Dermatoscopic Images Research Articles

Labial melanotic macules (LaMMs) are benign pigmented lesions that typically present as solitary patches on the lower lip. While their dermatoscopic features have been previously described, limited data exists on how these features vary across different skin types. To investigate the dermatoscopic differences of LaMMs in individuals with darker skin phototypes and compare them with lighter skin phototype LaMMs and mucosal melanomas. This retrospective study analyzes 64 LaMMs located at the mucocutaneous junction or on the mucous membrane, collected from 49 patients and four mucosal melanomas across three tertiary dermatology centers in Turkiye, Austria, and Australia between January 1, 2020, and December 31, 2022. Two investigators reviewed clinical and dermatoscopic images independently, followed by a consensus evaluation to determine the presence of dermatoscopic criteria. Lesions that were clinically equivocal and not confirmed by histopathology were excluded. Multiple LaMMs were observed in 30.6% of the patients, and the majority of LaMMs (91.8%) were located on the lower lip. Overall, 53.1% of lesions exhibited asymmetry, 54.7% showed more than one dermatoscopic pattern, and 68.8% demonstrated multiple colors. No significant differences were noted between darker (n=30, 61.2%) and lighter (n=19, 38,8%) patients with LaMMs regarding age, sex, lesion location, or the presence of multiple lesions. LaMMs in darker skin were more likely to be asymmetrical and displayed multiple dermatoscopic patterns. White structureless areas and gray dots/clods were more frequently observed in darker skin type. All labial melanomas exhibited asymmetry, as well as multiple patterns and colors on dermatoscopic examination. There was no significant difference in the rates of asymmetry, multiple patterns, or multiple colors compared to LaMMs included in the study. LaMMs in dark skin types more commonly present with asymmetry, multiple dermatoscopic patterns, and gray color.

The aim of the study: development of a screening method for patients aimed at early differential diagnosis of malignant skin neoplasms using dermatoscopy in combination with optoelectronic mobile equipment and algorithms for classifying dermatoscopic images based on machine learning methods.Materials and methods. To implement the detection of malignant neoplasms and classify them into the appropriate nosological group, machine learning methods, algorithms and optical recognition are used. The latter is used in the process of forming dermatoscopic images and training classification algorithms and models. The machine learning approaches are multi-class and binary cascade two-stage classification methods by classification algorithms based on the visual transformer architecture and neural network architecture.Results. During the experimental evaluation of the results of multi-class classification (eight types of malignant neoplasms), the best classification model with the visual transformer architecture was determined, characterized by the metrics Accuracy of 0.932 and F-measure of 0.891 on the formed dataset, including ISIC-2019 and our own set containing 657 images. The binary cascade two-stage classification for melanocytic and non-melanocytic neoplasms has Accuracy and F-measure values — of 0.954 and 0.948 (the first stage of classification) and for melanomas and nevi — 0.964 and 0.951, respectively (the second stage of classification).Conclusion. The obtained quantitative values of the malignant skin neoplasms detection accuracy by the developed screening examination method allow us to recommend the introduction of a multi-class classification for the primary division of a large volume of dermatoscopic images patients by nosological sign between medical specialists in the process of conducting mass (visiting) preventive examinations, and the introduction of a cascade binary classification in the an initial appointment conditions with limited access to specialized specialists to differentiate melanoma from other skin neoplasms. The developed screening examination method for patients can be introduced into medical practice as a system for supporting physician decision-making.

Introduction: Grover's disease (GD) is a rare acantholytic skin disorder typically characterized by pruritic vesicular or keratotic truncal papules, most commonly affecting older Caucasian males. Ultraviolet-induced fluorescence dermatoscopy (UVFD) and sub-ultraviolet reflectance dermatoscopy (sUVRD) are novel imaging techniques with potential diagnostic value in dermatology. Objectives: The objective of this study was to evaluate the dermatoscopic patterns of GD using UVFD and sUVRD techniques. Methods: A retrospective single-center cohort observational study was conducted including consecutive adult patients diagnosed with GD. Dermatoscopic images were obtained using a Dermlite DL5 dermatoscope paired with a smartphone for UVFD and a Casio DZ-D100 Dermocamera for sUVRD. Results: Among the 23 investigated patients (15 females, eight males; mean age 49.13 years), UVFD images frequently showed central polygonal bright scales with a greenish background. sUVRD images demonstrated hyporeflective polygonal scales, hyperreflective halos, and vascular patterns at the periphery. sUVRD was superior to UVFD and CD in the detection of semi-specific polygonal scales in GD. Eccrine duct involvement was observed in 76.31% of sUVRD images and 57.89% of matching conventional polarized dermatoscopy images. Contrary to the existing literature, female patients represented a higher percentage of the cohort. Twelve GD patients (52.2%) had a personal history of skin cancer, Conclusion: UVFD and sUVRD effectively characterized the unique features of GD lesions. Our findings suggest that GD may affect younger individuals and females more frequently than previously reported, potentially indicating underdiagnosis in this population. Incorporating dermatoscopy into routine examinations may improve the detection and management of GD.

