Imaging Of Lesions Research Articles

Comparing the Accuracy and Reliability of Detecting the Intensity of Spinal Inflammation on STIR Sequence with Apparent Diffusion Coefficient Values in Axial Spondyloarthritis

Objective: To compare the accuracy and reliability of detecting the intensity of spinal inflammation on short tau inversion recovery (STIR) with the apparent diffusion coefficient (ADC) values of the active magnetic resonance imaging (MRI) lesions in axial spondyloarthritis (axSpA). Materials and methods: Fifty active lesions in the STIR sequence of spinal MRI were identified. With reference to sites of active lesions in STIR, the corresponding region of interest (ROI) on the ADC map was drawn to determine the maximum ADC (ADC[Formula: see text]), mean ADC (ADC[Formula: see text]), normalized maximum (nADC[Formula: see text]), and mean (nADC[Formula: see text]). Four independent readers scored the identified active lesions as “intense” or “non-intense” according to the Spondyloarthritis Research Consortium of Canada (SPARCC) MRI index. They were compared to various ADC parameters for assessment of accuracy and reliability. Regression analyses were used to adjust potential factors that could affect ADC. Results: Significant differences were found in ADC[Formula: see text] between “intense” and “non-intense” lesions scored by three of the four readers (1,405.7 ± 271.4 vs. 1,165.8 ± 223.8, [Formula: see text] = 0.01; 1,420.7 ± 272.1 vs. 1,209.0 ± 248.5, [Formula: see text] = 0.01; 1,438.0 ± 307.2 vs. 1,213.6 ± 231.0, [Formula: see text] = 0.01). Only one reader could differentiate a difference in “intense” and “non-intense” lesions with respect to ADC[Formula: see text] (899.2 ± 248.3 vs. 711.0 ± 222.6, [Formula: see text] = 0.01) and nADC[Formula: see text] (4.4 ± 2.1 vs. 3.4 ± 1.4, [Formula: see text] = 0.05). Inter-reader agreements were slight to moderate (kappa = 0.07–0.45). Reliability substantially improved when only the lowest and highest 25th percentiles of ADC values were included (kappa = 0.17–0.75). Regression analyses showed that the “intense” lesions were associated with higher ADC values after adjustment for confounders. Conclusion: Reading of STIR MRI is limited by the lack of ability in differentiating subtle differences of spinal inflammation. ADC could be an alternative method.

Melanoma Skin Cancer Prediction Using Machine Learning Algorithm

The academic paper titled "Melanoma Skin Cancer Prediction Using Machine Learning Algorithms" provides a detailed analysis of the application of machine learning techniques in the timely detection and categorization of melanoma skin cancer. The research aims to enhance the accuracy and reliability of predicting the distinction between malignant melanomas and benign lesions by utilizing a diverse set of skin lesion images. The study examines various machine learning models, including Logistic Regression, Support Vector Machines (SVM), K-Nearest Neighbor (KNN), Decision Trees, Gaussian Naïve Bayes, and Ensemble Techniques, as well as advanced deep learning approaches like Convolutional Neural Networks (CNNs). Emphasizing the potential, the paper underscores the benefits of integrating advanced machine learning methods in clinical settings to improve the timely detection of melanoma, thus ultimately boosting patient outcomes and treatment efficacy.

Deep semantic segmentation and skin disease classification fromdermoscopic images based on a modernized deep learning network with multi-feature extraction

ABSTRACT In this research, an advanced deep learning model is proposed to segment and classify skin diseases. The skin images required for classification purposes are collected from public databases. Skin lesion segmentation is initially performed over collected images using Multi-head Attention-based MobileUNet, which helps locate lesions in dermoscopic images. The segmented images are fed to the feature extraction stage to get more representative features from the images to improve the classification performance. Features like texture, shape, and colour are extracted in this stage. The extracted features are given to the Adaptive Deep Capsule Network to classify the images. The segmented images are directly given to the Adaptive MobileNet to provide classification results. The final decision is made by averaging the classified outcome using the Adaptive Hybrid Network. The performance of classification is further enhanced by tuning the parameters from DCNet and MNet using the Enhanced Secretary Bird Optimisation. The performance validation finally shows the accurate performance in the skin disease classification model.

