Heterogeneous Images Research Articles

Federated Learning on Heterogeneous Images in Internet of Biomedical Things

Federated Learning on Heterogeneous Images in Internet of Biomedical Things

UTERINE SAROMAS Retrospective Studies on 7 Cases at the Hassan II University Hospital in Fez

Introduction: Uterine sarcomas are rare tumors, representing less than 3% of malignant tumors of the female genital tract and between 3 and 7% of malignant tumors of the uterine body. They are a heterogeneous group of tumors comprising different histological subtypes, but with a generally identical evolutionary profile. Unlike uterine epithelial tumors, and apart from low-grade endometrial stromal sarcomas, uterine sarcomas have a poor prognosis. They have a high rate of recurrence (between 50 and 70%), local and especially metastatic (in approximately 70% of cases). The objective of our work is to specify the epidemiological characteristics of uterine sarcomas, the diagnostic difficulties faced by practitioners at the clinical, radiological and anatomopathological stages, the modalities of management as well as the different determinants of the prognosis of these rare tumors. Patients And Methods: Seven cases of uterine sarcomas treated in the gynecology - obstetrics II department at the Hassan II University Hospital in Fez between 2019 and 2023 were analyzed retrospectively. Results: The average age of our patients varies between 50 and 75 years with an average of 62.2 years. Of the 7 patients, 90% are multiparous, 100% menopausal. 78% of our patients consulted for metrorrhagia. On the paraclinical level, we performed a pelvic ultrasound for all patients, 80% of the results of which were in favor of an enlarged uterus with a heterogeneous echogenic intracavitary image, hysteroscopy in 04 patients for exploration of postmenopausal metrorrhagia and magnetic resonance imaging in 06 patients. The diagnosis was made at the preoperative stage in five patients. 71% of our patients benefited from surgical treatment. The anatomopathological analysis of the surgical specimens showed that 05, 07 patients benefited from adjuvant radiotherapy, while a combined chemotherapy/radiotherapy postoperatively was indicated in 02 patients. One patient presented a recurrence having ...

The impact of artificial intelligence in the diagnosis and management of acoustic neuroma: A systematic review.

Schwann cell sheaths are the source of benign, slowly expanding tumours known as acoustic neuromas (AN). The diagnostic and treatment approaches for AN must be patient-centered, taking into account unique factors and preferences. The purpose of this study is to investigate how machine learning and artificial intelligence (AI) can revolutionise AN management and diagnostic procedures. A thorough systematic review that included peer-reviewed material from public databases was carried out. Publications on AN, AI, and deep learning up until December 2023 were included in the review's purview. Based on our analysis, AI models for volume estimation, segmentation, tumour type differentiation, and separation from healthy tissues have been developed successfully. Developments in computational biology imply that AI can be used effectively in a variety of fields, including quality of life evaluations, monitoring, robotic-assisted surgery, feature extraction, radiomics, image analysis, clinical decision support systems, and treatment planning. For better AN diagnosis and treatment, a variety of imaging modalities require the development of strong, flexible AI models that can handle heterogeneous imaging data. Subsequent investigations ought to concentrate on reproducing findings in order to standardise AI approaches, which could transform their use in medical environments.

Lensless On-Chip Chemiluminescence Imaging for High-Throughput Single-Cell Heterogeneity Analysis.

High-throughput single-cell heterogeneity imaging and analysis is essential for understanding complex biological systems and for advancing personalized precision disease diagnosis and treatment. Here, we present a miniaturized lensless chemiluminescence chip for high-throughput single-cell functional imaging with subcellular resolution. With the sensitive chemiluminescence sensing and wide field of view of contact lensless imaging, we demonstrated the chemiluminescent imaging of over 1000 single cells, and their membrane glycoprotein and the high-throughput single-cell heterogeneity of membrane protein imaging were examined for precision analysis. Furthermore, the functional adhesion and heterogeneity of single live cells were imaged and explored. This miniaturized lensless on-chip CL-CMOS imaging platform enables high-throughput single-cell imaging and analysis with high sensitivity and subcellular resolution, providing new techniques for the cellular study of biological heterogeneity and has potential application in precision disease diagnosis and treatment at the point-of-care settings.

