Manual Marking Research Articles

Enhancing pectus excavatum diagnosis with an automated batch evaluation tool for chest computed tomography images.

We aimed to implement a fully automatic computed tomography (CT) image-detection programming algorithm as a pectus excavatum (PE) diagnostic tool, facilitating comprehensive chest wall deformity evaluation. We developed our algorithm using MATLAB, leveraging the Hounsfield unit threshold and region growing methods. The MATLAB graphical user interface enables the direct use of our program. We validated the model using CT images of anthropomorphic phantoms and one normal individual. The measurement values obtained by our algorithm demonstrated very small differences compared to the known anthropomorphic phantom model data and manual measurement. For algorithm testing, 17,214 chest CT scans obtained from 57 PE patients were processed by the algorithm and independently reviewed by a radiologist and a thoracic surgeon. The measurements of transverse, anteroposterior, and sternum-to-vertebral distance of the thoracic cavity, along with the calculated data of four indices, exhibited high positive correlations (0.94-0.99). The asymmetry index and maximum anteroposterior hemithorax distance exhibited moderate correlation (0.40-0.83). Our automatic PE diagnostic tool demonstrated high accuracy; four chest wall deformity indices were obtained simultaneously without any initial manual marking, correlating well with manual measurements.

Development of an Algorithm for Semantic Segmentation of Earth Remote Sensing Data to Determine Phytoplankton Populations

Introduction. Computer vision is widely used for semantic segmentation of Earth remote sensing (ERS) data. The method allows monitoring ecosystems, including aquatic ones. Algorithms that maintain the quality of semantic segmentation of ERS images are in demand, specifically, to identify areas with phytoplankton, where water blooms— the cause of suffocation — are possible. The objective of the study is to create an algorithm that processes satellite data as input information for the formation and checking of mathematical models of hydrodynamics, which are used to monitor the state of water bodies. Various algorithms for semantic segmentation are described in the literature. New research focuses on enhancing the reliability of recognition — often using neural networks. This approach is modified in the presented work. To develop the direction, a new set of information from open sources and synthetic data are proposed. They are aimed at improving the generalization ability of the model. For the first time, the contour area of the phytoplankton population is compared to the database — and thus the boundary conditions are formed for the implementation of mathematical models and the construction of boundary-adaptive grids.Materials and Methods. The set of remote sensing images was supplemented with the author's augmentation algorithm in Python. Computer vision segmented areas of phytoplankton populations in the images. The U-Net convolutional neural network (CNN) was trained on the basis of NVIDIA Tesla T4 computing accelerators.Results. To automate the detection of phytoplankton distribution areas, a computer vision algorithm based on the U-Net CNN was developed. The model was evaluated by the calculated values of the main quality metrics related to segmentation tasks. The following metric values were obtained: Precision = 0.89, Recall = 0.88, F1 = 0.87, Dice = 0.87, and IoU = 0.79. Graphical visualization of the results of CNN learning on the training and validation sets showed good quality of model learning. This is evidenced by small changes in the loss function at the end of training. The segmentation performed by the model turned out to be close to manual marking, which indicated the high quality of the proposed solution. The area of the segmented region of the phytoplankton population was calculated by the area of one pixel. The result obtained for the original image was 51202.5 (based on information about the number of pixels related to the bloom of blue-green algae). The corresponding result of the modeling was 51312.Discussion and Conclusion. The study expands theoretical and practical knowledge on the use of convolutional neural networks for semantic segmentation of space imagery data. Given the results of the work, it is possible to assess the potential for automating the process of semantic segmentation of remote sensing data to determine the boundaries of phytoplankton populations using artificial intelligence. The use of the proposed computer vision model to obtain contours of water bloom due to phytoplankton will provide for the creation of databases — the basis for environmental monitoring of water resources and predictive modeling of hydrobiological processes.

Digital versus manual axis marking for toric phakic intraocular lens alignment: a prospective randomized intraindividual trial.