Skin Disease Detection Using Ensemble Learning

Skin diseases affect millions of people worldwide, making early detection and accurate diagnosis crucial for effective treatment. However, the process of diagnosing skin conditions often requires specialized dermatological expertise, which can be time-consuming and expensive. To address this challenge, we have developed a web-based Skin Disease Detection System using cutting-edge machine learning and ensemble techniques. Our system leverages the HAM10000 dataset, consisting of 10,000 dermoscopic images representing seven common skin lesion types. We employ an ensemble learning approach, integrating ResNet50, EfficientNetB0, and DenseNet121 as base models and combining their predictions using a logistic regression meta-model. This architecture enhances the accuracy and robustness of the predictions. The backend of our application is built with Flask, responsible for model processing, authentication, and handling API requests. The React frontend provides an intuitive user interface where users can upload skin lesion images and receive real-time predictions along with confidence scores. The system also incorporates authentication and token verification for secure user access, ensuring data privacy and integrity. Our project aims to provide an accessible, reliable, and cost-effective tool to assist healthcare professionals and individuals in the early detection of skin diseases, potentially improving patient outcomes and reducing the burden on dermatology services

Focal liver lesions: multiparametric microvasculature characterization via super-resolution ultrasound imaging

BackgroundNoninvasive and functional imaging of the focal liver lesion (FLL) vasculature at microscopic scales is clinically challenging. We investigated the feasibility of using super-resolution ultrasound (SR-US) imaging for visualizing and quantifying the microvasculature of intraparenchymal FLLs.MethodsPatients with FLLs between June 2022 and February 2023 were prospectively screened. Following bolus injection of microbubbles at clinical concentration, SR-US was performed using a high frame rate (350–500 Hz) modified ultrasound scanner and a convex array transducer with a central frequency of 3.1 MHz.ResultsIn total, 47 pathologically proven FLLs at a depth of 5.7 ± 1.7 cm (mean ± standard deviation) were included: 30 hepatocellular carcinomas (HCC), 11 liver metastases (LM), and 6 focal nodular hyperplasias (FNH). The smallest detectable vessel size of the hepatic microvasculature was 128.4 ± 18.6 μm (mean ± standard deviation) at a depth of 8 cm. Significant differences were observed among the three types of lesions in terms of pattern categories, vessel density, minimum flow velocity, and perfusion index. We observed higher vessel density for FNH versus liver parenchyma (p < 0.001) as well as fractal dimension and local flow direction entropy value for FNH versus HCC (p = 0.002 and p < 0.001, respectively) and for FNH versus LM (p = 0.006 and p = 0.002, respectively).ConclusionMultiparametric SR-US showed that these three pathological types of FLLs have specific microvascular phenotypes. Vessel density, fractal dimension and local flow direction entropy served as valuable parameters in distinguishing between benign and malignant FLLs.Trial registrationClinicalTrials.gov (NCT06018142).Relevance statementMultiparametric SR-US imaging offers precise morphological and functional assessment of the microvasculature of intraparenchymal focal liver lesions, providing insights into tumor heterogeneity and angiogenesis.Key PointsSuper-resolution (SR)-US imaging allowed morphological and functional evaluation of intraparenchymal hepatic lesion microvasculature.Hepatocellular carcinoma, liver metastasis, and focal nodular hyperplasia exhibit distinct microvascular architectures and hemodynamic profiles.Multiparametric microvasculature characterization via SR-US imaging facilitates the differentiation between benign and malignant microvascular phenotypes.Graphical

An XAI Approach to Melanoma Diagnosis: Explaining the Output of Convolutional Neural Networks with Feature Injection

