Article
Optics
Xinliang Zhai, Xiaoyan Wu, Yiwei Sun, Jianhong Shi, Guihua Zeng
Summary: This method utilizes unsupervised deep learning for noise reduction, enabling image reconstruction under challenging conditions, particularly suitable for biomedical imaging, with improved imaging speed and accuracy while maintaining high quality.
Article
Radiology, Nuclear Medicine & Medical Imaging
Yan-Ping Xue, Hyungseok Jang, Michal Byra, Zhen-Yu Cai, Mei Wu, Eric Y. Chang, Ya-Jun Ma, Jiang Du
Summary: The study developed a fully automated method for segmenting and mapping full-thickness cartilage using quantitative 3D ultrashort echo time MR imaging and a U-Net convolutional neural networks model. The method showed reliable results and could provide a comprehensive assessment of articular cartilage. Significant changes in quantitative biomarkers were observed in participants with osteoarthritis compared to normal controls.
EUROPEAN RADIOLOGY
(2021)
Article
Computer Science, Information Systems
Yuyang Liu, Suvodeep Mazumdar, Peter A. Bath
Summary: This study used statistical and unsupervised learning methods to identify scans of people with Alzheimer's disease, which can diagnose the disease accurately without requiring large amounts of labelled MRI scans. The research could help in the automatic diagnosis of Alzheimer's disease and provide a basis for diagnosing stable mild cognitive impairment and progressive mild cognitive impairment.
INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS
(2023)
Article
Medicine, General & Internal
Estefania Nunez, Valentin Fuster, Maria Gomez-Serrano, Jose Manuel Valdivielso, Juan Miguel Fernandez-Alvira, Diego Martinez-Lopez, Jose Manuel Rodriguez, Elena Bonzon-Kulichenko, Enrique Calvo, Alvaro Alfayate, Marcelino Bermudez-Lopez, Joan Carles Escola-Gil, Leticia Fernandez-Friera, Isabel Cerro-Pardo, Jose Maria Mendiguren, Fatima Sanchez-Cabo, Javier Sanz, Jose Maria Ordovas, Luis Miguel Blanco-Colio, Jose Manuel Garcia-Ruiz, Borja Ibanez, Enrique Lara-Pezzi, Antonio Fernandez-Ortiz, Jose Luis Martin-Ventura, Jesus Vazquez
Summary: This study identified circulating proteins that are associated with subclinical atherosclerosis and can predict the disease. These proteins offer potential for improving primary prevention strategies in areas where cardiovascular imaging is not available.
Article
Computer Science, Interdisciplinary Applications
Piyush Agarwal, Mohammad Aghaee, Melih Tamer, Hector Budman
Summary: This study proposes an unsupervised Statistical Process Control method based on deep learning, using a Multiway Partial Least Squares Autoencoder model trained with a genetic optimization algorithm to maximize fault detection rate. The effectiveness of this method is demonstrated on an industrial scale Penicillin process, and it outperforms linear fault detection algorithms when compared. The use of dynamic control limits significantly improves the detection rates for both linear and deep learning models.
COMPUTERS & CHEMICAL ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Mohamad Dhaini, Maxime Berar, Paul Honeine, Antonin Van Exem
Summary: This paper studies the problem of unsupervised domain adaptation for regression tasks and proposes a new approach based on dictionary learning. Experimental results show that the proposed method outperforms most of state-of-the-art methods on several benchmark datasets, especially when transferring knowledge from synthetic to real domains.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Genetics & Heredity
Ge Zhang, Zijing Xue, Chaokun Yan, Jianlin Wang, Huimin Luo
Summary: Gastric cancer, a type of complex disease with high mortality and limited effective treatments, is studied using gene expression and DNA methylation data to identify potential biomarkers through a feature selection approach. The study demonstrates superior performance in classification accuracy and further validates the effectiveness through biological analysis of selected genes.
