Current Imaging Research Articles

On-line correlative imaging of cryo-PALM and soft X-ray tomography for identification of subcellular structures

Correlative imaging of fluorescence microscopy and soft X-ray microscopy plays a crucial role in exploring the relationship between structure and function in cellular biology. However, the current correlative imaging methods are limited either to off-line or low-resolution fluorescence imaging. In this study, we developed an integrated on-line cryogenic photoactivated localization microscopy (cryo-PALM) system at a soft X-ray microscopy station. This design eliminates some critical issues such as sample damage and complex post-correlation arising from transferring samples between different cryostages. Furthermore, we successfully achieved correlative imaging of cryopreserved near-native cells, with a resolution of about 50 nm of cryo-PALM. Therefore, the developed on-line correlation imaging platform provides a powerful tool for investigating the intricate relationship between structure and function in biological and molecular interactions, as well as in other life science disciplines.

Deep underwater image compression for enhanced machine vision applications

Underwater image compression is fundamental in underwater visual applications. The storage resources of autonomous underwater vehicles (AUVs) and underwater cameras are limited. By employing effective image compression methods, it is possible to optimize the resource utilization of these devices, thereby extending the operational time underwater. Current image compression methods neglect the unique characteristics of the underwater environment, thus failing to support downstream underwater visual tasks efficiently. We propose a novel underwater image compression framework that integrates frequency priors and feature decomposition fusion in response to these challenges. Our framework incorporates a task-driven feature decomposition fusion module (FDFM). This module enables the network to understand and preserve machine-friendly information during the compression process, prioritizing task relevance over human visual perception. Additionally, we propose a frequency-guided underwater image correction module (UICM) to address noise issues and accurately identify redundant information, enhancing the overall compression process. Our framework effectively preserves machine-friendly features at a low bit rate. Extensive experiments across various downstream visual tasks, including object detection, semantic segmentation, and saliency detection, consistently demonstrated significant improvements achieved by our approach.

Theory and mitigation of motional eddy current in high-field eddy current shielding.

Eddy current shielding by a Faraday cage is an effective way to shield alternating-current magnetic fields in scientific instrumentation. In a strong static magnetic field, however, the eddy current in the conductive shield is subject to the Lorentz force, which causes the shield to vibrate. In addition to mechanical issues (e.g., acoustic noise), such vibration induces motional eddy current in the shield that can dominate the original, electromagnetic eddy current to undermine the conductor's shielding capability. In this work, we investigate a method to control motional eddy current by making cut-out patterns in the conductor that follow the electromagnetic eddy current image. This effectively limits the surface current of the plate to a single mode and prevents the proliferation of uncontrolled motion-induced surface currents that disrupts eddy current shielding. After developing a comprehensive theory of magneto-mechanical interaction in a conductive plate, the proposed method was tested on a flat-geometry testbed experiment inside a 3 T magnetic resonance imaging (MRI) magnet. It was found that the magnetic field generated by the motional eddy current was much more localized in space and frequency for a patterned-copper shield compared to a solid copper. The magnetic field of the patterned shield could be accurately predicted from the impedance measurement in the magnet. Implications of our results for improved shielding of gradient fields in high-field MRI are discussed.

Online welding status monitoring method of T-joint double-sided double arc welding based on multi-source information fusion

In modern manufacturing, double-sided double arc (DSDA) welding brings advantages in properties and efficiency for T-shaped joints. However, the double-side forming makes it difficult to detect the imperfections and ensure weld joint quality, which few studies have covered. Most existing methods deal with single-arc welding with limited sensing sources. To fill the gap, an online welding status monitoring method for DSDA welding is proposed based on the fusion of welding current, arc voltage, arc sound, and weld pool images. In pre-processing, the automatic weld pool region of interest (ROI) detection method based on the lightweight YOLO-L model and the waveform denoising algorithm are designed. In feature engineering, waveform signals are analyzed in both the time and frequency domain. A weld pool feature extractor based on a convolutional neural network (CNN) is proposed with good effectiveness and interpretability. The output feature combination is refined by Fisher-based selection and evaluation. In model building, an ensemble learning model based on three high-fit basic learners is proposed, with an accuracy of 98.538 %. The proposed model shows significant advantages over the single basic classifier and other ensemble methods. Experiments verify high precision and robustness, laying a foundation for accurate real-time monitoring of DSDA welding production.