Computer-aided diagnosis (CAD) systems, which combine medical image processing with artificial intelligence (AI) to support experts in diagnosing various diseases, emerged from the need to solve some of the problems associated with medical diagnosis, such as long timelines and operator-related variability. The most explored medical application is cancer detection, for which several CAD systems have been proposed. Among them, deep neural network (DNN)-based systems for skin cancer diagnosis have demonstrated comparable or superior performance to that of experienced dermatologists. However, the lack of transparency in the decision-making process of such approaches makes them “black boxes” and, therefore, not directly incorporable into clinical practice. Trying to explain and interpret the reasons for DNNs’ decisions can be performed by the emerging explainable AI (XAI) techniques. XAI has been successfully applied to DNNs for skin lesion image classification but never when additional information is incorporated during network training. This field is still unexplored; thus, in this paper, we aim to provide a method to explain, qualitatively and quantitatively, a convolutional neural network model with feature injection for melanoma diagnosis. The gradient-weighted class activation mapping and layer-wise relevance propagation methods were used to generate heat maps, highlighting the image regions and pixels that contributed most to the final prediction. In contrast, the Shapley additive explanations method was used to perform a feature importance analysis on the additional handcrafted information. To successfully integrate DNNs into the clinical and diagnostic workflow, ensuring their maximum reliability and transparency in whatever variant they are used is necessary.

Scatter correction for contrast-enhanced digital breast tomosynthesis with a dual-layer detector.

Contrast-enhanced digital breast tomosynthesis (CEDBT) highlights breast tumors with neo-angiogenesis. A recently proposed CEDBT system with a dual-layer (DL) flat-panel detector enables simultaneous acquisition of high-energy (HE) and low-energy (LE) projection images with a single exposure, which reduces acquisition time and eliminates motion artifacts. However, x-ray scatter degrades image quality and lesion detectability. We propose a practical method for accurate and robust scatter correction (SC) for DL-CEDBT. The proposed hybrid SC method combines the advantages of a two-kernel iterative convolution method and an empirical interpolation strategy, which accounts for the reduced scatter from the peripheral breast region due to thickness roll-off and the scatter contribution from the region outside the breast. Scatter point spread functions were generated using Monte Carlo simulations with different breast glandular fractions, compressed thicknesses, and projection angles. Projection images and ground truth scatter maps of anthropomorphic digital breast phantoms were simulated to evaluate the performance of the proposed SC method and three other kernel- and interpolation-based methods. The mean absolute relative error (MARE) between scatter estimates and ground truth was used as the metric for SC accuracy. DL-CEDBT shows scatter characteristics different from dual-shot, primarily due to the two energy peaks of the incident spectrum and the structure of the DL detector. Compared with the other methods investigated, the proposed hybrid SC method showed superior accuracy and robustness, with MARE of for all LE and HE projection images of different phantoms in both cranial-caudal and mediolateral-oblique views. After SC, cupping artifacts in the dual-energy image were removed, and the signal difference-to-noise ratio was improved by 82.0% for 8mm iodine objects. A practical SC method was developed, which provided accurate and robust scatter estimates to improve image quality and lesion detectability for DL-CEDBT.

Evaluation of the flying focal spot technology in a wide-angle digital breast tomosynthesis system.

We characterize the flying focal spot (FFS) technology in digital breast tomosynthesis (DBT), designed to overcome source motion blurring. A wide-angle DBT system with continuous gantry and focus motion ("uncompensated focus") and a system with FFS were compared for image sharpness and lesion detectability. The modulation transfer function (MTF) was assessed as a function of height in the projections and reconstructed images, along with lesion detectability using the contrast detail phantom for mammography (CDMAM) and the L1 phantom. For the uncompensated focus system, the spatial frequency for 25% MTF value ( ) measured at 2, 4, and 6cm in DBT projections fell by 35%, 49%, and 59%, respectively in the tube-travel direction compared with the FFS system. There was no significant difference in for the front-back and tube-travel directions for the FFS unit. The in-plane MTF in the tube-travel direction also improved with the FFS technology.The threshold gold thickness ( ) for the 0.16-mm diameter discs of contrast detail phantom for mammography (CDMAM) improved for the FFS system in DBT mode, especially at greater heights above the table; at 45 and 65mm improved by 16% and 24%, respectively, compared with the uncompensated focus system. In addition, improvements in calcification and mass detection in a structured background were observed for DBT and synthetic mammography. The FFS system demonstrated faster scan times (4.8s versus 21.7s), potentially reducing patient motion artifacts. The FFS technology offers isotropic resolution, improved small detail detectability, and faster scan times in DBT mode compared with the traditional continuous gantry and focus motion approach.

