Review
Chemistry, Analytical
Degaga Wolde Feyisa, Yehualashet Megersa Ayano, Taye Girma Debelee, Friedhelm Schwenker
Summary: Pulmonary tuberculosis (PTB) is a bacterial infection that affects the lung, and its diagnosis often relies on chest radiography. Deep learning (DL) techniques have been proposed to detect and mark areas of tuberculosis infection in chest X-rays, but fully supervised semantic segmentation is challenging due to the need for pixel-level labeled images. Weakly supervised localization techniques are gaining interest in localizing tuberculosis radiographic manifestations in chest X-rays.
Article
Computer Science, Artificial Intelligence
Daniel Moris, Joaquim de Moura, Jorge Novo, Marcos Ortega
Summary: Tuberculosis is an infectious disease that mainly affects the lung tissues. Using Generative Adversarial Network models for image generation can improve tuberculosis screening performance. Experimental results demonstrate that this method outperforms traditional approaches.
PATTERN RECOGNITION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Ines Feki, Sourour Ammar, Yousri Kessentini, Khan Muhammad
Summary: The COVID-19 pandemic has led to a need for efficient diagnosis methods, with deep learning proving to be valuable in analyzing chest X-ray images. This study introduces a collaborative federated learning framework for medical institutions to screen COVID-19 without sharing patient data, showing competitive results compared to traditional data-sharing models. By addressing privacy concerns and utilizing private data, this framework allows for the rapid development of powerful models for COVID-19 screening.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Subhrajit Dey, Rajarshi Roychoudhury, Samir Malakar, Ram Sarkar
Summary: Early detection of Tuberculosis is crucial in reducing mortality rates by preventing its spread to other body parts, and researchers are working on developing a computerized decision support system for efficient diagnosis.
APPLIED SOFT COMPUTING
(2022)
Article
Multidisciplinary Sciences
Robbie Sadre, Baskaran Sundaram, Sharmila Majumdar, Daniela Ushizima
Summary: This paper introduces a set of protocols to validate deep learning algorithms, focusing on emphasizing or hiding key regions of interest from CXR data to evaluate the classification performance for anomaly detection and its correlation to radiological signatures. Through a series of systematic tests, the weaknesses of current techniques are demonstrated, with perspectives on the advantages and limitations of automated radiography analysis when using heterogeneous data sources.
SCIENTIFIC REPORTS
(2021)
Review
Health Care Sciences & Services
K. C. Santosh, Siva Allu, Sivaramakrishnan Rajaraman, Sameer Antani
Summary: There has been significant progress in using machine learning techniques to analyze chest X-ray images for screening cardiopulmonary abnormalities over the past decade. In particular, there has been a strong interest in tuberculosis (TB) screening. This interest aligns with the advances in deep learning, specifically convolutional neural networks (CNNs). This article reviews the research studies published from 2016 to 2021, highlighting data collections, methodological contributions, promising methods, and challenges in TB screening using CXR images.
JOURNAL OF MEDICAL SYSTEMS
(2022)
Article
Immunology
Megan Palmer, Kenneth S. Gunasekera, Marieke M. van der Zalm, Julie Morrison, H. Simon Schaaf, Pierre Goussard, Anneke C. Hesseling, Elisabetta Walters, James A. Seddon
Summary: This article identifies key chest radiograph features that strongly support the diagnosis of confirmed intrathoracic tuberculosis in children.
CLINICAL INFECTIOUS DISEASES
(2022)
Article
Geriatrics & Gerontology
Jin Ryu, Sujeong Eom, Hyeon Chang Kim, Chang Oh Kim, Yumie Rhee, Seng Chan You, Namki Hong
Summary: This study aimed to develop a deep learning model based on chest X-rays to predict the presence of sarcopenia. The results showed that the model performed well in predicting muscle parameters and sarcopenia.
JOURNAL OF CACHEXIA SARCOPENIA AND MUSCLE
(2023)
Article
Computer Science, Information Systems
Vo Trong Quang Huy, Chih-Min Lin
Summary: This paper proposes a new deep learning model, CBAMWDnet, for detecting tuberculosis in chest X-ray images. The model combines the Convolutional Block Attention Module (CBAM) and the Wide Dense Net (WDnet) architecture to effectively capture spatial and contextual information in the images. Evaluation on a large dataset shows that the proposed model outperforms other models in terms of accuracy, sensitivity, precision, specificity, and F1 score. The model also demonstrates strong generalization ability, performing consistently well on different datasets.