Automatic 3D reconstruction of vertebrae from orthogonal bi-planar radiographs

When conducting spine-related diagnosis and surgery, the three-dimensional (3D) upright posture of the spine under natural weight bearing is of significant clinical value for physicians to analyze the force on the spine. However, existing medical imaging technologies cannot meet current requirements of medical service. On the one hand, the mainstream 3D volumetric imaging modalities (e.g. CT and MRI) require patients to lie down during the imaging process. On the other hand, the imaging modalities conducted in an upright posture (e.g. radiograph) can only realize 2D projections, which lose the valid information of spinal anatomy and curvature. Developments of deep learning-based 3D reconstruction methods bring potential to overcome the limitations of the existing medical imaging technologies. To deal with the limitations of current medical imaging technologies as is described above, in this paper, we propose a novel deep learning framework, ReVerteR, which can realize automatic 3D Reconstruction of Vertebrae from orthogonal bi-planar Radiographs. With the utilization of self-attention mechanism and specially designed loss function combining Dice, Hausdorff, Focal, and MSE, ReVerteR can alleviate the sample-imbalance problem during the reconstruction process and realize the fusion of the centroid annotation and the focused vertebra. Furthermore, aiming at automatic and customized 3D spinal reconstruction in real-world scenarios, we extend ReVerteR to a clinical deployment-oriented framework, and develop an interactive interface with all functions in the framework integrated so as to enhance human–computer interaction during clinical decision-making. Extensive experiments and visualization conducted on our constructed datasets based on two benchmark datasets of spinal CT, VerSe 2019 and VerSe 2020, demonstrate the effectiveness of our proposed ReVerteR. In this paper, we propose an automatic 3D reconstruction method of vertebrae based on orthogonal bi-planar radiographs. With the 3D upright posture of the spine under natural weight bearing effectively constructed, our proposed method is expected to better support doctors make clinical decision during spine-related diagnosis and surgery.

Cnn-trans model: A parallel dual-branch network for fundus image classification

The existence of fundus diseases not only endangers people’s vision, but also brings serious economic burden to the society. Fundus images are an objective and standard basis for the diagnosis of fundus diseases. With the continuous advancement of computer science, deep learning methods dominated by convolutional neural networks (CNN) have been widely used in fundus image classification. However, the current CNN-based fundus image classification research still has a lot of room for improvement: CNN cannot effectively avoid the interference of repeated background information and has limited ability to model the whole world. In response to the above findings, this paper proposes the CNN-Trans model. The CNN-Trans model is a parallel dual-branch network, which is the two branches of CNN-LSTM and Vision Transform (ViT). The CNN-LSTM branch uses Xception after transfer learning. As the original feature extractor, LSTM is responsible for dealing with the gradient disappearance problem in neural network iterations before the classification head, and then introduces a new type of lightweight attention mechanism between Xception and LSTM: Coordinate Attention, so as to emphasize the key information related to classification and suppress the less useful repeated background information; while the self-attention mechanism in the ViT branch is not limited by local interactions, it can establish long-distance dependence on the target and extract global features. Finally, the concatenation (Concat) operation is used to fuse the features of the two branches. The local features extracted by the CNN-LSTM branch and the global features extracted by the ViT branch form complementary advantages. After feature fusion, more comprehensive image feature information is sent to the to the classification layer. Finally, after a large number of experimental tests and comparisons, the results show that: the CNN-Trans model achieved an accuracy of 80.68% on the fundus image classification task, and the CNN-Trans model has a classification that is comparable to the state-of-the-art methods. performance..