In vivo study to evaluate an intelligent algorithm for time efficient detection of malignant melanoma using dermatofluoroscopy.

Dermatofluoroscopy is an optical non-invasive method of melanoma/nevus differentiation that has shown 89% sensitivity and 45% specificity in clinical trials, but long measurement duration hinders clinical use. An intelligent algorithm was developed to shorten the measurement time without compromising its diagnostic accuracy. It uses dermoscopic images of the skin lesions to be measured to select measurement points based on the assessment of color values. 27 patients with a total of 29 lesions suggestive of cutaneous melanoma were included in a clinical study and measured with both methods, conventional dermatofluoroscopy and the newly developed intelligent algorithm. The results were compared to the independent findings of two histopathologists to evaluate diagnostic accuracy and time saved. There was a median reduction of measurement points from 265 to 158 (40%). Meanwhile, the intelligent algorithm showed a higher diagnostic accuracy than conventional dermatofluoroscopy (AUC of 72% vs. 63%). The intelligent algorithm did not perform inferior to the conventional method while saving 40% of time. However, measurement times remain long compared to other non-invasive methods of diagnosing malignant melanoma. Further studies are needed to evaluate clinical suitability.

Predictive model for converting optic neuritis to multiple sclerosis; decision tree in focus.

Optic neuritis (ON) can be an initial clinical presentation of multiple sclerosis This study aims to provide a practical predictive model for identifying at-risk ON patients in developing MS. We utilized data from the Optic Neuritis Treatment Trial study, which enrolled 457 patients aged from 18 to 46 years, all diagnosed with acute ON. These patients underwent up to 15 years of neurological and ophthalmologic examinations and imaging. The selection of variables for the developing model was based on clinical importance and statistical significance, and any missing values were appropriately addressed. We developed a Decision Tree (DT) classifier as the primary model and manually tuned its hyperparameters for optimal performance. We employed SHapley Additive exPlanations (SHAP) for feature importance evaluation. All analysis was performed using Python version 3.10.9 and its associated libraries. A total of 388 patients completed the study, of which 154 developed clinically definite multiple sclerosis (CDMS). It was observed that 61% of patients with magnetic resonance imaging (MRI) lesions developed CDMS. The final variables selected for analysis were MRI lesions, neurologic history, ON type, gender, and visual field mean deviation. The DT model achieved an accuracy of 70.1% during cross-validation and 69.1% on the test set, with an area under the curve of 74.9% and 71.7%, respectively. Comparative analysis of DT with other models showed similar performance. SHAP analysis revealed that MRI lesions and ON type emerged as the two most significant features, with relative importance of 61% and 18%, respectively. The decision tree model, with satisfactory performance, effectively stratifies patients, based on baseline findings and offers valuable insights for informed decision-making by physicians.

Correlation between magnetic resonance imaging (MRI) features and lesion size in focal nodular hyperplasia of the liver.