Article
Computer Science, Artificial Intelligence
Saad Nafisah, Ghulam Muhammad
Summary: This study proposes an automatic tuberculosis detection system based on deep learning models. By using advanced segmentation networks to extract regions of interest from CXR images and employing different CNN models for classification, the system achieves high accuracy in TB detection.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Interdisciplinary Applications
Jianpeng Zhang, Yutong Xie, Guansong Pang, Zhibin Liao, Johan Verjans, Wenxing Li, Zongji Sun, Jian He, Yi Li, Chunhua Shen, Yong Xia
Summary: Clusters of viral pneumonia cases over a short period may indicate an outbreak. The study proposes a method to detect viral pneumonia using chest X-rays, which performs well on two different datasets without fine-tuning.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Immunology
Gamuchirai Tavaziva, Miriam Harris, Syed K. Abidi, Coralie Geric, Marianne Breuninger, Keertan Dheda, Aliasgar Esmail, Monde Muyoyeta, Klaus Reither, Arman Majidulla, Aamir J. Khan, Jonathon R. Campbell, Pierre-Marie David, Claudia Denkinger, Cecily Miller, Ruvandhi Nathavitharana, Madhukar Pai, Andrea Benedetti, Faiz Ahmad Khan
Summary: This study found that the accuracy of commercially available deep learning-based CAD for detecting tuberculosis varied between different populations, suggesting the need for tailored application based on specific patient characteristics.
CLINICAL INFECTIOUS DISEASES
(2022)
Article
Computer Science, Artificial Intelligence
Fabricio Aparecido Breve
Summary: This paper investigates the use of convolutional neural networks (CNNs) for identifying COVID-19 in chest X-ray images. By testing 21 different CNN architectures and employing ensemble methods, the study achieves superior results compared to previous research.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Rajat Mehrrotraa, M. A. Ansari, Rajeev Agrawal, Pragati Tripathi, Md Belal Bin Heyat, Mohammed Al-Sarem, Abdullah Yahya Mohammed Muaad, Wamda Abdelrahman Elhag Nagmeldin, Abdelzahir Abdelmaboud, Faisal Saeed
Summary: This paper proposes a method of using deep convolutional networks and machine learning algorithms to detect tuberculosis with low computational resources and basic imaging requirements. The model achieved high accuracy in identifying TB infected images and can be beneficial in identifying TB infections during the COVID-19 pandemic.
Article
Computer Science, Artificial Intelligence
Daniel Iglesias Moris, Jose Joaquim de Moura Ramos, Jorge Novo Bujan, Marcos Ortega Hortas
Summary: The current COVID-19 pandemic has caused more than 100 million cases and over two million deaths worldwide, urging the need for rapid and accurate diagnostic methods. Utilizing chest X-ray imaging can explore pathological structures, with portable devices being recommended over conventional fixed machinery. The subjectivity and fatigue of clinicians pose challenges in diagnosis, but computer-aided methodologies can enhance accuracy.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Optics
Christian Brunner, Andreas Duensing, Christian Schroeder, Michael Mittermair, Vladimir Golkov, Maximilian Pollanka, Daniel Cremers, Reinhard Kienberger
Summary: In this study, deep neural networks are applied to solve the challenge of information extraction from spectrograms recorded with the attosecond streak camera in time-resolved photoelectron spectroscopy. Extensive benchmarking on simulated data shows that the deep neural networks exhibit competitive retrieval quality and superior tolerance against noisy data conditions.
Article
Robotics
Lukas von Stumberg, Daniel Cremers
Summary: We present a monocular visual-inertial odometry system based on delayed marginalization and pose graph bundle adjustment. By delaying marginalization, we can obtain updated marginalization prior and new linearization points, and inject IMU information into marginalized states. Our system outperforms existing techniques in visual-inertial odometry.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Hartmut Bauermeister, Emanuel Laude, Thomas Moellenhoff, Michael Moeller, Daniel Cremers
Summary: Dual decomposition approaches in nonconvex optimization often encounter duality gaps. This paper eliminates the duality gap by reformulating the nonconvex task in the space of measures and approximating the infinite-dimensional problem using a piecewise polynomial discretization in the dual. The approach successfully reduces the duality gap and demonstrates scalability in the stereo matching problem.