Deep demosaicking convolution neural network and quantum wavelet transform-based image denoising

ABSTRACT Demosaicking is a popular scientific area that is being explored by a vast number of scientists. Current digital imaging technologies capture colour images with a single monochrome sensor. In addition, the colour images were captured using a sensor coupled with a Colour Filter Array (CFA). Furthermore, the demosaicking procedure is required to obtain a full-colour image. Image denoising and image demosaicking are the two important image restoration techniques, which have increased popularity in recent years. Finding a suitable strategy for multiple image restoration is critical for researchers. Hence, a deep learning (DL) based image denoising and image demosaicking is developed in this research. Moreover, the Autoregressive Circle Wave Optimization (ACWO) based Demosaicking Convolutional Neural Network (DMCNN) is designed for image demosaicking. The Quantum Wavelet Transform (QWT) is used in the image denoising process. Similarly, Quantum Wavelet Transform (QWT) is used to analyse the abrupt changes in the input image with noise. The transformed image is then subjected to a thresholding technique, which determines an appropriate threshold range. Once the threshold range has been determined, soft thresholding is applied to the resulting wavelet coefficients. After that, the extraction and reconstruction of the original image is carried out using the Inverse Quantum Wavelet Transform (IQWT). Finally, the fused image is created by combining the results of both processes using a weighted average. The denoised and demosaicked images are combined using the weighted average technique. Furthermore, the proposed QWT+DMCNN-ACWO model provided the ideal values of Peak signal-to-noise ratio (PSNR), Second derivative like measure of enhancement (SDME), Structural Similarity Index (SSIM), Figure of Merit (FOM) of 0.890, and computational time of 49.549 dB, 59.53 dB, 0.963, 0.890, and 0.571, respectively.

SIMFusion: A semantic information-guided modality-specific fusion network for MR Images

Multi-modal medical image fusion aims to integrate distinct imaging modalities to yield more comprehensive and precise medical images, which can benefit the subsequent image analysis tasks. However, prevailing state-of-the-art image fusion methods, despite substantial advancements, do not explicitly address the efficient handling of complementary information between modalities. Moreover, most current multi-modal medical image fusion methods encounter challenges when integrating with practical tasks, which lack the guidance of semantic information, thus hindering the generation of high-quality images for accurate identification of lesion areas. To deal with these challenges, this paper introduces a novel semantic information-guided modality-specific fusion network for multi-modal magnetic resonance (MR) images, named SIMFusion. Specifically, we propose the decomposition branch to capture common and specific features from MR images of different modalities for fusion to reduce information redundancy through correlated mutual information loss. Subsequently, we obtain their semantic characteristics via the semantic branch utilizing a pre-train segmentation network and, finally, achieve an adaptive balance between the two sets of features through a specialized fusion strategy. Extensive experiments demonstrate the superiority of SIMFusion over existing competing techniques on both BraTS2019 and ISLES2022 datasets, with 8.1% improvement in MI and 38.7% in VIFF in T2-T1ce image pairs, highlighting the potential of the proposed method as a promising solution for MR image fusion in practical applications. Our code will be released at https://github.com/Zhangxw-ustc/SIMFusion.

Automated Laryngeal Invasion Detector of Boluses in Videofluoroscopic Swallowing Study Videos Using Action Recognition-Based Networks.

We aimed to develop an automated detector that determines laryngeal invasion during swallowing. Laryngeal invasion, which causes significant clinical problems, is defined as two or more points on the penetration-aspiration scale (PAS). We applied two three-dimensional (3D) stream networks for action recognition in videofluoroscopic swallowing study (VFSS) videos. To detect laryngeal invasion (PAS 2 or higher scores) in VFSS videos, we employed two 3D stream networks for action recognition. To establish the robustness of our model, we compared its performance with those of various current image classification-based architectures. The proposed model achieved an accuracy of 92.10%. Precision, recall, and F1 scores for detecting laryngeal invasion (≥PAS 2) in VFSS videos were 0.9470 each. The accuracy of our model in identifying laryngeal invasion surpassed that of other updated image classification models (60.58% for ResNet101, 60.19% for Swin-Transformer, 63.33% for EfficientNet-B2, and 31.17% for HRNet-W32). Our model is the first automated detector of laryngeal invasion in VFSS videos based on video action recognition networks. Considering its high and balanced performance, it may serve as an effective screening tool before clinicians review VFSS videos, ultimately reducing the burden on clinicians.