This study compares visual and refractive outcomes and postoperative axis alignment for toric Implantable Collamer Lens (ICL) implantation in astigmatic myopia using manual versus digital marking techniques. Department of Ophthalmology, Medical University of Graz, Austria. Prospective randomized single-centered intraindividual comparison. Patients undergoing bilateral toric ICL implantation for myopia with astigmatism ≥ 0.5 diopters (D) were enrolled. Patients received both marking techniques and randomization was performed. Postoperative retroillumination photography assessed axis alignment, and visual and refractive parameters were evaluated. Duration of the surgeries was recorded. The study includes 20 patients and 40 eyes. Preoperative visual and refractive parameters showed no significant differences. Postoperatively, residual astigmatism (p=0.824) and spherical equivalent (p=0.309) were comparable. No notable differences between right and left eyes in terms of preoperative (p=0.371) and postoperative (p=0.824) astigmatism were observed. Although slight, corrected distance visual acuity (CDVA) increased in both groups postoperatively (p=0.381). Gain in CDVA was comparable between marking techniques (p=0.637). Safety and efficacy indices were comparable (p=0.991 and p=0.189, respectively). Postoperative axial misalignment was 2.8±3.1 degrees in the digital- and 4.4±5.1 degrees in the manual group (p=0.260). Vector analysis showed no significant differences between manual and digital marking. Duration remained uninfluenced by the marking technique (p=0.970) and side of surgery (p=0.682). In conclusion digital and manual marking techniques provided comparable results in axis alignment, surgical duration and visual and refractive outcomes. Both methods are viable for precise axis alignment, with digital marking offering a potential advantage in efficacy.

Comparison of two digital alignment systems for toric intraocular lens implantation

PurposeTo compare the two most used digital alignment systems regarding precision, repeatability and loss of track.Methods15 eyes of 15 patients older than 21 years with cataracts were included in this...

Analysis of white matter hyperintensities in Alzheimer’s disease and vascular dementia with magnetic resonance imaging

Objectives: The study aimed to analyze the brain white matter hyperintensities (WMHs) of patients with vascular dementia (VaD) and Alzheimer’s disease (AD) using magnetic resonance imaging to determine whether white matter lesions in the brain could be detected by a computer using image processing methods. Patients and methods: In this retrospective observational study, a unimodal, unsupervised, and automatic method was developed, and magnetic resonance imaging of 35 patients were examined. Of the 35 patients, 19 (14 males, 5 females; mean age: 73.2±6.7 years; range, 56 to 83 years) were picked from patients with AD or VaD who were admitted to a neurology clinic between January 2016 and December 2022 (Group 1). The remaining 16 patients (10 females, 5 males; mean age: 80.4±5.4 years; range, 69 to 92 years) were included from the ABVIB (Aging Brain: Vasculature, Ischemia, and Behavior) study from the ADNI (Alzheimer’s Disease Neuroimaging Initiative) database (Group 2). To calculate the volume of WMHs, a detailed analysis was conducted. Initially, skull stripping was performed, and then the brain was segmented. Afterward, two types of masks (Mask1 and Mask-2) were obtained by applying painting, decreasing, and blurring processes to the segmented white matter. These masks limited the region that was searched for WMHs. With this limitation, false positives that could arise from gray matter intensities were tried to be prevented. To evaluate the accuracy of WMH detection, a user interface was developed, and manual marking was conducted by an expert neurologist. After WMH detection, WMH volumes were calculated. Results: For Group 1, the similarity index was found to be 0.76 for Mask-1 and 0.80 for Mask-2, while for Group 2, the similarity index was 0.71 for Mask-1 and 0.87 for Mask-2. In patients with AD, the mean WMH lesion load (LL) was 15.16±16.59 mL. In patients with VaD, who were expected to suffer more from WMHs, the mean WMH LL was 29.22±11.40 mL. In Group 2, the mean WMH LL was 17.77±12.26 mL. Conclusion: This study may contribute to the literature since it is an automatic, unimodal, and unsupervised method that was applied to both a completely unique data set with many different scanning parameters and an open database data set.

