Small Imaging Research Articles

Individualized small bowel preparation for computed tomography enterography: A prospective randomized controlled trial.

The study aims to evaluate the feasibility of body mass index (BMI)-based individualized small bowel preparation for computed tomography enterography (CTE). In this prospective randomized controlled study, patients undergoing CTE were randomly assigned to the individualized group or standardized group. Those in individualized group were given different volumes of mannitol solution based on BMI (1000mL for patients with BMI<18.5kg/m2, 1500mL for patients with 18.5kg/m2≤BMI<25kg/m2 and 2000mL for patients with BMI≥25kg/m2) while patients in the standardized group were all asked to consume 1500-mL mannitol solution. CTE images were reviewed by two experienced radiologists blindly. Each segment of the small bowel was assessed for small bowel image quality and disease detection rates. Patients were invited to record a diary regarding adverse events and acceptance. A total of 203 patients were enrolled and randomly divided into two groups. For patients with BMI<18.5kg/m2, 1000-mL mannitol solution permitted a significantly lower rate of flatulence (P=0.045) and defecating frequency (P=0.011) as well as higher acceptance score (P=0.015), but did not affect bowel image quality and diseases detection compared with conventional dosage. For patients with BMI≥25kg/m2, 2000-mL mannitol solution provided better overall image quality (P=0.033) but comparable rates of adverse events and patients' acceptance compared with conventional dosage. Individualized bowel preparation could achieve both satisfactory image quality and patients' acceptance thus might be an acceptable alternative in CTE.

Geometric transformation adaptive optics (GTAO) for volumetric deep brain imaging through gradient-index lenses

The advance of genetic function indicators has enabled the observation of neuronal activities at single-cell resolutions. A major challenge for the applications on mammalian brains is the limited optical access depth. Currently, the method of choice to access deep brain structures is to insert miniature optical components. Among these validated miniature optics, the gradient-index (GRIN) lens has been widely employed for its compactness and simplicity. However, due to strong fourth-order astigmatism, GRIN lenses suffer from a small imaging field of view, which severely limits the measurement throughput and success rate. To overcome these challenges, we developed geometric transformation adaptive optics (GTAO), which enables adaptable achromatic large-volume correction through GRIN lenses. We demonstrate its major advances through in vivo structural and functional imaging of mouse brains. The results suggest that GTAO can serve as a versatile solution to enable large-volume recording of deep brain structures and activities through GRIN lenses.

Improving small sample medical image segmentation using CBAM: Insights from two datasets

Medical image segmentation is one of the key tasks in the medical field and is crucial for accurate lesion detection and treatment planning. However, the small sample problem has been one of the challenges in medical image segmentation. In this study, the small sample medical image segmentation problem is evaluated experimentally based on two different datasets, the VOC dataset and the esophageal medical images. In the VOC dataset experiments, we used the code from the BAM project on GitHub as a benchmark and compared it. Additionally, we improved the benchmark code by adding the Attention Mechanism (CBAM), placing its position before the support set undergoes global average pooling. By comparing the experimental results, we found that the model with the added CBAM achieved better segmentation results under small sample conditions. In the esophageal medical image experiments, the code from the BAM project was also used as a benchmark and a similar experimental design was performed. We further added CBAM by placing it before the global average pooling of the support set. The experimental results show that the model with the addition of CBAM achieves a significant improvement in the small sample segmentation task of esophageal medical images. To summarize, this study validates the effectiveness of using the BAM model in the small-sample medical image segmentation task on two different datasets. By adding CBAM, our model exhibits better segmentation performance under small sample conditions. These findings provide useful insights for small-sample medical image segmentation and are important for future research and practice.

Abstract B039: SAMPLER: Unsupervised representations of whole slide images for tumor phenotype prediction