Evaluating soluble Axl as a biomarker for glioblastoma: A pilot study.

With current imaging, discriminating tumor progression from treatment effect following immunotherapy or oncolytic virotherapy of glioblastoma (GBM) is challenging. A blood based diagnostic biomarker would therefore be helpful. Axl is a receptor tyrosine kinase that is highly expressed by many cancers including GBM. Axl expression is regulated through enzymatic cleavage of its extracellular domain. The resulting fragment can be detected in serum as soluble Axl (sAxl). sAxl levels can distinguish patients with melanoma, hepatocellular carcinoma, and pancreatic ductal adenocarcinoma from healthy controls. This is a pilot study to determine if sAxl is a candidate biomarker for GBM. The sAxl levels in the serum of 40 healthy volunteers and 20 GBM patients were determined using an enzyme-linked immunosorbent assay (ELISA). Pre- and post- operative sAxl levels were obtained. Volumetric MRI evaluation provided GBM tumor volume metrics. There was no significant difference in the sAxl levels of the volunteers (30.16±1.88 ng/ml) and GBM patients (30.74±1.96 ng/ml) p = 0.27. The postoperative sAxl levels were significantly higher than preoperative levels (32.32±2.26 ng/ml vs 30.74±1.96 ng/ml, p = 0.03). We found no correlation between tumor volume and sAxl levels. Axl expression was low or absent in 6 of 11 (55%) patient derived GBM cell lines. Given the wide range of Axl expression by GBM tumors, sAxl may not be a reliable indicator of GBM. However, given the small sample size in this study, a larger study may be considered.

Integrating Tendinous Pathophysiology Into Rotator Cuff Tears And Greater Trochanteric Pain Syndrome: A Narrative Review

This narrative review aims to use the similarities between the shoulder and hip joints to better understand why rotator cuff (RC) tendinopathy and hip abductor tendinopathy occur and inform about diagnosis and treatment of both orthopedic complaints. A search of the literature was conducted using GoogleScholar and Pubmed and initially followed a systematic review protocol, but the nature of the topic, current literature and data necessitated a narrative review. Reports that discussed pathomechanics of RC and gluteal tendinopathy individually, together and with other muscles groups were reviewed. It was found that the methods measuring and describing the processes of tendinopathy differ significantly, for each individual joint and between all joints. A review of a large body of quantifiable measures and theoretical ideas regarding tendinopathy was performed to address this lack of consensus in current literature. Initial literature yielded 74 articles. After review, only 43 articles were used from a broad range of approaches and methodologies. The review found a body of evidence suggesting that fibrocartilage overgrowth and compressive forces over bony structures cause tendonous pathology of the RC and hip abductor tendons. These findings support the idea that tendinopathy is often caused by intrinsic factors rather than the traditional view of external factors. Earlier treatment and improved outcomes without surgery are possible using current imaging technology to identify these intrinsic factors that affect tendinous properties.