A Mumford-Shah Model-based Method for Inpainting Markers in Ultrasound Images

Abstract During ultrasound diagnosis in clinics, clinicians often mark the lesion area, resulting in the presence of unnecessary objects in the collected ultrasound images. These markers potentially affect subsequent image analysis. To address this problem, we proposed an image inpainting method that combines image processing with Mumford-Shah algorithm to remove markers from ultrasound images. The proposed method consists of two parts. First, the input ultrasound images were processed by several image processing algorithms, including contrast enhancement, image binarization, edge detection, hole filling, and connected domain marking. These algorithms were combined to extract the mask of the manual marker in the ultrasound images. Second, the original ultrasound images and mask images were put into the Mumford-Shah model to implement ultrasound image inpainting by minimizing the following energy functional. The continuous iteration of Euler-Lagrange equation was employed to update the iteration item. If the difference between the two output results was less than the pre-set threshold, the output result could be regarded as the repaired ultrasound image. Experimental results show that the designed image processing method accurately extracted the binary mask, whether they are colourful or binary, number or line in ultrasound images. Meanwhile, the improved Mumford-Shah model effectively implemented the ultrasound image inpainting, with the accuracy of 96.55%. Therefore, the proposed method can be used to remove unnecessary objects in ultrasound images, which is beneficial to improve the accuracy of the artificial intelligence model.

Outcomes of astigmatic correction with and without two different cyclotorsion compensation methods in small incision lenticule extraction surgery

PurposeTo compare the astigmatic correction outcomes of small incision lenticule extraction (SMILE) surgery with or without two different cyclotorsion compensation methods. MethodsThis is a prospective randomized clinical trial. Patients with myopic astigmatism that underwent SMILE surgery were randomly divided into static cyclotorsion compensated group (SCC group), slit-lamp group and control group. In the SCC and slit-lamp groups, the intraoperative cyclotorsion was manually compensated with different limbal marking methods. In the control group, the cyclotorsion was not compensated. Visual acuity and manifest refraction were measured preoperatively and postoperatively. Astigmatic outcomes were estimated with vector analysis. ResultsA total of 94 eyes from 94 patients were analyzed postoperatively at the 3-month follow-up. Their mean preoperative cylinder was -1.56±0.86 D (range: -4.25 to -0.25 D). The mean preoperative spherical equivalent was -5.95±1.72 D (range: -10.50 to -2.75 D). All groups showed favorable results in the correction of myopic astigmatism. No statistically differences were found among three groups in postoperative visual acuity, refractive outcomes or vector parameters. ConclusionCyclotorsion compensation with two different manual limbal marking methods was helpful in aligning the surgical position in SMILE, but it was not as effective as expected for the correction of myopic astigmatism under well controlled surgical positioning.

Optical coherence tomography homography for detection of retinal displacement: a validation study.

Retinal displacement following rhegmatogenous retinal detachment (RRD) has been associated with inferior functional outcomes. Recent evidence using an overlay technique suggests that fundus-autofluorescence underestimates post-RRD repair retinal displacement. This study aims to validate the overlay technique in normal eyes and to determine its sensitivity and specificity at detecting retinal displacement. We conducted a retrospective case series involving 66 normal eyes, each with at least two separate infrared (IR) images at different time points. Overlay of the two images was based on manual marking of choroidal and optic nerve head (ONH) landmarks. For each set of two IR images, computer code for homography generated two outputs, flipping view video and an overlay picture. First, validation of choroidal/ONH alignment was performed using the flipping view video to ensure accurate manual markings. Then, two different masked graders (AB + IM) evaluated the overlays for presence of retinal displacement. 16 control eyes following RRD repair with detected retinal displacement on FAF imaging assessed sensitivity and specificity of the technique. 94% of overlays were found to be well aligned (62/66). 11 cases exhibited errors on flipping view analysis (choroidal/ONH misalignment). Those 11 cases had a significantly higher rate of retinal displacement (false positives) compared to cases without errors (8/11,72% Vs 54/55,98%,P = 0.001). Sensitivity and specificity of the overlay technique for detecting retinal displacement considering only adequate flipping view cases (n = 55) were calculated as 100% and 98%, respectively. IR overlay emerges as a reliable and valid method for detecting retinal displacement, exhibiting excellent sensitivity and specificity.

