAI MIDOG 2021 challenge aimed to develop scanner-agnostic MF detection algorithms and the winning algorithm outperformed six experts on the same task.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Cell Biology
Joep M. A. Bogaerts, Majke H. D. van Bommel, Rosella P. M. G. Hermens, Miranda P. Steenbeek, Joanne A. de Hullu, Jeroen A. W. M. van Der Laak, Michiel Simons
Summary: It is crucial to diagnose or exclude serous tubal intraepithelial carcinoma (STIC), a precursor lesion of tubo-ovarian high-grade serous carcinoma (HGSC), in a reliable and safe manner. This study aimed to optimize STIC diagnosis and increase reproducibility through a three-round Delphi study involving an international panel of expert gynecologic pathologists. The resulting consensus statements provide recommendations for more consistent STIC diagnosis.
Article
Computer Science, Artificial Intelligence
Peter Bandi, Maschenka Balkenhol, Marcory van Dijk, Michel Kok, Bram van Ginneken, Jeroen van der Laak, Geert Litjens
Summary: The researchers investigated how to efficiently utilize existing high-quality datasets in multi-task settings and strategies such as prevention of catastrophic forgetting for breast, colon, and head-and-neck cancer metastasis detection in lymph nodes. The results showed state-of-the-art performance in colon and head-and-neck cancer metastasis detection tasks. They also demonstrated the effectiveness of adapting networks from one cancer type to another to obtain multi-task metastasis detection networks, and that leveraging existing high-quality datasets can significantly improve performance on new target tasks while mitigating catastrophic forgetting.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Cell Biology
Leander van Eekelen, Geert Litjens, Konnie Hebeda
Summary: The increasing digitalization of routine diagnostic histological slides provides the potential to apply artificial intelligence (AI) in pathology, specifically in bone marrow histology. This perspective explores the potential tasks, investigations, and questions that can be addressed by applying AI to whole slide images of bone marrow biopsies. The discussion includes examples of current AI research using bone marrow biopsy slides and briefly touches upon the challenges of implementing AI-supported diagnostics.
Article
Multidisciplinary Sciences
John-Melle Bokhorst, Iris D. Nagtegaal, Filippo Fraggetta, Simona Vatrano, Wilma Mesker, Michael Vieth, Jeroen van der Laak, Francesco Ciompi
Summary: In this study, artificial intelligence was used to assist in the characterization and reporting of colorectal cancer biopsy samples. A segmentation method was developed to accurately depict tissue morphology and composition. This tool can support pathologists in risk stratification and has potential for other applications.
SCIENTIFIC REPORTS
(2023)
Article
Oncology
Ananda van der Kamp, Thomas de Bel, Ludo van Alst, Jikke Rutgers, Marry M. van den Heuvel-Eibrink, Annelies M. C. Mavinkurve-Groothuis, Jeroen van der Laak, Ronald R. de Krijger
Summary: Wilms tumor (WT) is a highly variable pediatric tumor, and artificial intelligence (AI) can contribute to accurate and reproducible histopathological assessment of WT through recognition of individual histopathological tumor components. A deep learning (DL) system was used to recognize 15 different normal and tumor components with high accuracy, showing potential for future automated WT classification.
Article
Oncology
John-Melle Bokhorst, Iris D. Nagtegaal, Inti Zlobec, Heather Dawson, Kieran Sheahan, Femke Simmer, Richard Kirsch, Michael Vieth, Alessandro Lugli, Jeroen van der Laak, Francesco Ciompi
Summary: Tumor budding is a histopathological biomarker associated with metastases and adverse survival outcomes in colorectal carcinoma (CRC) patients. We present a deep learning convolutional neural network model that automates the tumor budding detection and counting process, improving efficiency and reproducibility. We also explore novel metrics to quantify buds such as density and dispersion and report their prognostic value.
Article
Computer Science, Artificial Intelligence
Stephan Dooper, Hans Pinckaers, Witali Aswolinskiy, Konnie Hebeda, Sofia Jarkman, Jeroen van der Laak, Geert Litjens, BIGPICTURE Consortium
Summary: This paper proposes a method called StreamingCLAM, which trains a convolutional neural network on gigapixel images and achieves good results. The model is able to detect metastatic breast cancer and MYC-gene translocation in B-cell lymphoma, and offers interpretability through the attention mechanism.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Biochemistry & Molecular Biology
Julie E. M. Swillens, Iris D. Nagtegaal, Sam Engels, Alessandro Lugli, Rosella P. M. G. Hermens, Jeroen A. W. M. van der Laak
Summary: Computational pathology (CPath) algorithms can detect, segment, or classify cancer in whole slide images with accuracy comparable to or higher than pathologists. However, there are challenges to be overcome before these algorithms can be used in practice. International perspectives were explored through an eSurvey and interviews, revealing variations in opinions and experiences regarding barriers and facilitators. This diversity highlights the need for further quantitative research to determine important factors for adoption and initiate validation studies.