Abstract Deep learning has revolutionized digital pathology, allowing for automatic analysis of hematoxylin and eosin (H&amp;E) stained whole slide images (WSIs) for various tasks, such as the classification of cancers into clinical subtypes. In such analyses, WSIs are broken down into smaller images, hereafter called tiles, for efficient processing, with each tile encoded by a deep learning backbone. To reconstruct a slide level representation, a widely applied approach is to combine tile-level features using attention-based deep learning models for each downstream prediction task. These training strategies are (a) computationally intensive, (b) challenging to optimize, (c) sensitive to stain variations across datasets, and (d) require sufficiently large and labeled datasets for supervised training, which are not always available. We propose SAMpling of multiscale empirical distributions for LEarning Representations (SAMPLER), a fully statistical approach to generate WSI representations by encoding the empirical cumulative distribution function (CDF) of multiscale tile features. We evaluated this approach by training logistic regression classifiers from SAMPLER representations. SAMPLER-based classifiers were able to accurately separate subtypes of breast carcinoma (BRCA: AUC=0.911 ± 0.029), subtypes of non-small cell lung carcinoma (NSCLC: AUC=0.940±0.018), and subtypes of renal cell carcinoma (RCC: AUC=0.987±0.006) diagnostic slides of the cancer genome atlas (TCGA). Performance was similar to fully deep learning attention models but &amp;gt;100 times faster. We further validated out models on external test sets. Histopathological review confirms that SAMPLER-identified high attention tiles contain tumor morphological features specific to the tumor type, while low attention tiles contain fibrous stroma, blood, or tissue folding artifacts. SAMPLER is a fast and accurate approach for analyzing WSIs, with greatly improved scalability over attention methods to benefit digital pathology analysis. Citation Format: Patience Mukashyaka, Todd B. Sheridan, Ali Foroughi pour, Jeffrey H. Chuang. SAMPLER: Unsupervised representations of whole slide images for tumor phenotype prediction [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Translating Cancer Evolution and Data Science: The Next Frontier; 2023 Dec 3-6; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(3 Suppl_2):Abstract nr B039.

Emoji’s sentiment score estimation using convolutional neural network with multi-scale emoji images

Emojis are any small images, symbols, or icons that are used in social media. Several well-known emojis have been ranked and sentiment scores have been assigned to them. These ranked emojis can be used for sentiment analysis; however, many new released emojis have not been ranked and have no sentiment score yet. This paper proposes a new method to estimate the sentiment score of any unranked emotion emoji from its image by classifying it into the class of the most similar ranked emoji and then estimating the sentiment score using the score of the most similar emoji. The accuracy of sentiment score estimation is improved by using multi-scale images. The ranked emoji image data set consisted of 613 classes with 161 emoji images from three different platforms in each class. The images were cropped to produce multi-scale images. The classification and estimation were performed by using convolutional neural network (CNN) with multi-scale emoji images and the proposed voting algorithm called the majority voting with probability (MVP). The proposed method was evaluated on two datasets: ranked emoji images and unranked emoji images. The accuracies of sentiment score estimation for the ranked and unranked emoji test images are 98% and 51%, respectively.

Development of an integrated dual-modality 3D bioluminescence tomography and ultrasound imaging system for small animal tumor imaging.

Ultrasound (US) is a valuable tool for imaging soft tissue and visualizing tumor contours. Taking the benefits of US, we presented an integrated dual-modality imaging system in this paper that achieves three-dimensional (3D) bioluminescence tomography (BLT) with multi-view bioluminescence images and 3D US imaging. The purpose of this system is to perform non-invasive, long-term monitoring of tumor growth in 3D images. US images can enhance the accuracy of the 3D BLT reconstruction and the bioluminescence dose within an object. Furthermore, an integrated co-registered scanning geometry was used to capture the fused BLT and US images. We validated the system with an in vivo experiment involving tumor-bearing mice. The results demonstrated the feasibility of reconstructing 3D BLT images in the tumor region using 3D US images. We used the dice coefficient and locational error to evaluate the similarity between the reconstructed source region and the actual source region. The dice coefficient was 88.5%, and the locational error was 0.4 mm when comparing the BLT and 3D US images. The hybrid BLT/US system could provide significant benefits for reconstructing the source of tumor location and conducting quantitative analysis of tumor size.

Research Advancements on the Correlation Between Spontaneous Intracerebral Hemorrhage of Different Etiologies and Imaging Markers of Cerebral Small Vessel Disease.