Computer Aided Detection of Cancerous Areas through Steady Images of E.G.D(ESOPHAGOGASTRODUODENOSCOPY)

Abstract: Stomach cancer, or gastric cancer, remains a significant global health concern due to its high mortality rate and prevalence. Despite advancements in medical technology and treatment, it continues to rank as the fourth most common cancer worldwide and the second leading cause of cancer-related deaths. However, there has been a decline in both incidence and mortality rates over the past decades, attributed to improved awareness, screening programs, and medical interventions. Early detection is crucial, as stomach cancer often begins with symptomless lesions that can progress over time. Various imaging modalities are utilized for detection, with endoscopy emerging as the preferred method due to its ability to provide highresolution images and perform tissue sampling for accurate diagnosis. Recognizing the need for enhanced diagnostic support, a computer-aided analysis system has been developed, leveraging algorithms to enhance endoscopic images and detect cancerous areas more effectively. This system, employing MATLAB algorithms, enhances image quality and removes reflected flash spots, facilitating better visibility of lesions. The process involves manual marking of cancerous areas within images, dataset annotation using RoboFlow, and training the model using Google Collab for pattern recognition of cancerous lesions. Rigorous testing demonstrates high accuracy in real-time detection, promising significant advancements in early diagnosis and improved patient outcomes. This project represents a noteworthy stride in leveraging technology to combat stomach cancer, offering potential for further enhancements and broader clinical utility in the future.

Contrastive Prediction and Estimation of Deformable Objects based on Improved Resnet

Because the dynamic model of deformable linear object is complex, the learning based on visual model is difficult, and the feature information extraction is insufficient. Therefore, we propose a joint visual representation model using contrast learning of optimized encoder. We start with the encoder, add the residual structure to the encoder, optimize the extraction and compression of its feature information, and control its parameters to 3 million. In this way, we can not only obtain excellent feature information, but also have good efficiency. In the rope experiment, we collect information from the simulated environment without manual marking, extract features through the encoder and transmit them to the downstream task. Experiments show that the evaluation of our model at 135 ° and 45 ° is improved by about 50%.

Validation of Tracheal Sound-Based Respiratory Effort Monitoring for Obstructive Sleep Apnoea Diagnosis.

Background: Respiratory effort is considered important in the context of the diagnosis of obstructive sleep apnoea (OSA), as well as other sleep disorders. However, current monitoring techniques can be obtrusive and interfere with a patient's natural sleep. This study examines the reliability of an unobtrusive tracheal sound-based approach to monitor respiratory effort in the context of OSA, using manually marked respiratory inductance plethysmography (RIP) signals as a gold standard for validation. Methods: In total, 150 patients were trained on the use of type III cardiorespiratory polygraphy, which they took to use at home, alongside a neck-worn AcuPebble system. The respiratory effort channels obtained from the tracheal sound recordings were compared to the effort measured by the RIP bands during automatic and manual marking experiments. A total of 133 central apnoeas, 218 obstructive apnoeas, 263 obstructive hypopneas, and 270 normal breathing randomly selected segments were shuffled and blindly marked by a Registered Polysomnographic Technologist (RPSGT) in both types of channels. The RIP signals had previously also been independently marked by another expert clinician in the context of diagnosing those patients, and without access to the effort channel of AcuPebble. The classification achieved with the acoustically obtained effort was assessed with statistical metrics and the average amplitude distributions per respiratory event type for each of the different channels were also studied to assess the overlap between event types. Results: The performance of the acoustic effort channel was evaluated for the events where both scorers were in agreement in the marking of the gold standard reference channel, showing an average sensitivity of 90.5%, a specificity of 98.6%, and an accuracy of 96.8% against the reference standard with blind expert marking. In addition, a comparison using the Embla Remlogic 4.0 automatic software of the reference standard for classification, as opposed to the expert marking, showed that the acoustic channels outperformed the RIP channels (acoustic sensitivity: 71.9%; acoustic specificity: 97.2%; RIP sensitivity: 70.1%; RIP specificity: 76.1%). The amplitude trends across different event types also showed that the acoustic channels exhibited a better differentiation between the amplitude distributions of different event types, which can help when doing manual interpretation. Conclusions: The results prove that the acoustically obtained effort channel extracted using AcuPebble is an accurate, reliable, and more patient-friendly alternative to RIP in the context of OSA.

