Journal of Imaging最新文献

Human activity recognition (HAR) has emerged as a transformative field with widespread applications, leveraging diverse sensor modalities to accurately identify and classify human activities. This paper provides a comprehensive review of HAR techniques, focusing on the integration of sensor-based, vision-based, and hybrid methodologies. It explores the strengths and limitations of commonly used modalities, such as RGB images/videos, depth sensors, motion capture systems, wearable devices, and emerging technologies like radar and Wi-Fi channel state information. The review also discusses traditional machine learning approaches, including supervised and unsupervised learning, alongside cutting-edge advancements in deep learning, such as convolutional and recurrent neural networks, attention mechanisms, and reinforcement learning frameworks. Despite significant progress, HAR still faces critical challenges, including handling environmental variability, ensuring model interpretability, and achieving high recognition accuracy in complex, real-world scenarios. Future research directions emphasise the need for improved multimodal sensor fusion, adaptive and personalised models, and the integration of edge computing for real-time analysis. Additionally, addressing ethical considerations, such as privacy and algorithmic fairness, remains a priority as HAR systems become more pervasive. This study highlights the evolving landscape of HAR and outlines strategies for future advancements that can enhance the reliability and applicability of HAR technologies in diverse domains.

The objective of this study is to propose an advanced image enhancement strategy to address the challenge of reducing radiation doses in pediatric renal scintigraphy. Data from a public dynamic renal scintigraphy database were used. Based on noisier images, four denoising neural networks (DnCNN, UDnCNN, DUDnCNN, and AttnGAN) were evaluated. To evaluate the quality of the noise reduction, with minimal detail loss, the kidney signal-to-noise ratio (SNR) and multiscale structural similarity (MS-SSIM) were used. Although all the networks reduced noise, UDnCNN achieved the best balance between SNR and MS-SSIM, leading to the most notable improvements in image quality. In clinical practice, 100% of the acquired data are summed to produce the final image. To simulate the dose reduction, we summed only 50%, simulating a proportional decrease in radiation. The proposed deep-learning approach for image enhancement ensured that half of all the frames acquired may yield results that are comparable to those of the complete dataset, suggesting that it is feasible to reduce patients' exposure to radiation. This study demonstrates that the neural networks evaluated can markedly improve the renal scintigraphic image quality, facilitating high-quality imaging with lower radiation doses, which will benefit the pediatric population considerably.

Chest imaging plays a pivotal role in screening and monitoring patients, and various predictive artificial intelligence (AI) models have been developed in support of this. However, little is known about the effect of decreasing the radiation dose and, thus, image quality on AI performance. This study aims to design a low-dose simulation and evaluate the effect of this simulation on the performance of CNNs in plain chest radiography. Seven pathology labels and corresponding images from Medical Information Mart for Intensive Care datasets were used to train AI models at two spatial resolutions. These 14 models were tested using the original images, 50% and 75% low-dose simulations. We compared the area under the receiver operator characteristic (AUROC) of the original images and both simulations using DeLong testing. The average absolute change in AUROC related to simulated dose reduction for both resolutions was <0.005, and none exceeded a change of 0.014. Of the 28 test sets, 6 were significantly different. An assessment of predictions, performed through the splitting of the data by gender and patient positioning, showed a similar trend. The effect of simulated dose reductions on CNN performance, although significant in 6 of 28 cases, has minimal clinical impact. The effect of patient positioning exceeds that of dose reduction.

In recent years, there has been growing interest in taking advantage of the technological progress in information technology and computer science to enhance the synergy between multidisciplinary organisations with a mutual objective of improving scientific knowledge and engaging society in cultural activities. Such an example of collaboration networks includes those where governmental, scientific and cultural institutions work in unison to provide services that support research through the use of technology while disseminating information and promoting cultural heritage. Here, we present a case study implementing the results of the work between multidisciplinary departments of the NOVA University Lisbon and third-party cultural heritage organisations. In particular, a mobile and desktop PC application uses augmented reality to showcase results obtained from analysis of artwork by Amadeo de Souza-Cardoso using artificial intelligence. The mobile application is intended to be used to enhance museum visitors' experience and strengthen the link between scientific, governmental, and heritage organisations.

