Article
Radiology, Nuclear Medicine & Medical Imaging
Andrea Ponsiglione, Arnaldo Stanzione, Gaia Spadarella, Agah Baran, Luca Alessandro Cappellini, Kevin Groot Lipman, Peter Van Ooijen, Renato Cuocolo
Summary: The overall methodological rigor of radiomics studies in the ovarian field is not ideal, with a lack of prospective design and formal validation of results. This limits the reproducibility of results and potential translation to clinical setting.
EUROPEAN RADIOLOGY
(2023)
Article
Critical Care Medicine
Jonathan R. Weir-McCall, Elise Debruyn, Scott Harris, Nagmi R. Qureshi, Robert C. Rintoul, Fergus Gleeson, Fiona J. Gilbert
Summary: A lung cancer prediction convolutional neural network (LCP-CNN) demonstrates diagnostic performance comparable to PET with CT imaging in the diagnosis of solitary pulmonary nodules. The LCP-CNN outperforms dynamic contrast-enhanced (DCE) CT imaging in accuracy, while being equal to PET with CT imaging.
Review
Oncology
Laurens S. Ter Maat, Isabella A. J. van Duin, Sjoerd G. Elias, Paul J. van Diest, Josien P. W. Pluim, Joost J. C. Verhoeff, Pim A. de Jong, Tim Leiner, Mitko Veta, Karijn P. M. Suijkerbuijk
Summary: This systematic review summarizes the current evidence on imaging biomarkers that predict response and survival in patients treated with checkpoint inhibitors in all cancer types. The review found that several imaging biomarkers, including simple imaging factors and radiomic features or deep learning models, can be used in clinical decision making. However, further research is needed to identify more accurate biomarkers for predicting which patients will not benefit from checkpoint inhibition.
EUROPEAN JOURNAL OF CANCER
(2022)
Review
Neurosciences
Daichi Sone, Iman Beheshti
Summary: Epilepsy is a common neurological disorder characterized by seizures, and machine learning methods have the potential to provide reliable and optimal performance in clinical diagnoses and personalized medicine. Various machine learning methodologies are being examined and validated for precise and reliable clinical applications in epilepsy and neuroimaging. The review focuses on the clinical applications of ML models for brain imaging in epilepsy, addressing practical problems and suggesting future research directions.
FRONTIERS IN NEUROSCIENCE
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Hossein Arabi, Habib Zaidi
Summary: This study implemented deep learning-based metal artefact reduction to minimize metal artefacts in CT images, with DLI-MAR approach showing superior performance compared to DLP-MAR and NMAR. Metal artefacts in CT images can lead to quantitative bias and image artefacts in PET images, but DLI-MAR technique effectively reduced these adverse effects.
EUROPEAN RADIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Yuki Onozato, Takekazu Iwata, Yasufumi Uematsu, Daiki Shimizu, Takayoshi Yamamoto, Yukiko Matsui, Kazuyuki Ogawa, Junpei Kuyama, Yuichi Sakairi, Eiryo Kawakami, Toshihiko Iizasa, Ichiro Yoshino
Summary: This study developed and validated multiple machine learning models using radiomic features from preoperative [F-18]fluorodeoxyglucose positron emission tomography/computed tomography (PET/CT) images to predict the pathological invasiveness of lung cancer.
EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING
(2023)
Article
Multidisciplinary Sciences
Seungwon Choi, Seunghyuk Moon, Jongduk Baek
Summary: Several sinogram inpainting based metal artifact reduction methods have been proposed, but existing methods are not effective in the presence of data truncation. This paper introduces a new MAR method that involves using a newly synthesized sinogram to reduce metal artifact effectively. The proposed method shows significant improvement in reducing residual metal artifacts compared to previous methods.
Article
Radiology, Nuclear Medicine & Medical Imaging
Russell Frood, Matt Clark, Cathy Burton, Charalampos Tsoumpas, Alejandro F. Frangi, Fergus Gleeson, Chirag Patel, Andrew Scarsbrook
Summary: This study evaluated the ability of machine learning models derived from pre-treatment FDG PET/CT to predict outcomes in cHL patients. The results showed that a ridge regression model using a 1.5 x mean liver SUV segmentation had the highest performance, with a high AUC when evaluated on training, validation, and test datasets.
EUROPEAN RADIOLOGY
(2022)
Review
Radiology, Nuclear Medicine & Medical Imaging
Keyur Radiya, Henrik Lykke Joakimsen, Karl Oyvind Mikalsen, Eirik Kjus Aahlin, Rolv-Ole Lindsetmo, Kim Erlend Mortensen
Summary: Machine learning is increasingly being used in medical imaging, particularly in liver CT imaging. This study provides an overview of the field and answers questions about the application and performance of ML in liver CT imaging, as well as its clinical applications.