The purpose of this review is to identify the correlation between ICH and CSVD imaging markers under SMASH-U classification by searching and analyzing a large number of literatures in recent years, laying a theoretical foundation for future clinical research. At the same time, by collecting clinical data to evaluate patient prognosis, analyzing whether there are differences or supplements between clinical trial conclusions and previous theories, and ultimately guiding clinical diagnosis and treatment through the analysis of imaging biomarkers. In this review, by searching CNKI, Web of Science, PubMed, FMRS and other databases, the use of "spontaneous intracerebral hemorrhage", "hypertensive hemorrhagic cerebral small vessel disease", "cerebral small vessel disease imaging", "Based cerebral small vessel diseases", "SMASH the -u classification" and their Chinese equivalents for the main search term. We focused on reading and analyzing hundreds of relevant literatures in the last decade from August 2011 to April 2020, and also included some earlier literatures with conceptual data sources. After screening and ranking the degree of relevance to this study, sixty of them were cited for analysis and elaboration. In patients with ICH, the number of cerebral microbleeds in lobes, basal ganglia, and the deep brain is positively correlated with ICH volume and independently correlated with neurological functional outcomes; white matter hyperintensity severity is positively correlated with ICH recurrence risk; multiple lacunar infarction independently predict the risk of ICH; severe brain atrophy is an independent risk factor for a poor prognosis in the long term in patients diagnosed with ICH; and the number of enlarged perivascular spaces is correlated with ICH recurrence. However, small subcortical infarct and ICH are the subject of few studies. Higher CSVD scores are independently associated with functional outcomes at 90 days in patients diagnosed with ICH.

Fast and Precise Detection of Dense Soybean Seedlings Images Based on Airborne Edge Device

During the growth stage of soybean seedlings, it is crucial to quickly and precisely identify them for emergence rate assessment and field management. Traditional manual counting methods have some limitations in scenarios with large-scale and high-efficiency requirements, such as being time-consuming, labor-intensive, and prone to human error (such as subjective judgment and visual fatigue). To address these issues, this study proposes a rapid detection method suitable for airborne edge devices and large-scale dense soybean seedling field images. For the dense small target images captured by the Unmanned Aerial Vehicle (UAV), the YOLOv5s model is used as the improvement benchmark in the technical solution. GhostNetV2 is selected as the backbone feature extraction network. In the feature fusion stage, an attention mechanism—Efficient Channel Attention (ECA)—and a Bidirectional Feature Pyramid Network (BiFPN) have been introduced to ensure the model prioritizes the regions of interest. Addressing the challenge of small-scale soybean seedlings in UAV images, the model’s input size is set to 1280 × 1280 pixels. Simultaneously, Performance-aware Approximation of Global Channel Pruning for Multitask CNNs (PAGCP) pruning technology is employed to meet the requirements of mobile or embedded devices. The experimental results show that the identification accuracy of the improved YOLOv5s model reached 92.1%. Compared with the baseline model, its model size and total parameters were reduced by 76.65% and 79.55%, respectively. Beyond these quantitative evaluations, this study also conducted field experiments to verify the detection performance of the improved model in various scenarios. By introducing innovative model structures and technologies, the study aims to effectively detect dense small target features in UAV images and provide a feasible solution for assessing the number of soybean seedlings. In the future, this detection method can also be extended to similar crops.

Evaluation of the effectiveness of the Indian government’s policies to strengthen health warning labels on smokeless tobacco products: findings from the 2010–2019 Tobacco Control Project India Surveys

BackgroundSmokeless tobacco (SLT) packaging in India had a single symbolic (a scorpion) health warning label (HWL) in 2009 covering 40% of the front surface. In 2011, it was replaced with...

Vessel density mapping of small cerebral vessels on 3D high resolution black blood MRI

Small cerebral blood vessels are largely inaccessible to existing clinical in vivo imaging technologies. This study aims to present a novel analysis pipeline for vessel density mapping of small cerebral blood vessels from high-resolution 3D black-blood MRI at 3T. Twenty-eight subjects (10 under 35 years old, 18 over 60 years old) were imaged with the T1-weighted turbo spin-echo with variable flip angles (T1w TSE-VFA) sequence optimized for black-blood small vessel imaging with iso-0.5 mm spatial resolution (interpolated from 0.51×0.51×0.64 mm3) at 3T. Hessian-based vessel segmentation methods (Jerman, Frangi and Sato filter) were evaluated by vessel landmarks and manual annotation of lenticulostriate arteries (LSAs). Using optimized vessel segmentation, large vessel pruning and non-linear registration, a semiautomatic pipeline was proposed for quantification of small vessel density across brain regions and further for localized detection of small vessel changes across populations. Voxel-level statistics was performed to compare vessel density between two age groups. Additionally, local vessel density of aged subjects was correlated with their corresponding gross cognitive and executive function (EF) scores using Montreal Cognitive Assessment (MoCA) and EF composite scores compiled with Item Response Theory (IRT). Jerman filter showed better performance for vessel segmentation than Frangi and Sato filter which was employed in our pipeline. Small cerebral blood vessels including small artery, arterioles, small veins, and venules on the order of a few hundred microns can be delineated using the proposed analysis pipeline on 3D black-blood MRI at 3T. The mean vessel density across brain regions was significantly higher in young subjects compared to aged subjects. In the aged subjects, localized vessel density was positively correlated with MoCA and IRT EF scores. The proposed pipeline is able to segment, quantify, and detect localized differences in vessel density of small cerebral blood vessels based on 3D high-resolution black-blood MRI. This framework may serve as a tool for localized detection of small vessel density changes in normal aging and cerebral small vessel disease.