Although recent advances in CNNs and Transformers have significantly improved medical image segmentation, these models often struggle to balance segmentation accuracy, inference speed, and architectural simplicity. Lightweight MLP-based methods have emerged as a promising alternative, but they frequently lack the ability to capture fine-grained spatial context, leading to suboptimal boundary localization. To address this issue, a hybrid architecture can be introduced that integrates the computational efficiency of MLPs with the spatial feature extraction strengths of convolutional or transformer-based modules. This design aims to deliver high segmentation accuracy while preserving low latency and minimal architectural complexity, thereby enhancing applicability in real-time clinical settings. Medical image segmentation remains a challenging task requiring both accuracy and computational efficiency in clinical settings. This paper introduces FocusGate-Net, a novel hybrid architecture combining shifted token MLP blocks, convolutional feature extractors, and dual-attention mechanisms for robust medical image segmentation. Our approach leverages the spatial dependency modeling capabilities of MLP architectures while enhancing feature selectivity through Convolutional Block Attention Module (CBAM) and Attention Gate (AG) mechanisms. We evaluate FocusGate-Net on three diverse medical image datasets: ISIC2018 for skin lesion segmentation, PH2 for dermatoscopic images, and Kvasir-SEG for polyp segmentation. Comprehensive ablation studies verify the contribution of each architectural component, demonstrating the effectiveness of our hybrid design. When benchmarked against state-of-the-art models like UNet, UNet++, and ResUNet, FocusGate-Net achieves superior performance, with a Dice coefficient of 92.47% and IoU of 86.36% on ISIC2018. Furthermore, our model demonstrates exceptional cross-dataset generalization capability, achieving Dice scores of 97.25% on PH2 and 94.83% on Kvasir-SEG. These results highlight the potential of MLP-based hybrid architectures with attention mechanisms for improving medical image segmentation accuracy while maintaining computational efficiency suitable for clinical deployment.

Introduction Teledermatology, which utilizes communication technologies to remotely assess skin lesions, has become a vital tool in healthcare. This study aimed to compare the diagnostic accuracy of teledermatology versus face-to-face examination and explore factors influencing accuracy, such as teledermatoscopy use, dermatoscopy type, and clinical experience. Methods Fifty-seven cutaneous tumors were evaluated using handheld or digital dermatoscopy in face-to-face examinations, and preliminary diagnoses were recorded. A definitive diagnosis was established through histopathological examination, which served as the reference standard. Macro and dermatoscopic images were then sent to six teledermatologists for remote diagnosis, and findings were analyzed statistically. Results The preliminary diagnosis matched the histopathological diagnosis in 84.2% of face-to-face cases. Teledermatologists achieved 63.7% accuracy with macro images alone, increasing to 70.8% with dermatoscopic images. Teledermatology showed lower accuracy than face-to-face examination, regardless of whether teledermatoscopy was used (p < 0.05), but accuracy significantly improved with dermatoscopic images (p = 0.004). The teledermatology’s accuracy for malignancy prediction was comparable to face-to-face examination (p > 0.05). Dermatoscopy type did not significantly impact accuracy (p > 0.05), while longer clinical experience correlated with higher accuracy (p < 0.05). Interrater reliability was poor for specific diagnoses but improved when categorizing lesions as malignant or benign (κ = 0.192, κ = 0.683). Conclusion Although teledermatology performed below face-to-face examination in terms of specific diagnoses, it remained effective in distinguishing between benign and malignant cutaneous tumors. The inclusion of teledermatoscopy and longer clinical experience enhanced diagnostic accuracy.