FACSNet: Forensics aided content selection network for heterogeneous image steganalysis

The main goal of image steganalysis, as a technique of confrontation with steganography, is to determine the presence or absence of secret information in conjunction with the specific statistical characteristics of the carrier. With the development of deep learning technology in recent years, the performance of steganography has been gradually enhanced. Especially for the complex reality environment, the image content is mixed and heterogeneous, which brings great challenges to the practical application of image steganalysis technology. In order to solve this problem, we design a forensics aided content selection network (FACSNet) for heterogeneous image steganalysis. Considering the heterogeneous situation of real images, a forensics aided module is introduced to pre-categorise the images to be tested, so that the network is able to detect different categories of images in a more targeted way. The complexity of the images is also further analysed and classified using the content selection module to train a more adapted steganalyser. By doing this, the network is allowed to achieve better performance in recognising and classifying the heterogeneous images for detection. Experimental results show that our designed FACSNet is able to achieve excellent detection performance in heterogeneous environments, improving the detection accuracy by up to 7.14% points, with certain robustness and practicality.

A novel infrared and visible image fusion network based on cross-modality reinforcement and multi-attention fusion strategy

Infrared and visible image fusion focuses on the integration of complementary information, whose fused results with abundant texture details and salient targets can cater for human visual observation well. However, the loss of some important details is still a challenge in the image fusion process. In this paper, a novel method based on cross-modality reinforcement module and multi-attention fusion strategy is proposed to boost the end-to-end CNN backbone. Specifically, a cross-modality architecture is applied to compensate the spectral differences from heterogeneous images; and the multi-scale strip pooling is utilized as a further feature representation tool to model long-rang independencies precisely; then, a detail injection block is devised to fulfill the enhancement requirements of texture-contrast and target intensity; sequentially, a multi-attention fusion module is proposed to integrate features progressively. Extensive comparative experiments are conducted on several datasets to demonstrate the superiority of the proposed method in both quantitative metrics and visual perception, such as the average of visual information fidelity metric based on all the experiment samples reach to 0.964.

Iterative Training Sample Augmentation for Enhancing Land Cover Change Detection Performance With Deep Learning Neural Network.

Labeled samples are important in achieving land cover change detection (LCCD) tasks via deep learning techniques with remote sensing images. However, labeling samples for change detection with bitemporal remote sensing images is labor-intensive and time-consuming. Moreover, manually labeling samples between bitemporal images requires professional knowledge for practitioners. To address this problem in this article, an iterative training sample augmentation (ITSA) strategy to couple with a deep learning neural network for improving LCCD performance is proposed here. In the proposed ITSA, we start by measuring the similarity between an initial sample and its four-quarter-overlapped neighboring blocks. If the similarity satisfies a predefined constraint, then a neighboring block will be selected as the potential sample. Next, a neural network is trained with renewed samples and used to predict an intermediate result. Finally, these operations are fused into an iterative algorithm to achieve the training and prediction of a neural network. The performance of the proposed ITSA strategy is verified with some widely used change detection deep learning networks using seven pairs of real remote sensing images. The excellent visual performance and quantitative comparisons from the experiments clearly indicate that detection accuracies of LCCD can be effectively improved when a deep learning network is coupled with the proposed ITSA. For example, compared with some state-of-the-art methods, the quantitative improvement is 0.38%-7.53% in terms of overall accuracy. Moreover, the improvement is robust, generic to both homogeneous and heterogeneous images, and universally adaptive to various neural networks of LCCD. The code will be available at https://github.com/ImgSciGroup/ITSA.

Visible light infrared matching algorithm based on improved SuperPoint and linear converter

In order to solve the problems of high difficulty and mismatch rate in heterogeneous image matching of visible light and infrared images, a deep learning matching algorithm based on improved SuperPoint and linear transformer is proposed. The algorithm first introduces the idea of feature pyramid to construct a feature description branch based on the SuperPoint network structure, and trains it based on the hinge loss function, so as to better learn the multi-scale deep features of visible light and infrared images and increase the similarity of image point pair descriptors with the same name; in the feature matching module, the linear transformer is used to improve the SuperGlue matching algorithm and aggregate features to improve matching performance. The proposed algorithm is experimentally verified on multiple data sets. The results show that compared with the existing algorithms, the algorithm achieves better matching results and improves the matching accuracy.