Semantic segmentation in deep learning is a crucial area of research within computer vision, aimed at assigning specific labels to each pixel in an image. The segmentation of crops, plants, and weeds has significantly advanced the application of deep learning in precision agriculture, leading to the development of sophisticated architectures based on convolutional neural networks (CNNs). This study proposes a segmentation algorithm for identifying plants and weeds using broadband multispectral images. In the first part of this algorithm, we utilize the PIF-Net model for feature extraction and fusion. The resulting feature map is then employed to enhance an optimized U-Net model for semantic segmentation within a broadband system. Our investigation focuses specifically on scenes from the CAVIAR dataset of multispectral images. The proposed algorithm has enabled us to effectively capture complex details while regulating the learning process, achieving an impressive overall accuracy of 98.2%. The results demonstrate that our approach to semantic segmentation and the differentiation between plants and weeds yields accurate and compelling outcomes.

Background: This study aims to objectively evaluate the overall quality of colonoscopies using a specially trained deep learning-based semantic segmentation neural network. This represents a modern and valuable approach for the analysis of colonoscopy frames. Methods: We collected thousands of colonoscopy frames extracted from a set of video colonoscopy files. A color-based image processing method was used to extract color features from specific regions of each colonoscopy frame, namely, the intestinal mucosa, residues, artifacts, and lumen. With these features, we automatically annotated all the colonoscopy frames and then selected the best of them to train a semantic segmentation network. This trained network was used to classify the four region types in a different set of test colonoscopy frames and extract pixel statistics that are relevant to quality evaluation. The test colonoscopies were also evaluated by colonoscopy experts using the Boston scale. Results: The deep learning semantic segmentation method obtained good results, in terms of classifying the four key regions in colonoscopy frames, and produced pixel statistics that are efficient in terms of objective quality assessment. The Spearman correlation results were as follows: BBPS vs. pixel scores: 0.69; BBPS vs. mucosa pixel percentage: 0.63; BBPS vs. residue pixel percentage: -0.47; BBPS vs. Artifact Pixel Percentage: -0.65. The agreement analysis using Cohen's Kappa yielded a value of 0.28. The colonoscopy evaluation based on the extracted pixel statistics showed a fair level of compatibility with the experts' evaluations. Conclusions: Our proposed deep learning semantic segmentation approach is shown to be a promising tool for evaluating the overall quality of colonoscopies and goes beyond the Boston Bowel Preparation Scale in terms of assessing colonoscopy quality. In particular, while the Boston scale focuses solely on the amount of residual content, our method can identify and quantify the percentage of colonic mucosa, residues, and artifacts, providing a more comprehensive and objective evaluation.

Sediment transport and shoreline changes causing shoreline morphodynamic evolution are key indicators of a coastal structure's operational continuity. To reduce the computational costs associated with sediment transport modelling tools, a novel procedure based on the combination of a support vector machine for image classification and a trained neural network to extrapolate the shore evolution is presented here. The current study focuses on the coastal area over the Amir-Abad port, using high-resolution satellite images. The real conditions of the study domain between 2004 and 2023 are analysed, with the aim of investigating changes in the shore area, shoreline position, and sediment appearance in the harbour basin. The measurements show that sediment accumulation increases by approximately 49,000 m²/y. A portion of the longshore sediment load is also trapped and deposited in the harbour basin, disrupting the normal operation of the port. Afterwards, satellite images were used to quantitatively analyse shoreline changes. A neural network is trained to predict the remaining time until the reservoir is filled (less than a decade), which is behind the west arm of the rubble-mound breakwaters. Harbour utility services will no longer be offered if actions are not taken to prevent sediment accumulation.