EUROPEAN RADIOLOGY
(2023)
Article
Engineering, Electrical & Electronic
Ulugbek S. Kamilov, Charles A. Bouman, Gregery T. Buzzard, Brendt Wohlberg
Summary: Plug-and-play (PnP) priors are widely used frameworks for solving computational imaging problems by integrating physical models and learned models. PnP leverages high-fidelity physical sensor models and powerful machine learning methods for prior modeling of data, resulting in state-of-the-art reconstruction algorithms. PnP algorithms alternate between minimizing a data fidelity term for data consistency and imposing a learned regularizer, such as an image denoiser.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Computer Science, Interdisciplinary Applications
Qianyu Wu, Xu Ji, Yunbo Gu, Jun Xiang, Guotao Quan, Baosheng Li, Jian Zhu, Gouenou Coatrieux, Jean-Louis Coatrieux, Yang Chen
Summary: This paper introduces a novel Unsharp Structure Guided Filtering (USGF) method for reconstructing high-quality CT images directly from low-dose projections without clean references. Experimental results demonstrate that USGF achieves superior performance in terms of noise suppression and edge preservation.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Optics
Xiaojie Zhao, Yihong Li, Yan Han, Ping Chen, Jiaotong Wei
Summary: This paper proposes a statistical iterative spectral computed tomography (CT) imaging method based on blind separation of polychromatic projections to improve the accuracy of narrow-energy-width image decomposition. Experimental results show that the novel algorithm obtains more accurate narrow-energy-width images compared to state-of-the-art methods.
Review
Health Care Sciences & Services
Igino Simonetti, Piero Trovato, Vincenza Granata, Carmine Picone, Roberta Fusco, Sergio Venanzio Setola, Mauro Mattace Raso, Corrado Caraco, Paolo A. Ascierto, Fabio Sandomenico, Antonella Petrillo
Summary: Interval metastasis is a specific type of lymph node metastasis in patients with melanoma. Imaging techniques such as ultrasound, CT, MRI, lymphoscintigraphy, and PET are used for evaluation. Understanding these methods and collaborating with clinicians are essential for treatment and survival.
JOURNAL OF PERSONALIZED MEDICINE
(2022)
Article
Environmental Sciences
Zitong Huang, Takeshi Kurotori, Ronny Pini, Sally M. Benson, Christopher Zahasky
Summary: This article presents a method for quantifying the heterogeneous multiscale permeability in geologic porous media using positron emission tomography (PET) imaging. By utilizing a trained convolutional neural network (CNN) based on experimental data, accurate and computationally efficient permeability inversion maps can be generated. The results demonstrate the effectiveness of this approach by comparing the network-predicted permeability maps with experimental data.
WATER RESOURCES RESEARCH
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Piaoe Zeng, Annan Zhang, Le Song, Jianfang Liu, Huishu Yuan, Weifang Zhang
Summary: Spinal GCTTS is a very rare disease characterized by osteolytic bone destruction near the facet joint with a soft tissue mass and hypointensity on T2-weighted imaging. GCTTS shows high uptake of F-18-FDG, and PET/CT is helpful in detecting recurrent lesions.
INSIGHTS INTO IMAGING
(2021)
Editorial Material
Engineering, Electrical & Electronic
Bihan Wen, Saiprasad Ravishankar, Zhizhen Zhao, Raja Giryes, Jong Chul Ye
Summary: In recent years, there has been a growing interest in next-generation imaging systems and their combination with machine learning. Model-based imaging schemes, which incorporate physics-based forward models, noise models, and image priors, have laid the foundation in the emerging field of computational sensing and imaging. However, recent advances in machine learning techniques, such as large-scale optimization and deep neural networks, are increasingly being applied to improve the effectiveness and efficiency of computational imaging systems, leading to the redefinition of state-of-the-art computational imaging algorithms.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Engineering, Electrical & Electronic
Zhizhen Zhao, Jong Chul Ye, Yoram Bresler
Summary: Physics-informed generative modeling is a rapidly growing field in computational imaging, with various methods and applications. This review focuses on generative modeling techniques, such as variational autoencoders (VAEs) and generative adversarial networks (GANs), and recent advancements in score-based generative models. Through different imaging applications, the review demonstrates how these generative modeling techniques effectively incorporate the physics of the imaging problem to solve inverse problems.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Engineering, Electrical & Electronic
Guanhua Wang, Jeffrey A. Fessler
Summary: This paper presents an efficient and accurate method for computing approximate gradients involving NUFFTs, which improves the accuracy and efficiency of non-Cartesian MRI sampling trajectory learning. The proposed approach enables sampling optimization for larger image sizes and leads to improved image quality. It also shows that model-based image reconstruction methods can perform nearly as well as CNN-based methods when suitably optimized.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Engineering, Electrical & Electronic
Anish Lahiri, Gabriel Maliakal, Marc L. Klasky, Jeffrey A. Fessler, Saiprasad Ravishankar
Summary: This work focuses on image reconstruction when both the number of available CT projections and the training data is extremely limited. A sequential reconstruction approach is adopted, using an adversarially trained shallow network for 'destreaking' followed by a data-consistency update in each stage. To address the challenge of limited data, image subvolumes are used for training and patch aggregation during testing. To handle the computational challenge of 3D reconstruction, a hybrid 3D-to-2D mapping network is used for the 'destreaking' part. Comparisons with other methods indicate the potential of the proposed method in scenarios with highly limited projections and training data.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Zongyu Li, Yuni K. Dewaraja, Jeffrey A. Fessler
Summary: This article introduces an end-to-end unrolled iterative neural network training method for single-photon emission computerized tomography (SPECT) image reconstruction, which requires a memory-efficient forward-backward projector. The authors present an open-source Julia implementation of this projector, which uses only a fraction of the memory compared to a MATLAB-based projector. The study compares the performance of the proposed training method with other approaches using various phantoms and demonstrates that the end-to-end training with the Julia projector achieves the best reconstruction quality.