High-Precision Heterogeneous Satellite Image Manipulation Localization: Feature Point Rules and Semantic Similarity Measurement

Misusing image tampering software makes it easier to manipulate satellite images, leading to a crisis of trust and security concerns in society. This study compares the inconsistencies between heterogeneous images to locate tampered areas and proposes a high-precision heterogeneous satellite image manipulation localization (HSIML) framework to distinguish tampered from real landcover changes, such as artificial constructions, and pseudo-changes, such as seasonal variations. The model operates at the patch level and comprises three modules: The heterogeneous image preprocessing module aligns heterogeneous images and filters noisy data. The feature point constraint module mitigates the effects of lighting and seasonal variations in the images by performing feature point matching, applying filtering rules to conduct an initial screening to identify candidate tampered patches. The semantic similarity measurement module designs a classification network to assess RS image feature saliency. It determines image consistency based on the similarity of semantic features and implements IML using predefined classification rules. Additionally, a dataset for IML is constructed based on satellite images. Extensive experiments compared with existing SOTA models demonstrate that our method achieved the highest F1 score in both localization accuracy and robustness tests and demonstrates the capability for handling large-scale areas.

Variation in forest root image annotation by experts, novices, and AI

BackgroundThe manual study of root dynamics using images requires huge investments of time and resources and is prone to previously poorly quantified annotator bias. Artificial intelligence (AI) image-processing tools have been successful in overcoming limitations of manual annotation in homogeneous soils, but their efficiency and accuracy is yet to be widely tested on less homogenous, non-agricultural soil profiles, e.g., that of forests, from which data on root dynamics are key to understanding the carbon cycle. Here, we quantify variance in root length measured by human annotators with varying experience levels. We evaluate the application of a convolutional neural network (CNN) model, trained on a software accessible to researchers without a machine learning background, on a heterogeneous minirhizotron image dataset taken in a multispecies, mature, deciduous temperate forest.ResultsLess experienced annotators consistently identified more root length than experienced annotators. Root length annotation also varied between experienced annotators. The CNN root length results were neither precise nor accurate, taking ~ 10% of the time but significantly overestimating root length compared to expert manual annotation (p = 0.01). The CNN net root length change results were closer to manual (p = 0.08) but there remained substantial variation.ConclusionsManual root length annotation is contingent on the individual annotator. The only accessible CNN model cannot yet produce root data of sufficient accuracy and precision for ecological applications when applied to a complex, heterogeneous forest image dataset. A continuing evaluation and development of accessible CNNs for natural ecosystems is required.

S2HConv-Caps-BiGRU: Deep Learning-Based Heterogeneous Face Recognition Model with Divergent Stages

Real-time images of faces captured in different spectrum bands are considered heterogeneous images. Heterogeneous Face Recognition (HFR) matches faces across domains and is crucial to public safety. This paper proposes an HFR approach based on Deep Neural Networks (DNN). Feature maps are extracted from two images, such as gallery and sketch images, using Squirrel Search Heterogeneous Convolutional-Capsule- Bidirectional Gated Recurrent Unit (S2HConv-Caps-BiGRU). As a method of efficiently recognizing faces, coupled representation similarity metric (CRSM) will use the measure for the similarity of two feature maps. The experimental results will be evaluated to state-of-the-art (SOTA) statistical measures in terms of accuracy, recall, Jaccard score, dice score, mean square error (MSE), image similarity, performance and root mean square error (RMSE). Compared to other SOTA, the model produces the best results. The accuracy value of a CUFS dataset is 98.7%.