FRONTIERS IN GENETICS
(2021)
Article
Engineering, Electrical & Electronic
Zhijun He, Hongbo Zhao, Jianrong Wang, Wenquan Feng
Summary: In this paper, a multi-level progressive learning (MLPL) method is proposed for unsupervised vehicle re-identification (ReID), achieving good performance by utilizing only unlabeled target domain images. A multi-branch architecture is introduced to explore vehicle representations in different levels, while a density-based clustering method generates pseudo labels. A novel re-clustering method is also proposed to better mine the labels with high reliability. Furthermore, a dynamic progressive contrast learning (DPCL) strategy is designed to train the network based on the clustered labels. Comprehensive experiments on mainstream evaluation datasets show that our approach outperforms other existing unsupervised methods.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2023)
Article
Engineering, Electrical & Electronic
Yong Xu, Baoling Liu, Yuhui Quan, Hui Ji
Summary: This paper proposes a dataset-free unsupervised deep learning-based approach for background matting, which models the foreground and alpha matte using the priors encoded by two generative convolutional neural networks. The proposed method achieves competitive performance to recent supervised learning-based methods even without calling external training data.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Yi Zheng, Yong Zhou, Jiaqi Zhao, Ying Chen, Rui Yao, Bing Liu, Abdulmotaleb El Saddik
Summary: In person re-identification, the cost of supervised learning is high, making unsupervised methods more suitable. The key to solving the problem lies in finding a standard that effectively distinguishes the differences in image features between different pedestrian identities. The distance between features plays a crucial role in unsupervised learning and a deep learning method based on hierarchical clustering is improved by considering both feature and distance metrics.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Yan Xia, Nishant Ravikumar, John P. Greenwood, Stefan Neubauer, Steffen E. Petersen, Alejandro F. Frangi
Summary: The study presents a novel super-resolution algorithm based on conditional generative adversarial networks for generating high-quality cardiac MR images, which benefits subsequent image analyses and demonstrates superior performance in experiments.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Optics
Ye Tian, Ying Fu, Jun Zhang
Summary: In this paper, we propose an effective method based on channel attention convolutional neural network for under-sampled single-pixel imaging (SPI) to reconstruct high-quality object images directly from SPI measurements. The method takes advantage of unsupervised deep learning and effectively avoids over-fitting problem using SPI model constraint and total variation regularization. Extensive experimental results on simulation and real data demonstrate that the proposed method has superior performance in image quality, noise robustness, and generalization compared with the state-of-the-art SPI methods.
OPTICS AND LASER TECHNOLOGY
(2023)
Article
Cell Biology
Foo Wei Ten, Dongsheng Yuan, Nabil Jabareen, Yin Jun Phua, Roland Eils, Soeren Lukassen, Christian Conrad
Summary: Feature identification and manual inspection are still important in single-cell sequencing data analysis. We propose using ensembles of autoencoders and rank aggregation to extract consistent features in a less biased manner. Our method can complement conventional tools and work with overlapping clustering identity assignment for transitional cell types or fates.
FRONTIERS IN CELL AND DEVELOPMENTAL BIOLOGY
(2023)
Article
Chemistry, Multidisciplinary
Jeongwoo Ju, Heechul Jung, Junmo Kim
Summary: Modern DNN-based approaches have achieved impressive performance in computer vision tasks, but the need for extensive annotation imposes a high cost. Identifying and eliminating redundant examples can improve performance while reducing annotation requirements.