SIAM JOURNAL ON IMAGING SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Zhenzhang Ye, Bjoern Haefner, Yvain Queau, Thomas Moellenhoff, Daniel Cremers
Summary: This paper discusses the formulation of imaging and low-level vision problems as nonconvex variational problems and proposes convex relaxation methods to solve them. It extends a previous conference paper by introducing product-space relaxation and sublabel-accurate discretization, and demonstrates the use of a cutting-plane method to solve the resulting semi-infinite optimization problem. The journal version includes additional experiments, a more detailed algorithm outline, and a user-friendly introduction to functional lifting methods.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Hamid Rezatofighi, Tianyu Zhu, Roman Kaskman, Farbod T. Motlagh, Javen Qinfeng Shi, Anton Milan, Daniel Cremers, Laura Leal-Taixe, Ian Reid
Summary: This paper addresses the task of set prediction using deep feed-forward neural networks. It presents a novel approach for learning to predict sets with unknown permutation and cardinality using deep neural networks. The validity of the proposed approach is demonstrated on various vision problems.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Qadeer Khan, Idil Sueloe, Melis Oecal, Daniel Cremers
Summary: Supervised deep learning methods using image data have shown promise in vehicle control, but suffer from the need for labeled training data and poor performance on out-of-distribution scenarios. To address these issues, we propose a framework that leverages visual odometry to determine vehicle trajectory and uses this to infer steering labels. Additionally, synthesized images from deviated trajectories are included in the training distribution for improved neural network robustness.
APPLIED INTELLIGENCE
(2023)
Article
Robotics
Simon Klenk, Lukas Koestler, Davide Scaramuzza, Daniel Cremers
Summary: Estimating neural radiance fields (NeRFs) from ideal images has been extensively studied. However, most methods assume optimal illumination and camera motion, which are often violated in robotic applications. To address this, we propose E-NeRF, the first method that estimates NeRFs from a fast-moving event camera.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Lu Sang, Bjoern Haefner, Xingxing Zuo, Daniel Cremers
Summary: This paper presents a novel multi-view RGB-D based reconstruction method that utilizes a gradient signed distance field (gradient-SDF) to handle camera pose, lighting, albedo, and surface normal estimation. The proposed method optimizes the surface's quantities using its volumetric representation and validates two physically-based image formation models. Experimental results show that this method can recover high-quality surface geometry more accurately.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Florian Hofherr, Lukas Koestler, Florian Bernard, Daniel Cremers
Summary: We propose a method that combines neural implicit representations with neural ordinary differential equations to directly identify dynamic scene representations from visual observations. Our model requires less training data and has stronger generalization abilities than existing methods, and it can process high-resolution videos and synthesize photorealistic images. Additionally, our model can identify interpretable physical parameters and make long-term predictions.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Deepan Das, Qadeer Khan, Daniel Cremers
Summary: Ventriloquist-Net is a novel model for generating talking head images using a speech segment and a single face image, emphasizing on emotive expressions. It can handle in-the-wild source images and demonstrates state-of-the-art performance on unseen input data.
2022 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, ICIP
(2022)
Proceedings Paper
Automation & Control Systems
Mariia Gladkova, Nikita Korobov, Nikolaus Demmel, Aljosa Osep, Laura Leal-Taixe, Daniel Cremers
Summary: This paper proposes DirectTracker, a framework that effectively combines direct image alignment and sliding-window photometric bundle adjustment for 3D multi-object tracking, showing competitive performance.
2022 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Zhenzhang Ye, Tarun Yenamandra, Florian Bernard, Daniel Cremers
Summary: This paper proposes a trainable framework that uses graph neural networks to learn a deformable 3D geometry model from inhomogeneous image collections for graph matching tasks. The method outperforms recent learning-based approaches in terms of accuracy and cycle-consistency error, while also obtaining the underlying 3D geometry of the objects in the images.
THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / THE TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Marvin Eisenberger, Aysim Toker, Laura Leal-Taixe, Florian Bernard, Daniel Cremers
Summary: The Sinkhorn operator has gained popularity in computer vision and related fields due to its easy integration into deep learning frameworks. This article proposes an algorithm for obtaining analytical gradients of a Sinkhorn layer through implicit differentiation, allowing for any type of loss function and joint differentiation of target capacities and cost matrices. Error bounds for approximate inputs are also constructed. The results demonstrate improved stability, accuracy, and computational efficiency compared to automatic differentiation, particularly in resource-constrained scenarios like GPU memory.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Hang Li, Qadeer Khan, Volker Tresp, Daniel Cremers
Summary: This paper presents a computational framework inspired by the human brain to find the optimal low cost path between two nodes in a graph. The framework is able to handle unseen graphs and adapt to changes in node configurations during inference.
BRAIN INFORMATICS (BI 2022)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Dominik Roessle, Daniel Cremers, Torsten Schoen
Summary: This paper introduces a novel dynamic multi-modal and multi-instance network architecture that can learn intrinsic data fusion. By using Perceiver and Hopfield pooling, the proposed architecture outperforms the late fusion baseline by more than 40% accuracy in multi-modal setups, particularly on noisy data.
ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2022, PT I
(2022)