Thrombectomy for Stroke With Large Infarct on Noncontrast CT

Recent large infarct thrombectomy trials used heterogeneous imaging modalities and time windows for patient selection. Noncontrast computed tomographic (CT) scan is the most common stroke imaging approach. It remains uncertain whether thrombectomy is effective for patients with large infarcts identified using noncontrast CT alone within 24 hours of stroke onset. To evaluate the effect of thrombectomy in patients with a large infarct on a noncontrast CT scan within 24 hours of onset. Open-label, blinded-end point, bayesian-adaptive randomized trial with interim analyses for early stopping (futility or success) or population enrichment, which was conducted at 47 US academic and community-based stroke thrombectomy centers. Three hundred patients presenting within 24 hours with anterior-circulation, large-vessel occlusion and large infarct on noncontrast CT scan, with Alberta Stroke Program Early CT Scores of 2 to 5, were randomized to undergo thrombectomy or usual care. Enrollment occurred July 16, 2019 to October 17, 2022; final follow-up, January 25, 2023. The intervention patients (n = 152) underwent endovascular treatment using standard thrombectomy devices and usual medical care. Control patients (n = 148) underwent usual medical care alone. The primary efficacy end point was improvement in 90-day functional outcome measured using mean utility-weighted modified Rankin Scale (UW-mRS) scores (range, 0 [death or severe disability] to 10 [no symptoms]; minimum clinically important difference, 0.3). A bayesian model determined the posterior probability that the intervention would be superior to usual care; statistical significance was a 1-sided posterior probability of .975 or more. The primary adverse event end point was 90-day mortality; secondary adverse event end points included symptomatic intracranial hemorrhage and radiographic intracranial hemorrhage. The trial enrolled 300 patients (152 intervention, 148 control; 138 females [46%]; median age, 67 years), without early stopping or enrichment; 297 patients completed the 90-day follow-up. The mean (SD) 90-day UW-mRS score was 2.93 (3.39) for the intervention group vs 2.27 (2.98) for the control group with an adjusted difference of 0.63 (95% credible interval [CrI], -0.09 to 1.34; posterior probability for superiority of thrombectomy, .96). The 90-day mortality was similar between groups: 35.3% (53 of 150) for the intervention group vs 33.3% (49 of 147) for the control group. Six of 151 patients (4.0%) in the intervention group and 2 of 149 (1.3%) in the control group experienced 24-hour symptomatic intracranial hemorrhage. Fourteen patients of 148 (9.5%) in the intervention group vs 4 of 146 (2.7%) in the control group experienced parenchymal hematoma type 1 hemorrhages; 14 (9.5%) in the intervention group vs 5 (3.4%) in the control group experienced parenchymal hematoma type 2 hemorrhages; and 24 (16.2%) in the intervention group vs 9 (6.2%) in the control group experienced subarachnoid hemorrhages. Among patients with a large infarct on noncontrast CT within 24 hours, thrombectomy did not demonstrate improvement in functional outcomes. But the width of the credible interval around the effect estimate includes the possibility of both no important effect and a clinically relevant benefit, so the potential role of thrombectomy with this imaging approach and time window will likely require additional study. ClinicalTrials.gov Identifier: NCT03805308.

AFIRE: Adaptive FusionNet for illumination-robust feature extraction in heterogeneous imaging environments

The fusion of infrared and visible images aims to synthesize a fused image that incorporates richer information by leveraging the distinct characteristics of each modality. However, the disparate quality of input images in terms of infrared and visible light significantly impacts fusion performance. To address this issue, we propose a novel deep adaptive fusion method called Adaptive FusionNet for Illumination-Robust Feature Extraction (AFIRE). This method involves the interactive processing of two input features and dynamically adjusts the fusion weights based on varying illumination conditions. Specifically, we introduce a novel interactive extraction structure during the feature extraction stage for both infrared and visible light, enabling the capture of more complementary information. Additionally, we design a Deep Adaptive Fusion module to assess the quality of input features and perform weighted fusion through a channel attention mechanism. Finally, a new loss function is formulated by incorporating the entropy and median of input images to guide the training of the fusion network. Extensive experiments demonstrate that AFIRE outperforms state-of-the-art methods in preserving pixel intensity distribution and texture details. Source code is available at: https://www.github.com/ISCLab-Bistu/AFIRE.