Early detection of cancer is critical for effective treatment, particularly for aggressive malignancies like skin cancer and brain tumors. This research presents an integrated deep learning approach combining augmentation, segmentation, and classification techniques to identify diverse tumor types in skin lesions and brain MRI scans. Our method employs a fine-tuned InceptionV3 convolutional neural network trained on a multi-modal dataset comprising dermatoscopy images from the Human Against Machine archive and brain MRI scans from the ISIC 2023 repository. To address class imbalance, we implement advanced preprocessing and Generative Adversarial Network (GAN)-based augmentation. The model achieves 97% accuracy in classifying images across ten categories: seven skin cancer types, multiple brain tumor variants, and an “undefined” class. These results suggest clinical applicability for multi-cancer detection.

Background/IntroductionSkin lesion classification poses a critical diagnostic challenge in dermatology, where early and accurate identification has a direct impact on patient outcomes. While deep learning approaches have shown promise using dermatoscopic images alone, the integration of clinical metadata remains underexplored despite its potential to enhance diagnostic accuracy.MethodsWe developed a novel multimodal data fusion framework that systematically integrates dermatoscopic images with clinical metadata for the classification of skin lesions. Using the HAM10000 dataset, we evaluated multiple fusion strategies, including simple concatenation, weighted concatenation, self-attention mechanisms, and cross-attention fusion. Clinical features were processed through a customized Multi-Layer Perceptron (MLP), while images were analyzed using a modified Residual Networks (ResNet) architecture. Model interpretability was enhanced using Gradient-weighted Class Activation Mapping (Grad-CAM) visualization to identify the contribution of clinical attributes to classification decisions.ResultsCross-attention fusion achieved the highest classification accuracy, demonstrating superior performance compared to unimodal approaches and simpler fusion techniques. The multimodal framework significantly outperformed image-only baselines, with cross-attention effectively capturing inter-modal dependencies and contextual relationships between visual and clinical data modalities.Discussion/ConclusionsOur findings demonstrate that integrating clinical metadata with dermatoscopic images substantially improves the accuracy of skin lesion classification. However, challenges, including class imbalance and the computational complexity of advanced fusion methods, require further investigation.

Background: Skin cancer diagnosis faces critical challenges due to the visual similarity of lesions and dataset limitations. Methods: This study introduces HybridSkinFormer, a robust deep learning model designed to classify skin lesions from both clinical and dermatoscopic images. The model employs a two-stage architecture: a multi-layer ConvNet for local feature extraction and a residual-learnable multi-head attention module for global context fusion. A novel activation function (StarPRelu) and Enhanced Focal Loss (EFLoss) address neuron death and class imbalance, respectively. Results: Evaluated on a hybrid dataset (37,483 images across nine classes), HybridSkinFormer achieved state-of-the-art performance with an overall accuracy of 94.2%, a macro precision of 91.1%, and a macro recall of 91.0%, outperforming nine CNN and ViT baselines. Conclusions: Its ability to handle multi-modality data and mitigate imbalance highlights its clinical utility for early cancer detection in resource-constrained settings.

Background. Today, two main approaches to interpreting dermatoscopic images are used: descriptive and metaphorical terminology. The lack of a unified dermatoscopic terminology in the Russian language brings risks of ambiguous interpretation of signs, difficulties in describing dermatoscopic status, challenges in follow-up monitoring, and limited comparability of scientific data. Aim. To assess the terminology and protocols used in dermatoscopy when providing medical care to patients with cutaneous and subcutaneous diseases, skin neoplasms, and cosmetic skin defects in order to improve the quality of medical care. Methods. A special questionnaire (52 questions) was developed. Interviews may be performed in a mixed format. An anonymous survey of 402 physicians was conducted. Respondents were asked to describe three dermatoscopic images. Results. The vast majority of respondents noted the following aspects: the need for specialized terminology in order to describe and understand the dermatoscopic status (95%); lack of previously acquired knowledge and skills to use this diagnostic method freely (73.4%); and need for in-depth investigation of terminological nuances (89.1%). Most respondents (93.5%) supported the idea of establishing a unified dermatoscopic terminology in Russian. The specialists concerned use an extremely wide range of terms to describe the dermatoscopic presentation. More than a half of the respondents (58.5 %) reported difficulties in interpreting the dermatoscopic presentation of skin neoplasms described by another specialist. Conclusion. The analysis revealed no unified standards for dermatoscopic terminology in Russian. The development of a standardized terminological system for dermatoscopy in the Russian Federation requires consensus conferences involving the specialists concerned, elaboration of proposed advanced training programs on dermatoscopy as well as relevant methodological recommendations, and their consistent implementation in advanced training programs for doctors of various specialties (within medical universities and postgraduate education system).