APPLIED SCIENCES-BASEL
(2022)
Article
Biochemical Research Methods
Shanshan Wang, Ruoyou Wu, Cheng Li, Juan Zou, Ziyao Zhang, Qiegen Liu, Yan Xi, Hairong Zheng
Summary: This paper proposes a Physics-based unsupervised Contrastive Representation Learning (PARCEL) method to speed up parallel MR imaging. PARCEL captures the inherent features and representations for MR images by contrastively learning two branches of model-based unrolling networks, enabling accurate MR reconstruction without relying on fully sampled datasets.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Ophthalmology
Janice Sutton, Martin J. Menten, Sophie Riedl, Hvroje Bogunovic, Oliver Leingang, Philipp Anders, Ahmed M. Hagag, Sebastian Waldstein, Amber Wilson, Angela J. Cree, Ghislaine Traber, Lars G. Fritsche, Hendrik Scholl, Daniel Rueckert, Ursula Schmidt-Erfurth, Sobha Sivaprasad, Toby Prevost, Andrew Lotery
Summary: This study aims to utilize machine learning and advanced statistical modeling to discover biomarkers for disease progression in intermediate age-related macular degeneration (AMD) and report the natural history of its progression through multimodal retinal imaging.
Correction
Ophthalmology
Janice Sutton, Martin J. Menten, Sophie Riedl, Hrvoje Bogunovic, Oliver Leingang, Philipp Anders, Ahmed M. Hagag, Sebastian Waldstein, Amber Wilson, Angela J. Cree, Ghislaine Traber, Lars G. Fritsche, Hendrik Scholl, Daniel Rueckert, Ursula Schmidt-Erfurth, Sobha Sivaprasad, Toby Prevost, Andrew Lotery
Article
Ophthalmology
Martin Michl, Martina Neschi, Alexandra Kaider, Katja Hatz, Gabor Deak, Bianca S. S. Gerendas, Ursula Schmidt-Erfurth
Summary: The objective of this study was to assess agreement in evaluating OCT variables in leading macular diseases among OCT-certified graders. Even in optimized conditions, there is disease-dependent variability in biomarker evaluation, particularly for IRF in nAMD and DMO. Our findings highlight the variability in human expert OCT grading performance and the need for AI-based automated feature analyses.
Article
Computer Science, Information Systems
Botond Fazekas, Dmitrii Lachinov, Guilherme Aresta, Julia Mai, Ursula Schmidt-Erfurth, Hrvoje Bogunovic
Summary: Segmentation of Bruch's membrane (BM) on optical coherence tomography (OCT) is crucial for the diagnosis and follow-up of age-related macular degeneration (AMD), a leading cause of blindness. Existing automated methods lack anatomical coherence and confidence feedback, limiting their real-world applicability. To address this, we propose an end-to-end deep learning method that uses an Attention U-Net to output a probability density function and considers the natural curvature of the surface. Additionally, our method estimates uncertainty and interpolates A-scans with high uncertainty. Evaluation on internal and external datasets demonstrates superior performance and strong generalization ability.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Ophthalmology
Gregor S. Reiter, Hrvoje Bogunovic, Ferdinand Schlanitz, Wolf-Dieter Vogl, Philipp Seeboeck, Dariga Ramazanova, Ursula Schmidt-Erfurth
Summary: This study evaluated the quantitative impact of drusen and hyperreflective foci (HRF) volumes on mesopic retinal sensitivity in age-related macular degeneration (AMD) patients. The results showed a negative correlation between the volume of drusen and HRF and retinal sensitivity. The study highlights the use of AI-based methods to assess the impact of cellular changes on the development of AMD.
Article
Ophthalmology
Maximilian Pawloff, Bianca S. S. Gerendas, Gabor Deak, Hrvoje Bogunovic, Anastasiia Gruber, Ursula Schmidt-Erfurth
Summary: The purpose of this study was to evaluate the reliability of automated fluid detection in identifying retinal fluid activity in OCT scans of patients treated with anti-VEGF therapy for neovascular age-related macular degeneration. The results showed that the concordance between human expert grading and automated algorithm performance was high, indicating that deep learning-based segmentation of retinal fluid performs reliably on OCT images.
