Search

Background: Tele‐consultations are increasingly used for screening and diagnosis. Only a few studies have assessed dental students' visual attention to clinical images. Aim: To (i) determine dental students' gaze behavior, visual fixations, and diagnostic competence while viewing clinical images, and (ii) explore potential opportunities to strengthen the teaching–learning approaches. Design: Tobii Pro Nano‐device captured the eye‐tracking data for 65 dental undergraduate students in this cross‐sectional study. The predetermined areas of interest (AOI) for all five clinical photographs were uploaded onto Tobii software. All participants used a think‐aloud protocol with no restrictions to view time. Results: A total of 325 clinical pictures were analyzed, and the average view time was 189.25 ± 76.90 s. Most participants started at the center of the image (three frontal photos), spent a significant share of their view time on prominent findings, did not follow a systematic pattern, and exhibited diagnostic incompetence. Also, most participants followed a "Z" viewing pattern (oscillating movement from left to right) for the remaining two pictures. Conclusions: Subjects frequently fixated on the prominent AOI, however, failed to make the correct diagnosis. Their view patterns revealed no sequential viewing. Therefore, emphasizing knowledge about common dental abnormalities and focusing on full coverage of clinical pictures can improve dental students' diagnostic competence and view patterns. [ABSTRACT FROM AUTHOR]

Purpose: Breast density is a significant risk factor for breast cancer and can impact the sensitivity of screening mammography. Area‐based breast density measurements may not provide an accurate representation of the tissue distribution, therefore volumetric breast density (VBD) measurements are preferred. Dual‐energy mammography enables volumetric measurements without additional assumptions about breast shape. In this work we evaluated the performance of a dual‐energy decomposition technique for determining VBD by applying it to virtual anthropomorphic phantoms. Methods: The dual‐energy decomposition formalism was used to quantify VBD on simulated dual‐energy images of anthropomorphic virtual phantoms with known tissue distributions. We simulated 150 phantoms with volumes ranging from 50 to 709 mL and VBD ranging from 15% to 60%. Using these results, we validated a correction for the presence of skin and assessed the method's intrinsic bias and variability. As a proof of concept, the method was applied to 14 sets of clinical dual‐energy images, and the resulting breast densities were compared to magnetic resonance imaging (MRI) measurements. Results: Virtual phantom VBD measurements exhibited a strong correlation (Pearson's r>0.95$r > 0.95$) with nominal values. The proposed skin correction eliminated the variability due to breast size and reduced the bias in VBD to a constant value of −2%. Disagreement between clinical VBD measurements using MRI and dual‐energy mammography was under 10%, and the difference in the distributions was statistically non‐significant. VBD measurements in both modalities had a moderate correlation (Spearman's ρ$\rho \ $= 0.68). Conclusions: Our results in virtual phantoms indicate that the material decomposition method can produce accurate VBD measurements if the presence of a third material (skin) is considered. The results from our proof of concept showed agreement between MRI and dual‐energy mammography VBD. Assessment of VBD using dual‐energy images could provide complementary information in dual‐energy mammography and tomosynthesis examinations. [ABSTRACT FROM AUTHOR]

Similarly to the majority of deep learning applications, diagnosing skin diseases using computer vision and deep learning often requires a large volume of data. However, obtaining sufficient data for particular types of facial skin conditions can be difficult, due to privacy concerns. As a result, conditions like rosacea are often understudied in computer-aided diagnosis. The limited availability of data for facial skin conditions has led to the investigation of alternative methods of computer-aided diagnosis. In recent years, generative adversarial networks (GANs), mainly variants of StyleGANs, have demonstrated promising results in generating synthetic facial images. In this study, for the first time, a small dataset of rosacea with 300 full-face images was utilized to further investigate the possibility of generating synthetic data. Our experimentation demonstrated that the strength of R 1 regularization is crucial for generating high-fidelity rosacea images using a few hundred images. This was complemented by various experimental settings to ensure model convergence. We successfully generated 300 high-quality synthetic images, significantly contributing to the limited pool of rosacea images for computer-aided diagnosis. Additionally, our qualitative evaluations by 3 expert dermatologists and 23 non-specialists highlighted the realistic portrayal of rosacea features in the synthetic images. We also provide a critical analysis of the quantitative evaluations and discuss the limitations of solely relying on validation metrics in the field of computer-aided clinical image diagnosis. [ABSTRACT FROM AUTHOR]