IEEE TRANSACTIONS ON RADIATION AND PLASMA MEDICAL SCIENCES
(2023)
Editorial Material
Engineering, Electrical & Electronic
Bihan Wen, Saiprasad Ravishankar, Zhizhen Zhao, Raja Giryes, Jong Chul Ye
Summary: The March issue of the IEEE Signal Processing Magazine focuses on the second volume of the special issue on physics-driven machine learning for computational imaging. This issue includes nine articles out of the 47 submissions that were accepted.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Computer Science, Interdisciplinary Applications
Hyungjin Chung, Eun Sun Lee, Jong Chul Ye
Summary: We propose a new denoising method based on score-based reverse diffusion sampling, which outperforms traditional MMSE denoisers in terms of both image quality and adaptability to real-world situations.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Il Yong Chun, Zhengyu Huang, Hongki Lim, Jeffrey A. Fessler
Summary: Iterative neural networks (INN) are gaining attention for solving inverse problems in imaging, combining regression NNs and an iterative model-based image reconstruction (MBIR) algorithm. This paper proposes the first fast and convergent INN architecture, Momentum-Net, by generalizing a block-wise MBIR algorithm using momentum and majorizers with regression NNs. Momentum-Net guarantees convergence to a fixed-point for general differentiable (non)convex MBIR functions and convex feasible sets, under two asymptomatic conditions.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Guanhua Wang, Jon-Fredrik Nielsen, Jeffrey A. Fessler, Douglas C. Noll
Summary: This study proposes a generalized framework for optimizing 3D non-Cartesian sampling patterns via data-driven optimization. The method can optimize multiple properties of sampling patterns simultaneously and can be used with either model-based or learning-based reconstruction methods.
MAGNETIC RESONANCE IN MEDICINE
(2023)
Article
Computer Science, Interdisciplinary Applications
Sangjoon Park, Jong Chul Ye
Summary: The widespread application of artificial intelligence in health research is currently limited by data availability. To overcome this limitation, distributed learning methods such as federated learning (FL) and split learning (SL) are introduced, each with their own strengths and weaknesses in addressing data management and ownership issues. However, a recent proposal called federated split task-agnostic (F eSTA) learning suffers from high communication overhead. To address this, a new method called ${p}$ -F eSTA is presented, which uses ViT with random patch permutation and achieves improved multi-task learning performance without sacrificing privacy.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Interdisciplinary Applications
Hanvit Kim, Zongyu Li, Jiye Son, Jeffrey A. Fessler, Yuni K. Dewaraja, Se Young Chun
Summary: This article proposes a physics-guided weakly supervised training framework for fast and accurate scatter estimation in quantitative SPECT. The method utilizes a 100 x shorter MC simulation as weak labels and enhances them with deep neural networks, achieving comparable performance to the supervised counterpart while significantly reducing the computation in labeling.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Chanseok Lee, Gookho Song, Hyeonggeon Kim, Jong Chul Ye, Mooseok Jang
Summary: Reconstructing holographic images is difficult due to the ill posed inverse mapping problem. However, Lee and colleagues propose a deep learning method that incorporates a physical model and can handle physical perturbations in holographic image reconstruction, making it more reliable and applicable to a wider range of imaging problems.
NATURE MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Sangjoon Park, Eun Sun Lee, Kyung Sook Shin, Jeong Eun Lee, Jong Chul Ye
Summary: The increasing demand for AI systems to monitor human errors and abnormalities in healthcare presents challenges. This study presents a model called Medical X-VL, which is tailored for the medical domain and outperformed current state-of-the-art models in two medical image datasets. The model enables various zero-shot tasks for monitoring AI in the medical domain.
MEDICAL IMAGE ANALYSIS
(2024)
Article
Engineering, Electrical & Electronic
Hongki Lim, Yuni K. Dewaraja, Jeffrey A. Fessler
Summary: This study introduces a trained regularizer for SPECT reconstruction using CT imaging, which improves the imaging effect of low-count SPECT by incorporating CT-side information. Experimental results show that this method outperforms conventional methods in activity quantification, noise reduction, and root mean square error.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Gihyun Kwon, Jong Chul Ye
Summary: In this paper, a novel single-shot GAN adaptation method is proposed through unified CLIP space manipulations to address the overfitting or under-fitting issues when fine-tuning with a single target image. Experimental results demonstrate that the proposed method outperforms baseline models in generating diverse outputs with the target texture and allows more effective attribute editing.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)