Article
Computer Science, Interdisciplinary Applications
M. Niederleithner, L. de Sisternes, H. Stino, A. Sedova, T. Schlegl, H. Bagherinia, A. Britten, P. Matten, U. Schmidt-Erfurth, A. Pollreisz, W. Drexler, R. A. Leitgeb, T. Schmoll
Summary: Optical Coherence Tomography Angiography (OCTA) has the potential to replace invasive fluorescein angiography (FA) in ophthalmology, but it still lacks the field of view compared to fluorescence fundus photography techniques. To address this issue, a custom developed high-speed swept-source OCT (SS-OCT) system is presented, which can capture ultra-wide fields of view up to 90 degrees with high resolution. Additionally, a three-dimensional deep learning based algorithm is developed for denoising volumetric OCTA data sets, enhancing the visual appearance of angiograms.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Biochemical Research Methods
David Rivas-Villar, Alice R. Motschi, Michael Pircher, Christoph K. Hitzenberger, Markus Schranz, Philipp K. Roberts, Ursula Schmidt-Erfurth, Hrvoje Bogunovic
Summary: In this study, a novel automated pipeline for registering OCT images from different devices is proposed. The pipeline utilizes deep learning for multi-modal 2D en-face registration and retinal layer segmentation for Z-axis registration. The experimental results show high-quality registrations, with mean errors of approximately 46 μm for 2D registration and 9.59 μm for Z-axis registration. This registration method can be valuable for various clinical applications such as layer segmentation validation.
BIOMEDICAL OPTICS EXPRESS
(2023)
Article
Ophthalmology
Leonard M. Coulibaly, Gregor S. Reiter, Philipp Fuchs, Dmitrii Lachinov, Oliver Leingang, Wolf-Dieter Vogl, Hrvoje Bogunovic, Ursula Schmidt-Erfurth
Summary: This study investigated the progression of geographic atrophy secondary to nonneovascular age-related macular degeneration using artificial intelligence-based precision tools. The findings revealed that early lesions have slower growth rates and there are differences in growth dynamics compared to atrophic lesions in advanced stages.
OPHTHALMOLOGY RETINA
(2023)
Article
Ophthalmology
Marlene Hollaus, Michael Georgopoulos, Johannes Iby, Jonas Brugger, Oliver Leingang, Hrvoje Bogunovic, Ursula Schmidt-Erfurth, Stefan Sacu
Summary: This study aimed to analyze the short-term changes of mean photoreceptor thickness (PRT) on the ETDRS-grid after vitrectomy and membrane peeling in patients with epiretinal membrane (ERM). The results showed significant changes in PRT after surgery, but no correlation with BCVA was found.
Article
Multidisciplinary Sciences
Teresa Araujo, Guilherme Aresta, Ursula Schmidt-Erfurth, Hrvoje Bogunovic
Summary: This study performs a comparative analysis of uncertainty estimation methods for out-of-distribution (OOD) detection in automated screening and staging of age-related macular degeneration (AMD) using retinal OCT imaging. The combination of cosine distance in the feature space and a few-shot outlier exposure (OE) approach improves the near-OOD detection performance, providing robust and reliable diagnostic systems.
SCIENTIFIC REPORTS
(2023)
Article
Ophthalmology
Reinhard Told, Judith Kreminger, Ursula Schmidt-Erfurth, Roman Dunavoelgyi, Adrian Reumueller
Summary: This study investigated the relationship between choroidal melanoma characteristics and progression-free survival (PFS) in patients who underwent linear accelerator-based hypofractionated stereotactic photon radiotherapy. The results showed that thickness and largest basal diameter (LBD) were significantly associated with PFS. For monitoring treated choroidal melanoma, LBD assessments could be used instead of ultrasonography-based thickness measurements.
OPHTHALMOLOGY AND THERAPY
(2023)
Letter
Ophthalmology
Ursula Schmidt-Erfurth, Wolf-Dieter Vogl, Sophie Riedl, Julia Mai, Gregor S. Reiter, Dmitrii Lachinov, Hrvoje Bogunovic
OPHTHALMOLOGY RETINA
(2023)