ER-Targeted PET for Initial Staging and Suspected Recurrence in ER-Positive Breast Cancer

There are insufficient data comparing 16α-18F-fluoro-17β-estradiol (FES) positron emission tomography (PET) computed tomography (CT) with standard-of-care imaging (SOC) for staging locally advanced breast cancer (LABC) or evaluating suspected recurrence. To determine the detection rate of FES PET/CT and SOC for distant metastases in patients with estrogen receptor (ER)-positive LABC and recurrences in patients with ER-positive BC and suspected recurrence. This diagnostic study was conducted as a single-center phase 2 trial, from January 2021 to September 2023. The study design provided 80% power to find a 20% detection rate difference. Participants included patients with ER-positive LABC (cohort 1) or suspected recurrence (cohort 2). Data were analyzed from September 2023 to February 2024. Participants underwent both SOC imaging and experimental FES PET/CT. When there were suspicious lesions on imaging, 1 was biopsied for histopathological reference standard to confirm presence (true positive) or absence (false positive) of malignant neoplasm. The outcome of interest was the detection rate of FES PET CT vs SOC for distant metastases and recurrences. A total of 124 patients were accrued, with 62 in cohort 1 (median [IQR] age, 52 [32-84] years) and 62 in cohort 2 (median [IQR] age, 66 [30-93] years). In cohort 1, of 14 true-positive findings, SOC imaging detected 12 and FES detected 11 (P > .99). In cohort 2, of 23 true-positive findings, SOC detected 16 and FES detected 18 (P = .77). In 30 patients with lobular histology, of 11 true-positive findings, SOC detected 5 and FES detected 9 (P = .29). There were 6 false-positive findings on SOC and 1 false-positive finding on FES PET/CT (P = .13). In this diagnostic study with pathological findings as the reference standard, no difference was found between FES PET/CT and current SOC imaging for detecting distant metastases in patients with ER-positive LABC or recurrences in patients with ER-positive tumors and suspected recurrence. FES PET/CT could be considered for both clinical indications, which are not part of current Appropriate Use Criteria for FES PET. The findings regarding FES PET/CT in patients with lobular tumors, and for lower false positives than current SOC imaging, warrant further investigation.

Prevalence of financial hardship and health-related social needs among patients with missed radiology appointments

PurposeWe aimed to evaluate the prevalence of financial hardship and Health-Related Social Needs (HRSN) among patients who missed their radiology appointment. MethodsEnglish-speaking adult patients, with a missed outpatient imaging appointment at any of a tertiary care imaging centers between 11/2022 and 05/2023 were eligible. We measured self-reported general financial worry using Comprehensive Score for Financial Toxicity (COST), imaging hardship (worry that the current imaging is a financial hardship to patient and their family), material hardship (e.g., medical debt), cost-related care nonadherence, and HRSNs including housing instability, food insecurity, transportation problems, and utility help needs. Results282 patients were included (mean age 54.7 ± 15.0 years; 70.7 % female). Majority were non-Hispanic White (52.4 %), followed by Asian (23.0 %) and Hispanic (16.0 %) racial/ethnic background. Most missed appointments were patient-initiated (74.8 %); 13.5 % due to cost or insurance coverage and 6.4 % due to transportation and parking. Mean COST score was 26.8 with 44.4 % and 28.8 % reporting their illness and imaging as a source of financial hardship. 18.3 % and 35.2 % endorsed cost-related care nonadherence and material hardship. 32.7 % had at least one HRSNs with food insecurity the most common (25.4 %). Only 12.5 % were previously screened for financial hardship or HRSNs. Having comorbidity and living in more disadvantaged neighborhoods was associated with higher report of financial hardship and HRSNs. ConclusionFinancial hardship and HRSNs are common among those who miss radiology appointments. There needs to be more rigorous screening for financial hardship and HRSNs at every health encounter and interventions should be implemented to address these.

Cerebrospinal Fluid Leaks: Challenges in Localizing Spontaneous Spinal Leak Sites and Minimally Invasive Treatment

The confirmation of cerebrospinal fluid (CSF) leaks in the setting of spontaneous intracranial hypotension (SIH) by imaging involves a growing toolset of multimodal advanced spinal and skull base imaging techniques, for which exists a unique set of challenges for each CSF leak type. Furthermore, the repertoire of minimally invasive CSF leak treatment beyond nontargeted epidural blood patch administration has grown widely, with varied practices across institutions. This review describes current diagnostic imaging and treatment modalities as they apply to the challenges of CSF leak localization and management.