Numerous classification systems have been developed over the years, systems which not only provide assistance to dermatologists, but also enable individuals, especially those living in areas with low medical access, to get a diagnosis. In this paper, a Machine Learning model, which performs a binary classification, and, which for the remainder of this paper will be abbreviated as ML model, is trained and tested, so as to evaluate its effectiveness in giving the right diagnosis, as well as to point out the limitations of the given method, which include, but are not limited to, the quality of smartphone images, and the lack of FAIR image datasets for model training. The results indicate that there are many measures to be taken and improvements to be made, if such a system were to become a reliable tool in real-life circumstances. [ABSTRACT FROM AUTHOR]

IntroductionMalignant skin lesions pose a great threat to the health of patients. Due to the limitations of existing diagnostic techniques, such as poor accuracy and invasive operations, malignant skin lesions are highly similar to other skin lesions, with low diagnostic efficiency and high misdiagnosis rates. Automatic medical image classification using computer algorithms can effectively improve clinical diagnostic efficiency. However, existing clinical datasets are sparse and clinical images have complex backgrounds, problems with noise interference such as light changes and shadows, hair occlusions, etc. In addition, existing classification models lack the ability to focus on lesion regions in complex backgrounds.MethodsIn this paper, we propose a DBN (double branch network) based on a two-branch network model that uses a backbone with the same structure as the original network branches and the fused network branches. The feature maps of each layer of the original network branch are extracted by our proposed CFEBlock (Common Feature Extraction Block), the common features of the feature maps between adjacent layers are extracted, and then these features are combined with the feature maps of the corresponding layers of the fusion network branch by FusionBlock, and finally the total prediction results are obtained by weighting the prediction results of both branches. In addition, we constructed a new dataset CSLI (Clinical Skin Lesion Images) by combining the publicly available dataset PAD-UFES-20 with our collected dataset, the CSLI dataset contains 3361 clinical dermatology images for six disease categories: actinic keratosis (730), cutaneous basal cell carcinoma (1136), malignant melanoma (170) cutaneous melanocytic nevus (391), squamous cell carcinoma (298) and seborrheic keratosis (636).ResultsWe divided the CSLI dataset into a training set, a validation set and a test set, and performed accuracy, precision, sensitivity, specificity, f1score, balanced accuracy, AUC summary, visualisation of different model training, ROC curves and confusion matrix for various diseases, ultimately showing that the network performed well overall on the test data.DiscussionThe DBN contains two identical feature extraction network branches, a structure that allows shallow feature maps for image classification to be used with deeper feature maps for information transfer between them in both directions, providing greater flexibility and accuracy and enhancing the network's ability to focus on lesion regions. In addition, the dual branch structure of DBN provides more possibilities for model structure modification and feature transfer, and has great potential for development.

Clinical images are of utmost importance for the majority of dermatological research and publications. The rich collection of clinical images in medical journals may help in formulating machine learning programs in the future or facilitate image-based meta-analysis. However, the presence of a scale bar in those images is required for measuring the lesion from an image. We audited recent issues of three widely circulated Indian dermatology journals and found that among 345 clinical images, 2.61% had a scale with the unit. With this background, in this article, we provided three methods for capturing and processing clinical images with scale. This article would help dermatologists to think about incorporating a scale bar in the image for the progress of science.