Focal nodular hyperplasia (FNH) in the liver is a benign lesion and the relationship between lesion size and imaging features is yet to be established. We aimed to develop and validate a scoring system to assess the relationship between magnetic resonance imaging (MRI) features and lesion size in FNH. This cross-sectional study was conducted at Zhongshan Hospital, Fudan University in Shanghai, China, from August 2019 and March 2023. Three hundred and seven patients with 363 surgically confirmed FNHs were retrospectively enrolled (training set: 254 lesions, validation set: 109 lesions). Lesions were divided into large (>3 cm) and small (≤3 cm) groups according to the diameter. Multivariate logistic regression was used to assess imaging features associated with lesion size in the training set. A scoring system was constructed and verified. The discrimination and calibration performance of the scoring system were evaluated by the receiver operating characteristic and calibration curve. In the training set, a round appearance (P<0.001) and hyperintense on arterial, portal and delayed phase (P=0.004) were more frequently observed in small FNHs, whereas a lobulated appearance [odds ratio (OR) =4.155, 95% confidence interval (CI): 2.023-8.536; P<0.001], feeding artery (OR =7.083, 95% CI: 2.970-16.892; P<0.001), radiating septa (OR =3.747, 95% CI: 1.682-8.347; P=0.001), central scar (OR =2.838, 95% CI: 1.284-6.273; P=0.010), hyperintense on arterial phase, hyper to isointense on portal phase, and isointense on delayed phase (OR =3.539, 95% CI: 1.650-7.595; P=0.001) were significantly associated with large FNHs. A scoring system derived from these variables showed an area under the curve of 0.888 (95% CI: 0.843-0.924) and 0.896 (95% CI: 0.821-0.945) in training and validation set respectively. Through scoring MRI features, it can be observed that these features contribute differently to the diagnosis of FNH depending on its size.

Evaluation of the SwiftScan mode in bone single-photon emission computed tomography (SPECT) imaging: effect on imaging quality and a semi-quantitative analysis.

99mTc-stannous methylene diphosphonate (99mTc-MDP) bone single-photon emission computed tomography/computed tomography (SPECT/CT) imaging plays a crucial role in various clinical applications. Many strategies have been developed to reduce the injection activity and procedure time, improve the patient experience and reduce their anxiety prior to and during SPECT imaging. This study aimed to evaluate the SwiftScan mode and its effect on image quality, and diagnostic performance of malignant skeletal lesions in bone SPECT image. 99mTc-MDP SPECT/CT scans were acquired from the National Electrical Manufacturers Association (NEMA) phantom with a sphere-to-background ratio of 10:1, and 69 patients were enrolled in this retrospective study. The patients who showed abnormal uptake of focal 99mTc-MDP on whole-body bone planar scans, underwent local SPECT/CT scans. The image comparison and analysis were performed in three acquisition modes: SwiftScan; SwiftScan-50% (the SwiftScan mode with a 50% reduction in acquisition time); and step-and-shoot (SS). The image mean count, coefficient of variation (COV), radioactive concentration, and signal-to-background ratio (SBR) were measured for each phantom image. Visual assessment (using a 5-point Likert scale), semi-quantitative analysis [to determine the COV, mean of the standard uptake value (SUVmean), maximum of the standard uptake value (SUVmax), and SBR], and receiver operating characteristics (ROC) curve analysis were performed for clinical assessment. The SwiftScan mode obtained better quality images than the SwiftScan-50% and SS modes, and it also had better sphere clarity and lower background COVs (P<0.05). The mean counts of the SwiftScan, and the corresponding radioactive concentration and SBR were significantly higher than those of the SwiftScan-50% and SS (P<0.05), except in relation to the 13-10 mm spheres. A total of 101 abnormal 99mTc-MDP uptake skeletal lesions were included in the clinical study. The SwiftScan had significantly higher visual scores than both the SwiftScan-50% and SS (P<0.001). Additionally, there were significantly higher visual scores in the SwiftScan-50% than SS (P<0.05). The SwiftScan also had a lower COV, indicating reduced image noise. The SUVmax, SUVmean, and SBR for the skeletal lesions were significantly higher in the SwiftScan than the other two modes (P<0.05). The optimal diagnostic cut-off values of the SUVmax for identifying malignant skeletal lesions in the SwiftScan, SwiftScan-50%, and SS were 17.95, 14.75, and 9.94, respectively, but there was no significant difference among the three modes in terms of their differential diagnostic performance (all P>0.05). The SwiftScan showed significant improvements in image quality and the semi-quantitative analysis of bone SPECT imaging. It reduced the scanning time by 50% without compromising the image quality or diagnostic performance. These findings provide valuable insights that may inform the clinical application of the SwiftScan, enhancing qualitative and quantitative diagnosis in bone SPECT imaging.