Current density impedance imaging with PINNs

In this paper, we introduce CDII-PINNs, a computationally efficient method for solving CDII using PINNs in the framework of Tikhonov regularization. This method constructs a physics-informed loss function by merging the regularized least-squares output functional with an underlying differential equation, which describes the relationship between the conductivity and voltage. A pair of neural networks representing the conductivity and voltage, respectively, are coupled by this loss function. Then, minimizing the loss function provides a reconstruction. A rigorous theoretical guarantee is provided. We give an error analysis for CDII-PINNs and establish a convergence rate, based on prior selected neural network parameters in terms of the number of samples. The numerical simulations demonstrate that CDII-PINNs are efficient, accurate and robust to noise levels ranging from 1% to 20%.

Multidimensional Analysis of the Adult Human Heart in Health and Disease Using Hierarchical Phase-Contrast Tomography.

Background Current clinical imaging modalities such as CT and MRI provide resolution adequate to diagnose cardiovascular diseases but cannot depict detailed structural features in the heart across length scales. Hierarchical phase-contrast tomography (HiP-CT) uses fourth-generation synchrotron sources with improved x-ray brilliance and high energies to provide micron-resolution imaging of intact adult organs with unprecedented detail. Purpose To evaluate the capability of HiP-CT to depict the macro- to microanatomy of structurally normal and abnormal adult human hearts ex vivo. Materials and Methods Between February 2021 and September 2023, two adult human donor hearts were obtained, fixed in formalin, and prepared using a mixture of crushed agar in a 70% ethanol solution. One heart was from a 63-year-old White male without known cardiac disease, and the other was from an 87-year-old White female with a history of multiple known cardiovascular pathologies including ischemic heart disease, hypertension, and atrial fibrillation. Nondestructive ex vivo imaging of these hearts without exogenous contrast agent was performed using HiP-CT at the European Synchrotron Radiation Facility. Results HiP-CT demonstrated the capacity for high-spatial-resolution, multiscale cardiac imaging ex vivo, revealing histologic-level detail of the myocardium, valves, coronary arteries, and cardiac conduction system across length scales. Virtual sectioning of the cardiac conduction system provided information on fatty infiltration, vascular supply, and pathways between the cardiac nodes and adjacent structures. HiP-CT achieved resolutions ranging from gross (isotropic voxels of approximately 20 µm) to microscopic (approximately 6.4-µm voxel size) to cellular (approximately 2.3-µm voxel size) in scale. The potential for quantitative assessment of features in health and disease was demonstrated. Conclusion HiP-CT provided high-spatial-resolution, three-dimensional images of structurally normal and diseased ex vivo adult human hearts. Whole-heart image volumes were obtained with isotropic voxels of approximately 20 µm, and local regions of interest were obtained with resolution down to 2.3-6.4 µm without the need for sectioning, destructive techniques, or exogenous contrast agents. Published under a CC BY 4.0 license Supplemental material is available for this article. See also the editorial by Bluemke and Pourmorteza in this issue.

Image Processing Technology Based on Machine Learning

Machine learning is a relatively new field. With the deepening of people's research in this field, the application of machine learning is increasingly extensive. On the other hand, with the development of science and technology, image has become an indispensable medium of information transmission, and image processing technology is also booming. This paper introduces machine learning into image processing, and studies the image processing technology based on machine learning. This paper summarizes the current popular image processing technology, compares various image technology in detail, and explains the limitations of each image processing method. In addition, on the basis of image processing, this paper introduces machine learning algorithm, applies convolution neural network to feature extraction of image processing, and carries out simulation test. In the test, we select voc2007 dataset for image segmentation, Imagenet dataset for target detection, cifar100 dataset for image classification, and ROC curve for performance evaluation. The results show that the algorithm based on deep learning can achieve high accuracy in image segmentation, classification and target detection. The accuracy of image segmentation is 0.984, the accuracy of image classification is 0.987, and the accuracy of target detection is 0.986. Image processing based on machine learning has great advantages.