Clinical images are of utmost importance for the majority of dermatological research and publications. The rich collection of clinical images in medical journals may help in formulating machine learning programs in the future or facilitate image-based meta-analysis. However, the presence of a scale bar in those images is required for measuring the lesion from an image. We audited recent issues of three widely circulated Indian dermatology journals and found that among 345 clinical images, 2.61% had a scale with the unit. With this background, in this article, we provided three methods for capturing and processing clinical images with scale. This article would help dermatologists to think about incorporating a scale bar in the image for the progress of science. [ABSTRACT FROM AUTHOR]

Background: Labeling color fundus photos (CFP) is an important step in the development of artificial intelligence screening algorithms for the detection of diabetic retinopathy (DR). Most studies use the International Classification of Diabetic Retinopathy (ICDR) to assign labels to CFP, plus the presence or absence of macular edema (ME). Images can be grouped as referrable or nonreferrable according to these classifications. There is little guidance in the literature about how to collect and use metadata as a part of the CFP labeling process., Objective: This study aimed to improve the quality of the Multimodal Database of Retinal Images in Africa (MoDRIA) by determining whether the availability of metadata during the image labeling process influences the accuracy, sensitivity, and specificity of image labels. MoDRIA was developed as one of the inaugural research projects of the Mbarara University Data Science Research Hub, part of the Data Science for Health Discovery and Innovation in Africa (DS-I Africa) initiative., Methods: This is a crossover assessment with 2 groups and 2 phases. Each group had 10 randomly assigned labelers who provided an ICDR score and the presence or absence of ME for each of the 50 CFP in a test image with and without metadata including blood pressure, visual acuity, glucose, and medical history. Specificity and sensitivity of referable retinopathy were based on ICDR scores, and ME was calculated using a 2-sided t test. Comparison of sensitivity and specificity for ICDR scores and ME with and without metadata for each participant was calculated using the Wilcoxon signed rank test. Statistical significance was set at P<.05., Results: The sensitivity for identifying referrable DR with metadata was 92.8% (95% CI 87.6-98.0) compared with 93.3% (95% CI 87.6-98.9) without metadata, and the specificity was 84.9% (95% CI 75.1-94.6) with metadata compared with 88.2% (95% CI 79.5-96.8) without metadata. The sensitivity for identifying the presence of ME was 64.3% (95% CI 57.6-71.0) with metadata, compared with 63.1% (95% CI 53.4-73.0) without metadata, and the specificity was 86.5% (95% CI 81.4-91.5) with metadata compared with 87.7% (95% CI 83.9-91.5) without metadata. The sensitivity and specificity of the ICDR score and the presence or absence of ME were calculated for each labeler with and without metadata. No findings were statistically significant., Conclusions: The sensitivity and specificity scores for the detection of referrable DR were slightly better without metadata, but the difference was not statistically significant. We cannot make definitive conclusions about the impact of metadata on the sensitivity and specificity of image labels in our study. Given the importance of metadata in clinical situations, we believe that metadata may benefit labeling quality. A more rigorous study to determine the sensitivity and specificity of CFP labels with and without metadata is recommended., (©Simon Arunga, Katharine Elise Morley, Teddy Kwaga, Michael Gerard Morley, Luis Filipe Nakayama, Rogers Mwavu, Fred Kaggwa, Julius Ssempiira, Leo Anthony Celi, Jessica E Haberer, Celestino Obua. Originally published in JMIR Formative Research (https://formative.jmir.org), 18.09.2024.)

The early detection of Non-Melanoma Skin Cancer (NMSC) is crucial to achieve the best treatment outcomes. Shape is considered one of the main parameters taken for the detection of some types of skin cancer such as melanoma. For NMSC, the importance of shape as a visual detection parameter is not well-studied. A dataset of 993 standard camera images containing different types of NMSC and benign skin lesions was analysed. For each image, the lesion boundaries were extracted. After an alignment and scaling, Elliptic Fourier Analysis (EFA) coefficients were calculated for the boundary of each lesion. The asymmetry of lesions was also calculated. Then, multivariate statistics were employed for dimensionality reduction and finally computational learning classification was employed to evaluate the separability of the classes. The separation between malignant and benign samples was successful in most cases. The best-performing approach was the combination of EFA coefficients and asymmetry. The combination of EFA and asymmetry resulted in a balanced accuracy of 0.786 and an Area Under Curve of 0.735. The combination of EFA and asymmetry for lesion classification resulted in notable success rates when distinguishing between benign and malignant lesions. In light of these results, skin lesions' shape should be integrated as a fundamental part of future detection techniques in clinical screening. [ABSTRACT FROM AUTHOR]