MPCNN: a novel approach for detecting human Monkeypox from skin lesion images leveraging deep neural network

The global healthcare scenario encounters a substantial challenge caused by the widespread outbreak of Monkeypox affecting over 65 countries. Limited availability of polymerase chain reaction (PCR) tests and biochemical assays necessitates alternative strategies. This study explores the viability of computer-aided identification of Monkeypox through the analysis of skin lesion images, offering a potential solution, particularly in resource-constrained settings. Employing data augmentation techniques, we augment the dataset to enhance its robustness. Subsequently, we utilize various pre-trained deep learning models, including EfficientNetB3, VGG16, ResNet50, AlexNet, and EfficientNet for classification tasks related to Monkeypox and other diseases. The achieved accuracies for these models are 98.48%, 69.19%, 91.41%, 78.38%, and 94.44%, respectively. We introduce a novel modified convolutional neural network (CNN) architecture named MPCNN to further improve performance. Our proposed MPCNN model demonstrates exceptional accuracy, precisely identifying Monkeypox patients with a remarkable precision of 99.49%. This technological advancement in disease identification holds significant promise for enhancing healthcare strategies and response mechanisms in the context of global health concerns.

Segmentation of Plain CT Image of Ischemic Lesion based on Trans-Swin-UNet

The present study aims to build a hybrid convolutional neural network and transformer UNet-based model, Trans-Swin-UNet, to segment ischemic lesions of the plain computed tomography (CT) image. The model architecture is built based on TransUnet and has four main improvements. First, replace the decoder of TransUNet with a Swin transformer; second, add a Max Attention module into the skip connection; third, design a comprehensive loss function; and last, speed up the segmentation performance. The present study designs two experiments to evaluate the performance of the built model using both the self-collected and public plain CT image datasets. The model optimization experiment evaluates the improvements of Trans-Swin-UNet over TransUnet. The experimental results show that each improvement of the built model can achieve a better performance than TransUNet in terms of dice similarity coefficient (DSC), Jaccard coefficient (JAC), and accuracy (ACC). The comparison experiment compares the built model with four existing UNet-based models. The experimental results show that the built model had a DSC of 0.72±0.01, a JAC of 0.78±0.04, an ACC of 0.75±0.03 using the self-collected plain CT image dataset and a DSC of 0.73±0.02, a JAC of 0.79±0.03, an ACC of 0.76±0.02 using the public plain CT image dataset, achieving the best segmentation performance among five UNet-based neural network models. The two experimental results conclude that the built model could accurately segment ischemic lesions of the plain CT image. The limitations and future work of this study are also discussed.

Glutathione-activated near-infrared II fluorescent probe for lung metastatic diagnosis and intraoperative imaging of tumor

Accurate identification of intraoperative tumor lesions and effective treatment are crucial for improving surgical outcomes. Near-infrared (NIR) fluorescence imaging demonstrates advantages over traditional medical approaches in tumor interventions, garnering significant attention. However, clinically available imaging agents are generally limited by their "always on" characteristics, which can lead to non-specific imaging interference and "false-positive" results. In this context, we present a glutathione-activated NIR-II probe, LJ-GSH, designed for metastatic tumor imaging and specific imaging-guided tumor resection. LJ-GSH initially exhibits quenched fluorescence due to the weak electron-donating effect of the thiophenol moiety, which is recovered at 815/910 nm upon activation by the overexpressed levels of glutathione (GSH) in tumor cells and tissues, significantly enhancing the specificity of tumor imaging. This unique characteristic positions LJ-GSH as a reliable fluorescent sensor for monitoring GSH dynamics during physiological events. Notably, the probe's NIR-II emission feature markedly improves imaging contrast and resolution, facilitating real-time identification and imaging of lung metastatic lesions. With the aid of high-specific NIR-II imaging guidance, tumor tissues can be precisely resected, with the residual negative margin diameter reduced to approximately 0.2 mm. We envision that our tailored probe may offer an attractive option for clinical applications.