Exploratory Tau PET/CT with [11C]PBB3 in Patients with Suspected Alzheimer's Disease and Frontotemporal Lobar Degeneration: A Pilot Study on Correlation with PET Imaging and Cerebrospinal Fluid Biomarkers.

Accurately diagnosing Alzheimer's disease (AD) and frontotemporal lobar degeneration (FTLD) is challenging due to overlapping symptoms and limitations of current imaging methods. This study investigates the use of [11C]PBB3 PET/CT imaging to visualize tau pathology and improve diagnostic accuracy. Given diagnostic challenges with symptoms and conventional imaging, [11C]PBB3 PET/CT's potential to enhance accuracy was investigated by correlating tau pathology with cerebrospinal fluid (CSF) biomarkers, positron emission tomography (PET), computed tomography (CT), amyloid-beta, and Mini-Mental State Examination (MMSE). We conducted [11C]PBB3 PET/CT imaging on 24 patients with suspected AD or FTLD, alongside [11C]PiB PET/CT (13 patients) and [18F]FDG PET/CT (15 patients). Visual and quantitative assessments of [11C]PBB3 uptake using standardized uptake value ratios (SUV-Rs) and correlation analyses with clinical assessments were performed. The scans revealed distinct tau accumulation patterns; 13 patients had no or faint uptake (PBB3-negative) and 11 had moderate to pronounced uptake (PBB3-positive). Significant inverse correlations were found between [11C]PBB3 SUV-Rs and MMSE scores, but not with CSF-tau or CSF-amyloid-beta levels. Here, we show that [11C]PBB3 PET/CT imaging can reveal distinct tau accumulation patterns and correlate these with cognitive impairment in neurodegenerative diseases. Our study demonstrates the potential of [11C]PBB3-PET imaging for visualizing tau pathology and assessing disease severity, offering a promising tool for enhancing diagnostic accuracy in AD and FTLD. Further research is essential to validate these findings and refine the use of tau-specific PET imaging in clinical practice, ultimately improving patient care and treatment outcomes.

Selection of the Binding Object on the Current Image Formed by the Technical Vision System Using Structural and Geometric Features

Selection of the Binding Object on the Current Image Formed by the Technical Vision System Using Structural and Geometric Features

A Collaborative Self-supervised Domain Adaptation for Low-Quality Medical Image Enhancement.

Medical image analysis techniques have been employed in diagnosing and screening clinical diseases. However, both poor medical image quality and illumination style inconsistency increase uncertainty in clinical decision-making, potentially resulting in clinician misdiagnosis. The majority of current image enhancement methods primarily concentrate on enhancing medical image quality by leveraging high-quality reference images, which are challenging to collect in clinical applications. In this study, we address image quality enhancement within a fully self-supervised learning setting, wherein neither high-quality images nor paired images are required. To achieve this goal, we investigate the potential of self-supervised learning combined with domain adaptation to enhance the quality of medical images without the guidance of high-quality medical images. We design a Domain Adaptation Self-supervised Quality Enhancement framework, called DASQE. More specifically, we establish multiple domains at the patch level through a designed rule-based quality assessment scheme and style clustering. To achieve image quality enhancement and maintain style consistency, we formulate the image quality enhancement as a collaborative self-supervised domain adaptation task for disentangling the low-quality factors, medical image content, and illumination style characteristics by exploring intrinsic supervision in the low-quality medical images. Finally, we perform extensive experiments on six benchmark datasets of medical images, and the experimental results demonstrate that DASQE attains state-of-the-art performance. Furthermore, we explore the impact of the proposed method on various clinical tasks, such as retinal fundus vessel/lesion segmentation, nerve fiber segmentation, polyp segmentation, skin lesion segmentation, and disease classification. The results demonstrate that DASQE is advantageous for diverse downstream image analysis tasks.