Objectives To research and develop tools and methods for building cohorts of images linked to longitudinal healthcare records from real-world clinical images from the whole Scottish population. To provide this capability for the Scottish Medical Imaging service (provided by the Scottish National Safe Haven) to support research and AI projects. Approach Clinical images, especially when linked to routinely collected health data, are extremely useful for many types of research and AI development. However, finding and using clinical images for research data is challenging because: 1) Existing software used to search for images are designed for clinical care rather than research making it easy to find images for a particular patient. They are not designed to search for all images with particular characteristics e.g. slice thickness/scanning protocol/contrast agent/patient medication. 2) Reuse of clinical images for research requires de-identification, yet identifiable data can be present in many areas of the associated image file. Results The PICTURES (InterdisciPlInary Collaboration for efficienT and effective Use of clinical images in big data health care RESearch) 5-year programme has developed an architecture for building cohorts of images based upon research criteria and providing these in a di-identifiable form within a Safe Haven environment. There are 3 zones: • An identifiable zone which stores the raw image data and a MongoDB database which captures the metadata • A de-identified zone which provides a database and tools for cohort building which do not require imaging data expertise • Several Project Private Zones (PPZs) where researchers can install custom software and access the de-identified images for their project The architecture supports cohort building based upon features within pixel data, image metadata and linking to longitudinal health care records. Conclusion PICTURES is currently enhancing the cohort building user interface used by the National Safe Haven and supporting exemplar projects. The SMI service is live and accepting requests for more information. The software is open source and we welcome the use of the platform by other Safe Havens/research groups.

Background: Images are an important element of any scientific publication, more so for a very visual specialty such as dermatology. Dermatology journals generally include detailed instructions related to images submitted with the manuscripts, including technical aspects such as file format, resolution, and editing. We aimed to review and critically analyze instructions for images in the top 50 dermatology journals [as per the latest Clarivate journal citation report (JCR) ranking]. Methods: The top 50 journals as per the latest Clarivate JCR were included in the study. Instructions for images were reviewed for each of the journals. The main points analyzed included file type, resolution, size limits, editing limits, support for post-processing of images and details regarding patient consent. Results: Only half the journals (25) had clear, detailed, and specific instructions for the images. Only one journal specifically mentioned pixels per inch (PPI) as the descriptive term, and the remaining used dots per inch (DPI), whereas 6 did not mention either. Twenty-three journals did not mention the recommended size of the image. The most common minimum resolution mentioned was 300 DPI (41 journals). Although 24 of the journals mentioned editing limits, none of the journals elaborated on image plagiarism in the instructions. Twenty-one of the fifty did not mention a clear policy on patient consent. Conclusions: Image submission guidelines for dermatology journals need to be made more elaborate yet easier to understand. Developing a consensus, followed by standardization of these submission guidelines, can help both authors and journals. [ABSTRACT FROM AUTHOR]

The demand for remote health monitoring will significantly increase in the near future due to a decrease in the doctor/patient ratio as the world population grows. In remote health monitoring, one of the major goals is to send patient's data to clinical experts at geographically distant locations. In this scenario, the importance of implementing security on patients' clinical information increases, so that this could not be either changed or read by an unauthorized person. Rivest Cipher (RC5) algorithm is a secure and simple encryption algorithm. Due to its simplicity, fast encryption, and low memory requirements, it is considered as a suitable block cipher for resource constraint environments, such as body area networks (BAN), although it can be broken by various attacks due to poor diffusion. In this paper, we propose a chaotic-based key scheduling algorithm for the RC5 in which round keys are generated based on 2-D chaotic maps and used as a symmetric key during the encryption and decryption process. Cipher feedback mode is adapted to further increase the diffusion property of the cipher. This chaos-based key is very difficult to trace, and the cipher is extremely hard to break. A strict security analysis of the proposed cipher is performed against several attacks. The experimental results have shown that the impact of this approach is significantly better than the conventional security mechanisms in case of BAN when applied to critical clinical images.