Telediagnosis performance of specialists in oral medicine and general dental practitioner using images of oral mucosa lesions in Chile.

The objective of this research was to evaluate the performance of specialists in oral medicine doing diagnosis of oral mucosa lesions through digital images comparing general dental practitioner in Chile. 20 oral medicine specialists from the national registry of specialists in Chile and 20 general dental practitioners were invited to participate. Each participant reviewed 33 cases with digital images of oral mucosa lesions and was asked to submit diagnostic hypotheses. The proportions of correct diagnoses and diagnostic accuracy were determined. Analyses were performed using STATA 16.0. Specialists presented a higher total proportion of correct diagnoses than general dental practioners (86.5% vs 49.2%). Specialists also showed higher sensitivity (88.5% vs 59.3%) and greater specificity (85.8% vs 48.6 %) than general dental practioners in the diagnosis of oral cancer and oral potentially malignant disorders compared to benign oral lesions. In Chile, oral medicine specialists are a reliable alternative to provide diagnostic guidance through e-consult, but its margin of error must be considered.

Transformer Dil-DenseUnet: An Advanced Architecture for Stroke Segmentation

We propose a novel architecture, Transformer Dil-DenseUNet, designed to address the challenges of accurately segmenting stroke lesions in MRI images. Precise segmentation is essential for diagnosing and treating stroke patients, as it provides critical spatial insights into the affected brain regions and the extent of damage. Traditional manual segmentation is labor-intensive and error-prone, highlighting the need for automated solutions. Our Transformer Dil-DenseUNet combines DenseNet, dilated convolutions, and Transformer blocks, each contributing unique strengths to enhance segmentation accuracy. The DenseNet component captures fine-grained details and global features by leveraging dense connections, improving both precision and feature reuse. The dilated convolutional blocks, placed before each DenseNet module, expand the receptive field, capturing broader contextual information essential for accurate segmentation. Additionally, the Transformer blocks within our architecture address CNN limitations in capturing long-range dependencies by modeling complex spatial relationships through multi-head self-attention mechanisms. We assess our model’s performance on the Ischemic Stroke Lesion Segmentation Challenge 2015 (SISS 2015) and ISLES 2022 datasets. In the testing phase, the model achieves a Dice coefficient of 0.80 ± 0.30 on SISS 2015 and 0.81 ± 0.33 on ISLES 2022, surpassing the current state-of-the-art results on these datasets.

AFCF-Net: A novel U-Net based asymmetric feature calibration and fusion network for skin lesion image segmentation.

Skin lesion segmentation plays a pivotal role in the diagnosis and treatment of skin diseases. By using deep neural networks to segment lesion areas, doctors can more accurately assess the severity of health-related conditions of patients and promptly implement appropriate treatment measures, thereby enhancing treatment outcomes and improving the quality of life (QoL) of patients. However, existing segmentation networks still face challenges in balancing segmentation performance and efficiency. To address this issue, a novel network, named AFCF-Net, is proposed in this paper for skin lesion segmentation tasks. Firstly, the proposed network employs a newly designed spatial channel feature calibration convolution (SCFCConv) to enhance its ability to perceive spatial and channel features. Secondly, AFCF-Net utilizes newly designed feature symmetric fusion convolution (FSFConv) in skip connections to selectively fuse features from different levels, thereby enhancing its sensitivity to texture, edges, and other detailed features. In addition, a feature attention recombination module (FARM) is added to the bottleneck of the proposed network to comprehensively acquire and utilize contextual information at different scales, thus improving the network's generalization ability. Finally, a newly designed multi-level feature aggregation branch is introduced as an additional decoder for AFCF-Net to supplement key features lost during the original decoding process. Experiments, conducted on four skin image datasets, demonstrate that the proposed AFCF-Net network achieves better segmentation performance with fewer parameters and computational resources, compared to state-of-the-art segmentation networks. Additionally, AFCF-Net exhibits stronger generalization ability.