Optical imaging is an excellent non-invasive method for viewing visceral organs. Most importantly, it is safer as compared to ionizing radiation-based methods like X-rays. By making use of the properties of photons, this technique generates high-resolution images of cells, molecules, organs, and tissues using visible, ultraviolet, and infrared light. Moreover, optical imaging enables real-time evaluation of soft tissue properties, metabolic alterations, and early disease markers in real time by utilizing a variety of techniques, including fluorescence and bioluminescence. Innovative biocompatible fluorescent probes that may provide disease-specific optical signals are being used to improve diagnostic capabilities in a variety of clinical applications. However, despite these promising advancements, several challenges remain unresolved. The primary obstacle includes the difficulty of developing efficient fluorescent probes, and the tissue autofluorescence, which complicates signal detection. Furthermore, the depth penetration restrictions of several imaging modalities limit their use in imaging of deeper tissues. Additionally, enhancing biocompatibility, boosting fluorescent probe signal-to-noise ratios, and utilizing cutting-edge imaging technologies like machine learning for better image processing should be the main goals of future research. Overcoming these challenges and establishing optical imaging as a fundamental component of modern medical diagnoses and therapeutic treatments would require cooperation between scientists, physicians, and regulatory bodies.

An attempt was made to quantify soil properties using hyperspectral remote-sensing techniques and machine-learning algorithms. In total, 100 soil samples representing various locations and soil-nutrient statuses were collected, and the samples were analyzed for soil pH, EC, soil organic carbon, available nitrogen (AN), available phosphorus (AP), and available potassium (AK) by following standard methods. Soil had a wide range of properties, i.e., pH varied from 5.62 to 8.49, EC varied from 0.08 to 1.78 dS/m, soil organic carbon varied from 0.23 to 0.94%, available nitrogen varied from 154 to 344 kg/ha, available phosphorus varied from 9.5 to 25.5 kg/ha, and available potassium varied from 131 to 747 kg/ha. The same set of soil samples were subjected to spectral reflectance measurement using SVC GER 1500 Spectroradiometer (spectral range: 350 to 1050 nm). The measured spectral signatures of various soils were organized for developing a spectral library and for deriving various spectral indices to correlate with soil properties to quantify the nutrients. The soil samples were partitioned into 60:40 ratios for training and validation, respectively. In order to select optimum bands (wavelength) from the soil spectra, we have employed metaheuristic algorithms i.e., Particle Swarm Optimization (PSO), Moth-Flame optimization (MFO), Flower Pollination Optimization (FPO), and Battle Royale Optimization (BRO) algorithm. Further partial least square regression (PLSR) was used to find the latent variable and to evaluate various algorithms for their performance in predicting soil properties. The results indicated that nutrients could be quantified from spectral reflectance measurement with fair to good accuracy through the Battle Royale Optimization technique with a R2 value of 0.45, 0.32, 0.48, 0.21, 0.71, and 0.35 for pH, EC, soil organic carbon, available-N, available-P, and available-K, respectively.

Contrast-enhanced mammography (CEM) combines morphological and functional imaging, enhancing breast cancer (BC) diagnosis. This study investigates the relationship between CEM morphodynamic features and histopathological characteristics of BC. In this prospective study, 50 female patients (mean age: 57.2 ± 13.7 years) with BI-RADS 4-5 lesions underwent CEM followed by surgical excision between December 2022 and May 2024. Low-energy and recombined CEM images were analyzed for breast composition, lesion characteristics, and enhancement patterns, while histopathological evaluation included tumor size, histotype, grade, lymphovascular invasion, and immunophenotype. Spearman rank correlation and multivariable regression analysis were used to evaluate the relationship between CEM findings and histopathological characteristics. Tumor size on CEM strongly correlated with histopathological tumor size (ρ = 0.788, p < 0.001) and was associated with high-grade lesions (p = 0.017). Non-circumscribed margins were linked to a Luminal-B subtype (p = 0.001), while high lesion conspicuity was associated with Luminal-B and triple-negative BC (p = 0.001) and correlated with larger tumors (ρ = 0.517, p < 0.001). Background parenchymal enhancement was negatively correlated with age (ρ = -0.286, p = 0.049). CEM provides critical insights into BC, demonstrating significant relationship between imaging features and histopathological characteristics. These findings highlight CEM's potential as a reliable tool for tumor size estimation, subtype characterization, and prognostic assessment, suggesting its role as an alternative to MRI, particularly for patients with contraindications.