Data distillation in computational pathology by choosing few representants of the original variance: A use case in ovarian cancer

In computational pathology, a typical Whole Slide Image may easily reach a size of 3–4 Gigabytes, while a database, with hundred of cases, might reach Terabytes. In this scenario, training any model is expensive and therefore the possibility of reducing the size of the training data is appealing. This paper presents a method to summarize the information of a data set, a function of the variance, by selecting the most informative samples. The first step of the whole strategy consists in projecting small image patches (100 × 100) to a feature space and discretize it with a simple k-means to obtain the feature space vocabulary. The groups obtained by the k-means are the vocabulary words and therefore any small patch is represented by the centroid of the group to which such patch is projected. A second step, a probabilistic Latent Semantic Analysis constructs groups of words known as topics by computing frequencies of words in documents, which are larger patches containing between 450 and 500 small patches. A third step collects the documents representing 80 % of the topic variance, their patches are assembled and a Singular Value Decomposition (SVD) is applied to these patches. The Data Distillation process chooses only the small patches belonging to the topics showing higher variance in the matrix of eigenvalues S from the SVD decomposition. The method efficacy was evaluated by comparing the performance of models trained either with 40 % of an actual ovarian cancer dataset selected using this method and the entire dataset without any selection. Results show the F-score obtained with these two sets was similar, about 0.87, with different classifiers, namely Support Vector Machine and a Multilayer Perceptron.

Quantitative Analysis of Multimodal MRI Markers and Clinical Risk Factors for Cerebral Small Vessel Disease Based on Deep Learning.

Cerebral small vessel disease lacks specific clinical manifestations, and extraction of valuable features from multimodal images is expected to improve its diagnostic accuracy. In this study, we used deep learning techniques to segment cerebral small vessel disease imaging markers in multimodal magnetic resonance images and analyze them with clinical risk factors. We recruited 211 lacunar stroke patients and 83 control patients. The patients' cerebral small vessel disease markers were automatically segmented using a V-shaped bottleneck network, and the number and volume were calculated after manual correction. The segmentation results of the V-shaped bottleneck network for white matter hyperintensity and recent small subcortical infarction were in high agreement with the ground truth (DSC>0.90). In small lesion segmentation, cerebral microbleed (average recall=0.778; average precision=0.758) and perivascular spaces (average recall=0.953; average precision=0.923) were superior to lacunar infarct (average recall=0.339; average precision=0.432) in recall and precision. Binary logistic regression analysis showed that age, systolic blood pressure, and total cerebral small vessel disease load score were independent risk factors for lacunar stroke (P<0.05). Ordered logistic regression analysis showed age was positively correlated with cerebral small vessel disease load score and total cholesterol was negatively correlated with cerebral small vessel disease score (P<0.05). Lacunar stroke patients exhibited higher cerebral small vessel disease imaging markers, and age, systolic blood pressure, and total cerebral small vessel disease score were independent risk factors for lacunar stroke patients. V-shaped bottleneck network segmentation network based on multimodal deep learning can segment and quantify various cerebral small vessel disease lesions to some extent.