Diagnosis and treatment experience of Chlamydia psittaci pneumonia: A multicenter retrospective study in China

BackgroundIn recent years, clinical cases of Chlamydia psittaci pneumonia have gradually increased. Chlamydia psittaci pneumonia can quickly progress to severe pneumonia, leading to respiratory failure. Chlamydia psittaci cannot be detected by commonly used detection methods, leading to difficulties in clinical diagnosis and treatment, which may eventually develop into severe Chlamydia psittaci pneumonia. Therefore, we should improve our diagnostic and treatment capabilities for this disease.MethodsThis article retrospectively studied 61 cases of pneumonia from 12 different provinces in China, and classified them into severe pneumonia and non-severe pneumonia. Divided into two groups, with 23 cases of severe pneumonia and 38 cases of non-severe pneumonia. We compared the two groups in terms of imaging, laboratory testing, treatment, prognosis, etc., hoping to provide better explanations and help clinical physicians better diagnose the disease.ResultsThere is no difference in age, gender, contact history, smoking history, or basic disease between severe and non-severe Chlamydia psittaci pneumonia. The most prominent symptom is fever, with a median body temperature of 39.7 ℃, and other symptoms similar to common respiratory infections. White blood cells and neutrophils may not show significant elevation, but C-reactive protein (CRP) and procalcitonin (PCT) will both show significant elevation, often accompanied by a decrease in arterial oxygen pressure. Some patients may experience liver and kidney dysfunction, but there is no statistically significant difference between the two groups. Doxycycline remains the preferred drug for severe Chlamydia psittaci pneumonia. The imaging mainly shows consolidation and bronchial inflation sign, and may also present with pleural effusion.ConclusionsThe imaging of Chlamydia psittaci pneumonia usually shows consolidation with bronchial inflation sign, which can also be manifested as ground glass changes or solid masses. Atypical images are prone to misdiagnosis in the early stages. There are some differences in laboratory examination and imaging between severe pneumonia and non-severe pneumonia, which can help identify severe patients in the early stage. Metagenomic next-generation sequencing (mNGS) played an important role in the diagnosis of all cases in this article. Timely treatment has a good prognosis for the disease, and imaging lesions can be completely absorbed.

A Dual-Branch Residual Network with Attention Mechanisms for Enhanced Classification of Vaginal Lesions in Colposcopic Images

Vaginal intraepithelial neoplasia (VAIN), linked to HPV infection, is a condition that is often overlooked during colposcopy, especially in the vaginal vault area, as clinicians tend to focus more on cervical lesions. This oversight can lead to missed or delayed diagnosis and treatment for patients with VAIN. Timely and accurate classification of VAIN plays a crucial role in the evaluation of vaginal lesions and the formulation of effective diagnostic approaches. The challenge is the high similarity between different classes and the low variability in the same class in colposcopic images, which can affect the accuracy, precision, and recall rates, depending on the image quality and the clinician’s experience. In this study, a dual-branch lesion-aware residual network (DLRNet), designed for small medical sample sizes, is introduced, which classifies vaginal lesions by examining the relationship between cervical and vaginal lesions. The DLRNet model includes four main components: a lesion localization module, a dual-branch classification module, an attention-guidance module, and a pretrained network module. The dual-branch classification module combines the original images with segmentation maps obtained from the lesion localization module using a pretrained ResNet network to fine-tune parameters at different levels, explore lesion-specific features from both global and local perspectives, and facilitate layered interactions. The feature guidance module focuses the local branch network on vaginal-specific features by using spatial and channel attention mechanisms. The final integration involves a shared feature extraction module and independent fully connected layers, which represent and merge the dual-branch inputs. The weighted fusion method effectively integrates multiple inputs, enhancing the discriminative and generalization capabilities of the model. Classification experiments on 1142 collected colposcopic images demonstrate that this method raises the existing classification levels, achieving the classification of VAIN into three lesion grades, thus providing a valuable tool for the early screening of vaginal diseases.