A Web Knowledge-Driven Multimodal Retrieval Method in Computational Social Systems: Unsupervised and Robust Graph Convolutional Hashing

Multimodal retrieval has received widespread consideration since it can commendably provide massive related data support for the development of computational social systems (CSSs). However, the existing works still face the following challenges: 1) rely on the tedious manual marking process when extended to CSS, which not only introduces subjective errors but also consumes abundant time and labor costs; 2) only using strongly aligned data for training, lacks concern for the adjacency information, which makes the poor robustness and semantic heterogeneity gap difficult to be effectively fit; and 3) mapping features into real-valued forms, which leads to the characteristics of high storage and low retrieval efficiency. To address these issues in turn, we have designed a multimodal retrieval framework based on web-knowledge-driven, called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">unsupervised and robust graph convolutional hashing</i> (URGCH). The specific implementations are as follows: first, a “ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">secondary semantic self-fusion</i> ” approach is proposed, which mainly extracts semantic-rich features through pretrained neural networks, constructs the joint semantic matrix through semantic fusion, and eliminates the process of manual marking; second, a “ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">adaptive computing</i> ” approach is designed to construct enhanced semantic graph features through the knowledge-infused of neighborhoods and uses graph convolutional networks for knowledge fusion coding, which enables URGCH to sufficiently fit the semantic modality gap while obtaining satisfactory robustness features; Third, combined with hash learning, the multimodality data are mapped into the form of binary code, which reduces storage requirements and improves retrieval efficiency. Eventually, we perform plentiful experiments on the web dataset. The results evidence that URGCH exceeds other baselines about <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1\%$</tex-math> </inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$3.7\%$</tex-math> </inline-formula> in mean average precisions (MAPs), displays superior performance in all the aspects, and can meaningfully provide multimodal data retrieval services to CSS.

Impact of advanced image-guided systems on correction of corneal astigmatism during FLACS.

To determine the effectiveness of the correction of corneal astigmatism using the Alcon Image Guidance (IG) system vs manual marking in the orientation of femtosecond laser-assisted cataract surgery (FLACS). Private practice, Albany, New York. Retrospective review. Patients who underwent FLACS with IG and without IG were from January 2018 to June 2022 investigated. Variables including preoperative keratometry values, cylinder, spherical equivalent (SE), and visual acuity (VA) were collected, as well as the cyclorotation angle delta registered by IG, postoperative refractive cylinder, SE, and VA. The primary outcome was postoperative refractive cylinder in patients with IG compared with those without IG. A total of 160 eyes were included, with 103 eyes that had IG and 57 eyes that did not have IG. Postoperative cylinder was similar in those with IG (0.31 ± 0.36 diopters [D]) compared with those without IG (0.31 ± 0.37 D) ( P = .97). Average cyclorotation in the IG group was 2.82 ± 3.03 degrees. When cyclorotation was stratified into 3 groups (<2.8 degrees, 2.8 to 8.5 degrees, >8.5 degrees), no differences were found in postoperative refractive cylinder ( P = .35). Patients who underwent FLACS with IG had similar postoperative cylinder outcomes compared with those without IG. This study suggests that the accommodation of cyclotorsion using an advanced IG system is similar to that obtained with manual marking techniques in patients having 2 D or less of astigmatism corrected.

Synthetic data augmentation for high-resolution X-ray welding defect detection and classification based on a small number of real samples

Deep learning has become the dominant technology in most computer vision tasks. These methods often rely on a large number of labeled sample datasets for training, and in the field of non-destructive testing of welds in industrial manufacturing, weld images with defects are very scarce, and it is still a challenging challenge to construct high-resolution weld defect datasets that meet the requirements. To overcome this limitation, a new data augmentation method for high-resolution X-ray welding defect classification and synthesis based on a small number of real samples is proposed to realize the data augmentation of industrial nondestructive inspection X-ray film defect images. Firstly, to overcome the scarcity of the weld X-ray defect classification dataset, the weld X-ray defect classification dataset (Weld Defect Classification, WDC) is constructed. Secondly, the performance of 16 common deep classification models on WDC datasets is explored. Then, the images of the real local welding defects and the non-defective weld area are fused at random locations, and two data augmentation modes, (Single Image Single Defect, SISD) and (Single Image Multi Defects, SIMD), can generate defect files and annotation files (Visual Object Classes, VOC) at the same time, which can save a lot of time for manual marking. Finally, compared with the traditional data augmentation method, the proposed method can effectively improve the accuracy of defect detection and generalization, the mAP (Mean Average Precision, mAP) @0.5 of YOLOV8X (You Only Look Once, YOLO) and YOLOV5.6.1X is 66.6% and 72.8%, which provides an effective solution for data sample generation in the industrial field.