DCFusion: Difference correlation-driven fusion mechanism of infrared and visible images

In end-to-end image fusion models, the loss function significantly impacts performance. However, most loss functions treat salient and background regions in source images equally, failing to distinguish complementary areas in multimodal images. This limits the model’s ability to effectively integrate information from these regions. Therefore, we propose difference correlation-driven fusion mechanism of infrared and visible images, which called DCFusion. Specifically, the model utilizes a dual-branch interactive network that dynamically fuses cross-modal multi-scale complementary information through element-wise multiplication, effectively integrating region-specific information. We introduce a two-stage method for generating salient target masks that adaptively focus on high-contrast regions in infrared images by analyzing pixel contrasts in local areas. Furthermore, we utilize the salient target masks to create heterogeneous images and design the LSCD loss function to minimize the information gap between the heterogeneous images and the fused image, thereby enhancing the model’s interpretability. Experiments on the RoadScene and TNO datasets show that DCFusion surpasses with existing representativity fusion approaches, achieving state-of-the-art performance in both subjective visual and objective evaluations. Our code will be publicly available at https://github.com/MinLila/DCFusion.

Label-free 3D molecular imaging of living tissues using Raman spectral projection tomography

The ability to image tissues in three dimensions (3D) with label-free molecular contrast at the mesoscale would be a valuable capability in biology and biomedicine. Here, we introduce Raman spectral projection tomography (RSPT) for volumetric molecular imaging with optical sub-millimeter spatial resolution. We have developed a RSPT imaging instrument capable of providing 3D molecular contrast in transparent and semi-transparent samples. We also created a computational pipeline for multivariate reconstruction to extract label-free spatial molecular information from Raman projection data. Using these tools, we demonstrate imaging and visualization of phantoms of various complex shapes with label-free molecular contrast. Finally, we apply RSPT as a tool for imaging of molecular gradients and extracellular matrix heterogeneities in fixed and living tissue-engineered constructs and explanted native cartilage tissues. We show that there exists a favorable balance wherein employing Raman spectroscopy, with its advantages in live cell imaging and label-free molecular contrast, outweighs the reduction in imaging resolution and blurring caused by diffuse photon propagation. Thus, RSPT imaging opens new possibilities for label-free molecular monitoring of tissues.

Flood inundation monitoring using multi-source satellite imagery: a knowledge transfer strategy for heterogeneous image change detection

Flood emergency mapping is essential for flood management, often requiring near real-time extraction of large-scale flood extents by combining pre- and post-event multi-source remote sensing images. Pre-event optical imagery delineates the normal water extent, providing a benchmark for estimation of post-flood changes. Synthetic aperture radar (SAR) imagery provides rapid and accurate interpretation of flood inundation extent during the rainy period. Post-classification comparison is one of the fundamental methods for flood mapping of multi-source imagery, as the image characteristics of SAR and optical imagery are inherently different. However, the accuracy of flood mapping depends on the accuracy of water body delineation from single temporal imagery, which can lead to error propagation in flood extent determination. Change detection methods can extract flood extent directly from pre- and post-event imagery, but the models need to learn a consistent flood feature description between optical and SAR imagery from a limited flood dataset. Here, we propose an automatic method for flood inundation extent extraction using pre- and post-event multi-source imagery that requires only a small number of pseudo-change labels. The methodology adopts a pseudo-Siamese network framework as a Heterogeneous - Flood Inundation Extraction Network (H-FIENet), to detect flood extent between the pre‐ and post-event heterogeneous images. Spatially inconsistent multi-source pre- and post-event imagery was used to create a pseudo flood dataset, and this dataset was used to transfer water body feature knowledge from the pre-trained model to the flood extraction model using a cross-task knowledge transfer strategy. Pre- and post-event images from different satellite sources and different flood scenarios were used to evaluate the performance of H-FIENet. We found that: (1) The methodology produced an overall accuracy of over 0.94 for flood inundation extraction from multi-source heterogeneous pre- and post-event imagery. (2) H-FIENet can detect both expansion and recession of the flood extent using any dual-temporal imagery. Our work makes it possible to automate time-series flood monitoring from multi-source remote sensing imagery.

From Syntactic to Semantic Interoperability Using a Hyperontology in the Oncology Domain.