16-channel snapshot multispectral imaging based on integrated Fabry Perot microcavity array

&lt;sec&gt;By designing and fabricating a narrow-band Fabry-Perot multi-beam interference spectroscopic microcavity array, and integrating it with a visible light detector focal plane array, we demonstrate a small compact multispectral imaging detector. The micro-cavity filter array with 4×4 basic repeating units and a total of 2048×2048 pixels is obtained on a quartz substrate by the four-fractal combination lithograph-etching process. Then the micro miniatured multispectral imaging detector is formed by fitting with the detector chip. The depth and precision of the etching will determine the distribution and offset of the central wavelength of the narrowband spectral channel respectively. The results show that the etching rate of reactive ion is (3.6 ± 0.2) Å/s, and the process is stable and controllable. Due to the different etching depths, the basic repeating unit forms 16 different levels of steps, and the process achieves the design expectation well.&lt;/sec&gt;&lt;sec&gt;The results are obtained as follows: the response spectrum peak of the microcavity array sample varies from 520 to 680 nm, the free spectrum range is 160 nm, the full width at the half-peak is less than 10 nm, the transmittance is about 70%, the relative half-width of the transmittance peak at 590 nm is 1.19%, and the waveform coefficient is 2.78. A 16-channel multispectral camera is constructed by using the optical micro-precision assembly device to realize the precise alignment and the fitting of the micro-cavity filter array and the image sensor. Xenon lamp and monochromator are used as tunable wavelength monochromatic cooperative light source to detect the effect of 16-channel snapshot multispectral imaging on a pixel scale. The results show that the multispectral imaging detector has 16 different narrow-band response spectra. The characteristic spectrum of the target can be clearly distinguished by spectral channel.&lt;/sec&gt;&lt;sec&gt;When imaging the target with known spectral characteristics, for a certain frame of multi-spectral image, selecting a suitable spectral channel can eliminate the background in the field of view through image subtraction and improve the contrast of the target. In the dark room condition, we take the LED light source with center wavelength varying in a range between 528 and 589 nm as the target (the wavelength coincides with the working wavelength of the spectral channel), and effectively suppress the background through the spectral differential intensity subtraction, which can improve the accuracy and sensitivity of the target capture. The 16-channel snapshot multi-spectral imaging detector based on integrated Fabry-Perot microcavity array has the advantages of small size, high integration and strong environmental adaptability, and is expected to play a role in realizing the real-time detection of weak moving targets, auxiliary diagnosis of skin surface observation, and high dynamic range imaging of target observation under backlight conditions.&lt;/sec&gt;

Vascularized characteristics and functional regeneration of three-dimensional cell reconstruction of oral mucosa equivalents based on vascular homeostasis phenotypic modification

Our prior research has effectively developed tissue-engineered vascularized oral mucosa equivalents (VOME); however, challenges such as low repeatability and stability, as well as the inability to accurately replicate the complexity of real blood vessels, were encountered. Therefore, this study aimed to screen the VOME and native oral mucosa vascular homeostasis phenotypes by tandem mass tag-tagged proteomics associated with laser capture microdissection and human angiogenesis antibody array technology. Then, lentiviruses were constructed and stably transfected with vascular endothelial-like cells (VELCs) to detect angiogenic capacity. HE, EdU Apollo tracer staining, immunofluorescence staining, scanning electron microscopy, biomechanical testing, and a small animal ultrasound imaging system were used to analyze the characteristics of vascularization homeostasis and monitor functional regeneration of the vascularized homeostatic phenotypic oral mucosal equivalents (VHPOME). The results showed that PGAM1, COL5A1, ANG, and RNH1 are potential specific angiogenesis phenotypes. High expression of PGAM1, COL5A1, and ANG and/or low expression of RNH1 can promote the angiogenesis of VOME. ANG/shRNH1 has the most significant tube-like structure-formation ability. The expression of PGAM1, COL5A1, and ANG in the VHPOME group was higher than that of the control group, and the expression of RNH1 was lower than that of the control group. COL5A1/ANG can significantly improve the mechanical properties. The blood flow signal was most significant in the ANG/shRNH1 group. PGAM1, COL5A1, ANG, shRNH1, PGAM1/ANG, COL5A1/ANG, PGAM1/shRNH1, PGAM1/shRNH1, COL5A1/shRNH1, and ANG/shRNH1 may be the targets for establishing vascularization homeostasis and functional regeneration of oral mucosal equivalent genes (groups), and ANG/shRNH1 has the most significant effect.

Application of quantum dots in brain diseases and their neurotoxic mechanism.