Modern computerized systems for the diagnosis of skin neoplasms are mainly focused on issuing recommendations to patients, but the application of designated systems in clinical practice remains limited. It is supposed that it is connected with the lack of qualitative researches of such systems and low trust of doctors to non-transparent mechanisms of their work. The creation of a medical decision support system based on the logic of the doctor's diagnostic search can solve this problem. An important task of the system is to recognize the color of globules of skin neoplasms, but the methods of solving the task have not yet been described in scientific publications. The application of the method of automated color recognition of globules on dermatoscopic images of skin neoplasms is considered, which allows recognizing globules by color in accordance with a palette of 7 colors (blue, yellow-white, brown, red, orange, nude, black). An original set of 9 color features has been developed as part of this method. The Random Forest method was applied to classify the images based on the feature (globule color). According to the results of the experiment conducted with a sample of 313 images, the classification accuracy was 91 %. The developed method can be implemented programmatically within the framework of a modified pattern analysis algorithm, and this method can also be used as part of a medical decision support system for the diagnosis of skin cancer.

In the human body, the skin is the main organ. Nearly 30–70% of individuals globally have skin-related health issues, for whom efficient and effective analysis is essential. A general method dermatologists use for analyzing skin illnesses is dermoscopy, which permits surveillance of the hidden structures of skin injuries, i.e., an area suffering from an illness whose effects are unseen to the naked eye. Dermoscopy is generally employed for cancers and other kinds of skin cancers with pigment. Yet, access to a dermoscopy is demanding in resource-poor areas and unnecessary for many general skin diseases. So, developing an effective skin disease analysis method that depends upon effortlessly accessible clinical imaging would be helpful and deliver lower-cost, common access to many individuals. Recently, computer-aided diagnosis (CAD) approaches have been effectively employed to detect skin cancers in dermatoscopic imaging. The CAD-based techniques will be beneficial for helping professionals detect and classify skin lesions. This paper presents an Advanced Skin Lesion Classification using Block-Scrambling-Based Encryption with a Fusion of Transfer Learning Models and a Hippopotamus Optimization (SLCBSBE-FTLHO) model. The main aim of the SLCBSBE-FTLHO model relies on automating the diagnostic procedures of skin lesions using optimal DL approaches. At first, the block-scrambling-based encryption (BSBE) technique is utilized in the image encryption pre-processing stage, and then the decryption process is performed. The feature extraction process employs the fusion of MobileNetV2, GoogLeNet, and AlexNet techniques. Furthermore, the conditional variational autoencoder (CVAE) method is implemented for skin lesion classification. To optimize the CVAE model performance, the hippopotamus optimization (HO) model is utilized for hyperparameter tuning to ensure that the optimum hyperparameters are chosen for enhanced accuracy. To exhibit the improved performance of the SLCBSBE-FTLHO approach, a comprehensive experimental analysis is conducted under the skin cancer ISIC dataset. The comparative study of the SLCBSBE-FTLHO approach portrayed a superior accuracy value of 99.48% over existing models.

The BRAAFF-Annotated Acral Lesions Dataset (BALD): A Curated Set of Dermatoscopic Images of Acral Melanoma and Nevi from Various Sources.

MILK10k: A Hierarchical Multimodal Imaging-Learning Toolkit for Diagnosing Pigmented and Nonpigmented Skin Cancer and its Simulators.

Skin diseases, particularly melanoma and other types of pigmented lesions, constitute a significant portion of global health concerns due to their prevalence and potential severity. In recent years, deep learning (DL) has revolutionized image classification tasks in the medical domain, particularly using Convolutional Neural Networks (CNNs) for skin lesion analysis. However, traditional CNNs are limited to capturing spatial features, often overlooking sequential patterns and complex contextual cues inherent in dermatological imagery. This study explores the automated classification of pigmented skin lesions using the HAM10000 dataset, a diverse collection of 10,015 dermatoscopic images spanning seven diagnostic categories. Addressing challenges in computational dermatology, we leverage MobileNet-V2 and InceptionV3 deep learning architectures, optimized via transfer learning and advanced preprocessing techniques. Comparative evaluation is performed between baseline CNN models and their Long Short-Term Memory (LSTM)-augmented variants to assess improvements in classification performance through sequential feature modeling. Results indicate that LSTM integration enhances contextual feature learning, improving accuracy for underrepresented lesion classes, with InceptionV3+LSTM achieving the highest classification accuracy.