Interoperability is crucial to overcoming various challenges of data integration in the healthcare domain. While OMOP and FHIR data standards handle syntactic heterogeneity among heterogeneous data sources, ontologies support semantic interoperability to overcome the complexity and disparity of healthcare data. This study proposes an ontological approach in the context of the EUCAIM project to support semantic interoperability among distributed big data repositories that have applied heterogeneous cancer image data models using a semantically well-founded Hyperontology for the oncology domain.

Determining Thresholds for Optimal Adaptive Discrete Cosine Transformation

The discrete cosine transform (DCT) is widely used for image and video compression. Lossy algorithms such as JPEG, WebP, BPG and many others are based on it. Multiple modifications of DCT have been developed to improve its performance. One of them is adaptive DCT (ADCT) designed to deal with heterogeneous image structure and it may be found, for example, in the HEVC video codec. Adaptivity means that the image is divided into an uneven grid of squares: smaller ones retain information about details better, while larger squares are efficient for homogeneous backgrounds. The practical use of adaptive DCT algorithms is complicated by the lack of optimal threshold search algorithms for image partitioning procedures. In this paper, we propose a novel method for optimal threshold search in ADCT using a metric based on tonal distribution. We define two thresholds: pm, the threshold defining solid mean coloring, and ps, defining the quadtree fragment splitting. In our algorithm, the values of these thresholds are calculated via polynomial functions of the tonal distribution of a particular image or fragment. The polynomial coefficients are determined using the dedicated optimization procedure on the dataset containing images from the specific domain, urban road scenes in our case. In the experimental part of the study, we show that ADCT allows a higher compression ratio compared to non-adaptive DCT at the same level of quality loss, up to 66% for acceptable quality. The proposed algorithm may be used directly for image compression, or as a core of video compression framework in traffic-demanding applications, such as urban video surveillance systems.

Enhanced Characterization of Fractured Zones in Bedrock Using Hydraulic Tomography through Joint Inversion of Hydraulic Head and Flux Data

Hydraulic tomography (HT) is a promising technique for high-resolution imaging of subsurface heterogeneity, which addresses the limitations of traditional methods, such as borehole drilling and geophysical surveys. This study focuses on the application of HT to detect and characterize fractured zones in bedrock and addresses the gap in the understanding of the role of distributed flux data in the joint inversion of hydraulic head and flux data. By conducting synthetic injection tests and using sequential successive linear estimators for inversion, the study explores the effectiveness of combining limited head data with distributed temperature sensing (A-DTS)-derived flux data. The findings highlight the fact that integrating flux data significantly enhances the accuracy of identifying fracture permeability characteristics, even when head data is sparse. This approach not only improves the resolution of hydraulic conductivity fields but also offers a cost-effective strategy for practical field applications. The results underscore the potential of HT to enhance our understanding of groundwater flow and contaminant transport in fractured media, which has important implications for carbon capture, enhanced geothermal systems, and radioactive waste disposal.

A handheld polarimetric imaging device and calibration technique for accurate mapping of terahertz Stokes vectors

In recent years, handheld and portable terahertz instruments have been in rapid development for various applications ranging from non-destructive testing to biomedical imaging and sensing. For instance, we have deployed our Portable Handheld Spectral Reflection (PHASR) Scanners for in vivo full-spectroscopic imaging of skin burns in large animal models in operating room settings. In this paper, we debut the polarimetric version of the PHASR Scanner, and describe a generalized calibration technique to map the spatial and spectral dependence of the Jones matrix of an imaging scanner across its field of view. Our design is based on placement of two orthogonal photoconductive antenna (PCA) detectors separated by a polarizing beam splitter in the PHASR Scanner housing. We show that as few as three independent measurements of a well-characterized polarimetric calibration target are sufficient to determine the polarization state of the incident beam at the sample location, as well as to extract the Jones propagation matrix from the sample location to the detectors. We have tested the accuracy of our scanner by validating polarimetric measurements obtained from a birefringent crystal rotated to various angles, as compared to the theoretically predicted response of the sample. This new version of our PHASR scanner can be used for high-speed imaging and investigation of heterogeneity of polarization-sensitive samples in the field.