Use of patient clinical photographs requires specific attention to confidentiality and privacy. Although there are policies and procedures for publishing clinical images, there is little systematic evidence about what patients and health professionals actually think about consent for publishing clinical images. We investigated the opinions of three stakeholder groups (patients, students and doctors) at 3 academic healthcare institutions and 37 private practices in Croatia (total 791 participants: 292 patients, 281 medical and dental students and 281 doctors of medicine or dental medicine). The questionnaire contained patient photographs with different levels of anonymization. All three respondent groups considered that more stringent forms of permission for were needed identifiable photographs than for those with higher levels of anonymization. When the entire face was presented in a photo only 33% of patients considered that written permission was required, compared with 88% of the students and 89% of the doctors. Opinions about publishing patient photographs differed among the three respondent samples: almost half of the patients thought no permission was necessary compared with one-third of students and doctors. These results show poor awareness of Croatian patients regarding the importance of written informed consent as well as unsatisfactory knowledge of health professionals about policies on the publication of patients' data in general. In conclusion, there is a need for increasing awareness of all stakeholders to achieve better protection of patient privacy rights in research and publication. [ABSTRACT FROM AUTHOR]

Purpose: Automated assessment of perceptual image quality on clinical Computed Tomography (CT) data by computer algorithms has the potential to greatly facilitate data‐driven monitoring and optimization of CT image acquisition protocols. The application of these techniques in clinical operation requires the knowledge of how the output of the computer algorithms corresponds to clinical expectations. This study addressed the need to validate algorithmic image quality measurements on clinical CT images with preferences of radiologists and determine the clinically acceptable range of algorithmic measurements for abdominal CT examinations. Materials and methods: Algorithmic measurements of image quality metrics (organ HU, noise magnitude, and clarity) were performed on a clinical CT image dataset with supplemental measures of noise power spectrum from phantom images using techniques developed previously. The algorithmic measurements were compared to clinical expectations of image quality in an observer study with seven radiologists. Sets of CT liver images were selected from the dataset where images in the same set varied in terms of one metric at a time. These sets of images were shown via a web interface to one observer at a time. First, the observer rank ordered the CT images in a set according to his/her preference for the varying metric. The observer then selected his/her preferred acceptable range of the metric within the ranked images. The agreement between algorithmic and observer rankings of image quality were investigated and the clinically acceptable image quality in terms of algorithmic measurements were determined. Results: The overall rank‐order agreements between algorithmic and observer assessments were 0.90, 0.98, and 1.00 for noise magnitude, liver parenchyma HU, and clarity, respectively. The results indicate a strong agreement between the algorithmic and observer assessments of image quality. Clinically acceptable thresholds (median) of algorithmic metric values were (17.8, 32.6) HU for noise magnitude, (92.1, 131.9) for liver parenchyma HU, and (0.47, 0.52) for clarity. Conclusions: The observer study results indicated that these algorithms can robustly assess the perceptual quality of clinical CT images in an automated fashion. Clinically acceptable ranges of algorithmic measurements were determined. The correspondence of these image quality assessment algorithms to clinical expectations paves the way toward establishing diagnostic reference levels in terms of clinically acceptable perceptual image quality and data‐driven optimization of CT image acquisition protocols. [ABSTRACT FROM AUTHOR]

Two‐dimensional transforms are extensively used for speckle noise reduction (despeckling) in ultrasound images. This work proposes a double filter bank structure consisting of a discrete wavelet packet transform (DWPT) and directional filter bank (DFB) along with a fuzzy‐based clustering technique to despeckle ultrasound images. Wavelet packet transform can efficiently reject noises based on grey scale relational thresholding and DFB can efficiently preserve edge information. In this study, instead of conventional thresholding methods, fuzzy‐based clustering techniques are applied for noise rejection. The algorithm provides a consistent improvement over the competing state‐of‐the‐art speckle reduction algorithms due to the improved ability to preserve geometric features while rejecting speckle noise adaptively. The authors claim is validated by applying a number of clinical images with performance indices such as peak signal to noise ratio, mean structural similarity, signal to mean square error, speckle signal to noise ratio, speckle suppression index and edge preservation index. [ABSTRACT FROM AUTHOR]