A Decoder Structure Guided CNN‐Transformer Network for face super‐resolution

AbstractRecent advances in deep convolutional neural networks have shown improved performance in face super‐resolution through joint training with other tasks such as face analysis and landmark prediction. However, these methods have certain limitations. One major limitation is the requirement for manual marking information on the dataset for multi‐task joint learning. This additional marking process increases the computational cost of the network model. Additionally, since prior information is often estimated from low‐quality faces, the obtained guidance information tends to be inaccurate. To address these challenges, a novel Decoder Structure Guided CNN‐Transformer Network (DCTNet) is introduced, which utilises the newly proposed Global‐Local Feature Extraction Unit (GLFEU) for effective embedding. Specifically, the proposed GLFEU is composed of an attention branch and a Transformer branch, to simultaneously restore global facial structure and local texture details. Additionally, a Multi‐Stage Feature Fusion Module is incorporated to fuse features from different network stages, further improving the quality of the restored face images. Compared with previous methods, DCTNet improves Peak Signal‐to‐Noise Ratio by 0.23 and 0.19 dB on the CelebA and Helen datasets, respectively. Experimental results demonstrate that the designed DCTNet offers a simple yet powerful solution to recover detailed facial structures from low‐quality images.

A Client-Cloud-Chain Data Annotation System of Internet of Things for Semi-Supervised Missing Data

With continuous progress in science and technology, a large amount of data are produced in all fields of the world at anytime and anywhere. These data are unmarked and lack marking information, while manual marking is time-consuming and laborious. Herein, this paper introduces a distributed semi-supervised labeling framework. This framework addresses the issue of missing data by proposing an attribute-filling method based on subspace learning. Furthermore, this paper presents a distributed semi-supervised learning strategy that trains sub-models (private models) within each sub-system. Finally, this paper develops a distributed graph convolutional neural network fusion technique with enhanced interpretability grounded on the attention mechanism. This paper assigns weights of importance to the edges of each layer in the graph neural network based on sub-models and public data, thereby enabling distributed and interpretable graph convolutional attention. Extensive experimentation using public datasets demonstrates the superiority of the proposed scheme over other state-of-the-art baselines, achieving a reduction in loss of 50% compared to the original approach.

An overview of algorithms for discovering new categories of problems based on unlabeled data

In traditional machine learning, supervised and semi-supervised learning is designed to be used in a closed-world setting where the training data is fixed and does not change over time. Unfortunately, these methods still require a large number of labels for the categories to be categorized, which is expensive and impractical. A new category discovery algorithm is designed so that it can discover new categories while classifying and recognizing labeled images. In this case, the machine can automatically identify new categories without manual marking of image feature categories, which can greatly reduce the cost of image classification. Kai Han et al. named this problem a new category discovery problem and proposed that deep clustering can be used to solve it well. This paper focuses on the comparison of two commonly used robust baselines in the new category discovery and proposes that adding a post-processing model can better improve the accuracy of the model result. This paper applied the relaxed contrast learning method to the Ranking Statistics, and the accuracy of CIFAR-100 is improved by 6%.

Real-time portrait image retouching extended from DualBLN

Towards the target of image retouching tasks, the three-dimension LookUp Table (3D LUT) is a discrete 3D lattice that represents the color mapping to correct the inconsistent color of images produced from varying devices and environments. We design a network named Dual Branch LUT-aware Network (DualBLN), which innovatively introduces the information representing the pixel value mapping in 3D LUT into the real-time retouching procedure, learning the adaptive weights and the multiple 3D LUTs with strong representation capability. Specifically, we elaborate two branches to conduct feature extraction of the raw image and 3D LUTs, which consider the detail of the dual inputs, then generate the precise weights to fuse LUTs for graceful retouching. To integrate the information of the image and the 3D LUTs meticulously, we adopt bilinear pooling to alleviate the conflict between both features when fusing them from two branches of the network, which avoids the feature distortion caused by concatenation or summation directly. In addition, to meet the needs of portrait photo retouching (PPR) for portrait area priority (HRP), we propose Human Region Mask Generator (HRMG) module to solve the shortcomings of the current PPR task that requires the manual marking of portrait region, which achieve the PPR task more conveniently and efficiently. Extensive experiments verify the effectiveness of our method, e.g., the PSNR metric receives over 0.6 gains on high-resolution and 1.4 gains on the human-centered region at most compared with prior state-of-art works.