The early-stage diagnosis and therapy of brain diseases pose a persistent challenge in the field of biomedicine. Quantum dots (QDs), nano-luminescent materials known for their small size and fluorescence imaging capabilities, present promising capabilities for diagnosing, monitoring, and treating brain diseases. Although some investigations about QDs have been conducted in clinical trials, the concerns about the toxicity of QDs have continued. In addition, the lack of effective toxicity evaluation methods and systems and the difference between in vivo and in vitro toxicity evaluation hinder QDs application. The primary objective of this paper is to introduce the neurotoxic effects and mechanisms attributable to QDs. First, we elucidate the utilization of QDs in brain disorders. Second, we sketch out three pathways through which QDs traverse into brain tissue. Ultimately, expound upon the adverse consequences of QDs on the brain and the mechanism of neurotoxicity in depth. Finally, we provide a comprehensive summary and outlook on the potential development of quantum dots in neurotoxicity and the difficulties to be overcome.

The protective effect of traditional Chinese medicine Jinteng Qingbi granules on rats with rheumatoid arthritis.

Introduction: Jinteng Qingbi granules (JTQBG), a traditional Chinese medicine formulation, are widely used for the treatment of rheumatoid arthritis (RA) due to their satisfactory therapeutic efficacy. However, the underlying mechanism of action remains unclear. This study aims to investigate the protective effects of JTQBG against RA and elucidates its potential molecular mechanisms. Methods: A collagen-induced arthritis (CIA) rat model was utilized, and JTQBG (1.25, 2.5, 5g/kg/day) or methotrexate (MTX, 1mg/kg/week) was orally administered. The rats' weight, arthritis index (AI), and paw volume were measured weekly. Synovial hyperplasia of the joints was detected using a small animal ultrasound imaging system. Joint destruction was assessed using an X-ray imaging system. Histopathological examinations were performed using hematoxylin-eosin (H&E), Saffron-O and fast green staining. Serum inflammatory cytokines were detected using ELISA. Furthermore, 4D label-free quantitative proteomics of synovial tissues and non-targeted metabolomics of blood serum were conducted to analyze the molecular mechanisms. Results: JTQBG exerted a significant therapeutic effect on CIA rats by reducing inflammatory cell infiltration, synovial hyperplasia, cartilage erosion, and bone destruction. It also decreased the spleen index, inhibited hyperplasia of the white pulp, and decreased the serum levels of IL-1β and IL-18. Proteomics analysis identified 367 differentially expressed proteins (DEPs) between the Model and Normal groups, and 71 DEPs between the JTQBG and Model groups. These DEPs were significantly enriched in the NF-κB pathway. 11 DEPs were significantly reversed after treatment with JTQBG. Western blot results further validated the expression levels of Nfkb1, Pdk1, and Pecam1, and analyzed the expression levels of p-IKK, p-IκBα, and IκBα. The therapeutic efficacy of JTQBG was partly attributed to the suppression of the NF-κB pathway in synovial tissues. Serum metabolomics identified 17 potential biomarkers for JTQBG treatment of CIA rats, which were closely related to Alanine, aspartate and glutamate metabolism, Tryptophan metabolism, Ascorbate and aldarate metabolism, Arginine metabolism, and Inositol phosphate metabolism. Conclusion: Our findings demonstrated that JTQBG was effective against RA by alleviating synovial inflammation, synovial hyperplasia, and joint destruction. The anti-RA properties of JTQBG were likely attributed to the inhibition of the NF-κB pathway and the regulation of serum metabolite disorders.

Deep learning-enhanced characterization of bubble dynamics in proton exchange membrane water electrolyzers.

The ever-increasing utility of imaging technology in proton exchange membrane water electrolyzer research raises the demand for rapid and precise image analysis. In particular, for optical video recordings, the challenge primarily lies in the large number of frames that impede the delineation of bubble dynamics with standard methods. In order to address this problem, the present study supports the automation of data analysis to facilitate swift, comprehensive, and measurable insights from captured imagery. We present a deep learning-based framework to perform high-throughput analyses of bubble dynamics using optical images of proton exchange membrane water electrolyzers. Leveraging a relatively small annotated imaging dataset of just 35 images, various configurations of the U-Net architecture were trained to perform bubble segmentation tasks. The best model achieved a precision of 95%, a recall of 78%, and an F1-score of 86% on the validation set. Subsequent to segmentation, the methodology enabled the rapid extraction of parameters such as time-resolved bubble area, size distributions, bubble position probability density, and individual bubble shape analytics. The findings underscore the potential of deep learning to enhance the analysis of polymer electrolyte membrane water electrolyzer imaging, offering a path toward more efficient and informative evaluations in electrochemical research.