In this study, a novel probabilistic, geometric and dynamic shape‐based level sets method is proposed. The shape prior is coupled with the intensity information to enhance the segmentation results. The two‐dimensional principal component analysis method is applied on the training shapes to represent the shape variation with enough number of shape projections in the training step. The shape model is constructed using the implicit representation of the projected shapes. A new energy functional is proposed (i) to embed the shape model into the image domain and (ii) to estimate the shape coefficients. The proposed method is validated on synthetic and clinical images with various challenges such as the noise, occlusion and missing information. The authors compare their method with some of related works. Experiments show that the proposed segmentation method is more accurate and robust than other alternatives under different challenges. ∗ Note: Colour figures are available in the online version of this paper.

Diffusion imaging is critical for detecting acute brain injury. However, normal apparent diffusion coefficient (ADC) maps change rapidly in early childhood, making abnormality detection difficult. In this article, we explored clinical PACS and electronic healthcare records (EHR) to create age-specific ADC atlases for clinical radiology reference. Using the EHR and three rounds of multiexpert reviews, we found ADC maps from 201 children 0-6 years of age scanned between 2006 and 2013 who had brain MRIs with no reported abnormalities and normal clinical evaluations 2+ years later. These images were grouped in 10 age bins, densely sampling the first 1 year of life (5 bins, including neonates and 4 quarters) and representing the 1-6 year age range (an age bin per year). Unbiased group-wise registration was used to construct ADC atlases for 10 age bins. We used the atlases to quantify (a) cross-sectional normative ADC variations; (b) spatiotemporal heterogeneous ADC changes; and (c) spatiotemporal heterogeneous volumetric changes. The quantified age-specific whole-brain and region-wise ADC values were compared to those from age-matched individual subjects in our study and in multiple existing independent studies. The significance of this study is that we have shown that clinically acquired images can be used to construct normative age-specific atlases. These first of their kind age-specific normative ADC atlases quantitatively characterize changes of myelination-related water diffusion in the first 6 years of life. The quantified voxel-wise spatiotemporal ADC variations provide standard references to assist radiologists toward more objective interpretation of abnormalities in clinical images. Our atlases are available at . Hum Brain Mapp 38:3052-3068, 2017. © 2017 Wiley Periodicals, Inc. [ABSTRACT FROM AUTHOR]

The development of low profile gamma-ray detectors has encouraged the production of small field of view (SFOV) hand-held imaging devices for use at the patient bedside and in operating theatres. Early development of these SFOV cameras was focussed on a single modality—gamma ray imaging. Recently, a hybrid system—gamma plus optical imaging—has been developed. This combination of optical and gamma cameras enables high spatial resolution multi-modal imaging, giving a superimposed scintigraphic and optical image. Hybrid imaging offers new possibilities for assisting clinicians and surgeons in localising the site of uptake in procedures such as sentinel node detection. The hybrid camera concept can be extended to a multimodal detector design which can offer stereoscopic images, depth estimation of gamma-emitting sources, and simultaneous gamma and fluorescence imaging. Recent improvements to the hybrid camera have been used to produce dual-modality images in both laboratory simulations and in the clinic. Hybrid imaging of a patient who underwent thyroid scintigraphy is reported. In addition, we present data which shows that the hybrid camera concept can be extended to estimate the position and depth of radionuclide distribution within an object and also report the first combined gamma and Near-Infrared (NIR) fluorescence images. [ABSTRACT FROM AUTHOR]