Development of a deep learning model for early gastric cancer diagnosis using preoperative computed tomography images.

Gastric cancer is a highly prevalent and fatal disease. Accurate differentiation between early gastric cancer (EGC) and advanced gastric cancer (AGC) is essential for personalized treatment. Currently, the diagnostic accuracy of computerized tomography (CT) for gastric cancer staging is insufficient to meet clinical requirements. Many studies rely on manual marking of lesion areas, which is not suitable for clinical diagnosis. In this study, we retrospectively collected data from 341 patients with gastric cancer at the First Affiliated Hospital of Wenzhou Medical University. The dataset was randomly divided into a training set (n=273) and a validation set (n=68) using an 8:2 ratio. We developed a two-stage deep learning model that enables fully automated EGC screening based on CT images. In the first stage, an unsupervised domain adaptive segmentation model was employed to automatically segment the stomach on unlabeled portal phase CT images. Subsequently, based on the results of the stomach segmentation model, the image was cropped out of the stomach area and scaled to a uniform size, and then the EGC and AGC classification models were built based on these images. The segmentation accuracy of the model was evaluated using the dice index, while the classification performance was assessed using metrics such as the area under the curve (AUC) of the receiver operating characteristic (ROC), accuracy, sensitivity, specificity, and F1 score. The segmentation model achieved an average dice accuracy of 0.94 on the hand-segmented validation set. On the training set, the EGC screening model demonstrated an AUC, accuracy, sensitivity, specificity, and F1 score of 0.98, 0.93, 0.92, 0.92, and 0.93, respectively. On the validation set, these metrics were 0.96, 0.92, 0.90, 0.89, and 0.93, respectively. After three rounds of data regrouping, the model consistently achieved an AUC above 0.9 on both the validation set and the validation set. The results of this study demonstrate that the proposed method can effectively screen for EGC in portal venous CT images. Furthermore, the model exhibits stability and holds promise for future clinical applications.

Extraction and classification of channels from LiDAR in plains by channel tracking

Accurate and detailed maps of irrigation and drainage networks in plains are imperative for efficient water management. Conducting field surveys of these networks proves to be expensive and labor-intensive. Besides, manual extraction of this information from remote sensing data lacks the necessary speed and accuracy. While LiDAR data offers highly detailed insights into terrain morphology, extracting water channels from it is not a straightforward task. The commonly employed slope-based algorithms, primarily reliant on elevation gradients, falter in plains due to minimal height variations such as local depressions or sinks, which can lead to misinterpretations. The primary goal of this study is to automate the extraction of the network using high-resolution Digital Terrain Models (DTMs) derived from LiDAR. Though the number of parameters is limited, users have the flexibility to finely adjust hydrologically relevant and intuitive parameters based on their specific channel preferences. The algorithm commences with a curvature analysis of the surface. Random seed points are strategically chosen from locations where the local curvature is categorized as ‘valley.’ A cross-section of the DTM is then extracted aligned with the direction of maximum curvature. The section's morphology is thoroughly evaluated to determine its validity as a water channel section. If deemed valid, points on both sides of the thalweg point along the zero-curvature direction undergo testing. The channel is traced in both directions, as long as the cross-sections maintain their validity. Upon the cessation of tracking, new hypotheses are examined until all ‘valley’ type points have been exhausted. The study demonstrates that the algorithm can also calculate channel capacity or perform channel classification using the acquired channel thalweg points and cross-sections. To validate the algorithm's effectiveness, tests are conducted across three distinct areas within the Bergama Plain in Turkey. Manual markings establish the ground truth for the channels. The precision and recall values achieve 100%, with the exception of Test Area 2 (77% precision). Additionally, the outcomes surpass those of flow-based algorithm implemented in the same areas. The proposed method proves its utility in plains and remains open to further refinement and development.