Solving a classification problem using Transfer learning on small image datasets

In this study, we explore the efficiency of transfer learning in training convolutional neural networks (CNNs) for image classification tasks, focusing on a dataset with limited size. Utilizing the VGG16 model, we investigate the impact of varying the number of trainable layers on classification accuracy. Our specific classification challenge involves distinguishing between images of cars with open and closed hoods. The study demonstrates that employing transfer learning enables the pretrained VGG16 model to achieve over 97% accuracy in this binary classification task, even with a training set of just 1000 image samples. This research was conducted solely by the undersigned, who also represents the sole author of this paper.

Minimalist and High-Quality Panoramic Imaging With PSF-Aware Transformers.

High-quality panoramic images with a Field of View (FoV) of 360° are essential for contemporary panoramic computer vision tasks. However, conventional imaging systems come with sophisticated lens designs and heavy optical components. This disqualifies their usage in many mobile and wearable applications where thin and portable, minimalist imaging systems are desired. In this paper, we propose a Panoramic Computational Imaging Engine (PCIE) to achieve minimalist and high-quality panoramic imaging. With less than three spherical lenses, a Minimalist Panoramic Imaging Prototype (MPIP) is constructed based on the design of the Panoramic Annular Lens (PAL), but with low-quality imaging results due to aberrations and small image plane size. We propose two pipelines, i.e. Aberration Correction (AC) and Super-Resolution and Aberration Correction (SR&AC), to solve the image quality problems of MPIP, with imaging sensors of small and large pixel size, respectively. To leverage the prior information of the optical system, we propose a Point Spread Function (PSF) representation method to produce a PSF map as an additional modality. A PSF-aware Aberration-image Recovery Transformer (PART) is designed as a universal network for the two pipelines, in which the self-attention calculation and feature extraction are guided by the PSF map. We train PART on synthetic image pairs from simulation and put forward the PALHQ dataset to fill the gap of real-world high-quality PAL images for low-level vision. A comprehensive variety of experiments on synthetic and real-world benchmarks demonstrates the impressive imaging results of PCIE and the effectiveness of the PSF representation. We further deliver heuristic experimental findings for minimalist and high-quality panoramic imaging, in terms of the choices of prototype and pipeline, network architecture, training strategies, and dataset construction. Our dataset and code will be available at https://github.com/zju-jiangqi/PCIE-PART.

GHA-DenseNet prediction and diagnosis of malignancy in femoral bone tumors using magnetic resonance imaging

Background and objectiveDue to their aggressive nature and poor prognosis, malignant femoral bone tumors present considerable hurdles. Early treatment commencement is essential for enhancing vital and practical outcomes. In this investigation, deep learning algorithms will be used to analyze magnetic resonance imaging (MRI) data to identify bone tumors that are malignant. MethodologyThe study cohort included 44 patients, with ages ranging from 17 to 78 (22 women and 22 males). To categorize T1 and T2 weighted MRI data, this paper presents an improved DenseNet network model for the classification of bone tumor MRI, which is named GHA-DenseNet. Based on the original DenseNet model, the attention module is added to solve the problem that the deep convolutional model can reduce the loss of key features when capturing the location and content information of femoral bone tumor tissue due to the limitation of local receptive field. In addition, the sparse connection mode is used to prune the connection mode of the original model, so as to remove unnecessary and retain more useful fast connection mode, and alleviate the overfitting problem caused by small dataset size and image characteristics. In a clinical model designed to anticipate tumor malignancy, the utilization of T1 and T2 classifier output values, in combination with patient-specific clinical information, was a crucial component. ResultsThe T1 classifier's accuracy during the training phase was 92.88% whereas the T2 classifier's accuracy was 87.03%. Both classifiers demonstrated accuracy of 95.24% throughout the validation phase. During training and validation, the clinical model's accuracy was 82.17% and 81.51%, respectively. The clinical model's receiver operating characteristic (ROC) curve demonstrated its capacity to separate classes. ConclusionsThe proposed method does not require manual segmentation of MRI scans because it makes use of pretrained deep learning classifiers. These algorithms have the ability to predict tumor malignancy and shorten the diagnostic and therapeutic turnaround times. Although the procedure only needs a little amount of radiologists' involvement, more testing on a larger patient cohort is required to confirm its efficacy.