Introduction: Artificial intelligence in oncology has gained a lot of interest in recent years. Early detection of Oral squamous cell carcinoma (OSCC) is crucial for early management to attain a better prognosis and overall survival. Machine learning (ML) has also been used in oral cancer studies to explore the discrimination between clinically normal and oral cancer., Materials and Methods: A dataset comprising 360 clinical intra-oral images of OSCC, Oral Potentially Malignant Disorders (OPMDs) and clinically healthy oral mucosa were used. Clinicians trained the machine learning model with the clinical images (n=300). Roboflow software (Roboflow Inc, USA) was used to classify and annotate images along with Multi-class annotation and object detection models trained by two expert oral pathologists. The test dataset (n=60) of new clinical images was again evaluated by two clinicians and Roboflow. The results were tabulated and Kappa statistics was performed using SPSS v23.0 (IBM Corp., Armonk, NY).  Results: Training dataset clinical images (n=300) were used to train the clinicians and Roboflow algorithm. The test dataset (n=60) of new clinical images was again evaluated by the clinicians and Roboflow. The observed outcomes revealed that the Mean Average Precision (mAP) was 25.4%, precision 29.8% and Recall 32.9%. Based on the kappa statistical analysis the 0.7 value shows a moderate agreement between the clinicians and the machine learning model. The test dataset showed the specificity and sensitivity of the Roboflow machine learning model to be 75% and 88.9% respectively.  Conclusion: In conclusion, machine learning showed promising results in the early detection of suspected lesions using clinical intraoral images and aids general dentists and patients in the detection of suspected lesions such as OPMDs and OSCC that require biopsy and immediate treatment., Competing Interests: The authors have declared that no competing interests exist., (Copyright © 2023, Y et al.)

Introduction: Malignant skin lesions pose a great threat to the health of patients. Due to the limitations of existing diagnostic techniques, such as poor accuracy and invasive operations, malignant skin lesions are highly similar to other skin lesions, with low diagnostic efficiency and high misdiagnosis rates. Automatic medical image classification using computer algorithms can effectively improve clinical diagnostic efficiency. However, existing clinical datasets are sparse and clinical images have complex backgrounds, problems with noise interference such as light changes and shadows, hair occlusions, etc. In addition, existing classification models lack the ability to focus on lesion regions in complex backgrounds., Methods: In this paper, we propose a DBN (double branch network) based on a two-branch network model that uses a backbone with the same structure as the original network branches and the fused network branches. The feature maps of each layer of the original network branch are extracted by our proposed CFEBlock (Common Feature Extraction Block), the common features of the feature maps between adjacent layers are extracted, and then these features are combined with the feature maps of the corresponding layers of the fusion network branch by FusionBlock, and finally the total prediction results are obtained by weighting the prediction results of both branches. In addition, we constructed a new dataset CSLI (Clinical Skin Lesion Images) by combining the publicly available dataset PAD-UFES-20 with our collected dataset, the CSLI dataset contains 3361 clinical dermatology images for six disease categories: actinic keratosis (730), cutaneous basal cell carcinoma (1136), malignant melanoma (170) cutaneous melanocytic nevus (391), squamous cell carcinoma (298) and seborrheic keratosis (636)., Results: We divided the CSLI dataset into a training set, a validation set and a test set, and performed accuracy, precision, sensitivity, specificity, f1score, balanced accuracy, AUC summary, visualisation of different model training, ROC curves and confusion matrix for various diseases, ultimately showing that the network performed well overall on the test data., Discussion: The DBN contains two identical feature extraction network branches, a structure that allows shallow feature maps for image classification to be used with deeper feature maps for information transfer between them in both directions, providing greater flexibility and accuracy and enhancing the network's ability to focus on lesion regions. In addition, the dual branch structure of DBN provides more possibilities for model structure modification and feature transfer, and has great potential for development., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2023 Wang, Qi, Sun, Li and Yao.)

Orbital CT showed a subperiosteal abscess compressing superior-rectus and superior-oblique muscles, and opacification of bilateral sinuses (Fig. Graph: Figure 1 Photographs of ocular movement deficits of the left eye with various eye movements - directions of attempted gaze shown by closest arrow to images (panel a). Orbital cellulitis should be suspected if proptosis, diplopia, pain with eye movement, or ophthalmoplegia are present.[1]-[3] Direct extension from sinus infection is the most common source of orbital cellulitis and subperiosteal abscess.[1], [2], [4] For both conditions, orbital imaging is necessary for definitive diagnosis. [Extracted from the article]