As the incidence of skin cancer continues to rise globally, effective automated classification methods become crucial for early detection and timely intervention. Lightweight neural networks such as MobileNetV3 offer promising solutions due to their minimal parameters, making them suitable for environment with low resource. This study aims to develop an automated multiclass skin cancer classification system by enhancing MobileNetV3 with the Convolutional Block Attention Module (CBAM). The primary goal is to achieve high classification accuracy without significantly increasing computational demands. We employed Bayesian optimization to automatically fine-tune model parameters and applied targeted data augmentation techniques to address class imbalance. CBAM was integrated to highlight diagnostically relevant regions within images. The proposed method was evaluated using the ISIC 2024 SLICE-3D dataset, which includes over 400,000 dermatoscopic images categorized into benign, basal cell carcinoma, melanoma, and squamous cell carcinoma classes. Preprocessing involved standardized resizing, normalization, and extensive geometric and photometric augmentations. Results demonstrated that our method achieved an accuracy of 98.97%, precision of 98.99%, recall of 98.97%, and an F1-score of 98.98%, surpassing previous state-of-the-art models by 1.86–6.52%. Remarkably, this improvement was achieved with minimal additional parameters due to the effective integration of CBAM. These results represent an advancement in automated medical image analysis, particularly for low resource settings, by combining lightweight CNNs with attention mechanisms and systematic hyperparameter exploration.

Melanoma represent one of the most perilous types of skin cancer due to their rapid progression and the challenges associated with diagnosis. This research used the HAM10000 dataset to demonstrate Convolutional Neural Networks (CNNs), the most sophisticated deep learning model for classifying skin cancer lesions. During this investigation, we gathered 10,015 dermatoscopic images and classified them into seven separate kinds of skin lesions. The model performs feature extraction and classification hierarchically using fully connected, pooling, and convolutional layers. This endeavor has resulted in an impressive 98.57% training accuracy and 93.34% validation accuracy, representing a substantial improvement over the previously used approach. Essential performance metrics, such as accuracy, recall, and F1-score, demonstrate the model's efficacy in detecting different types of skin cancer. We obtained high accuracy, an F1 score, and sub-optimal recall. The evidence indicates that CNN-based approaches may facilitate early diagnosis, improve treatment results, and reduce dermatologists' workloads. This study's results contribute to the advancement of skin cancer research

Early detection of skin cancer, particularly melanoma, is crucial for improving patient outcomes. This study presents a comprehensive comparative analysis of deep learning architectures for automated skin lesion classification. This study evaluated eight models across four architectural categories: traditional convolutional neural networks, residual networks, vision transformers and novel hybrid approaches combining CNNs with transformers. Using the International Skin Imaging Collaboration dataset with 3,267 dermatoscopic images, the study assessed both classification performance and computational efficiency. Hybrid architectures consistently outperformed individual models, with the ResNet+ViT hybrid achieving the highest accuracy of 85.1% and AUC of 0.932. However, these performance improvements came with significant computational costs, highlighting important trade-offs for practical deployment. MobileNetV2 offered the best efficiency while maintaining a reasonable accuracy of 79.1%, making it suitable for resource-constrained environments. The findings demonstrate that combining complementary feature extraction mechanisms from CNNs and transformers creates more robust representations for skin lesion classification, approaching the diagnostic accuracy of dermatologists reported in the literature. This research contributes valuable insights for the development of automated diagnostic support tools that could improve early detection rates and enhance patient outcomes in skin cancer management.

ABSTRACT The Hospital Appointment System is a web-based solution designed to streamline the process of scheduling and managing appointments between patients and healthcare providers. Traditional appointment systems often face challenges such as long wait times, miscommunication, and inefficiency in managing patient records. This system addresses these issues by allowing patients to register, view available doctors, and book appointments online based on doctor availability. Doctors can manage their schedules and view patient histories through a secure portal. The platform includes features such as real-time notifications, user authentication, appointment rescheduling, and digital record keeping, which collectively enhance patient experience and reduce administrative workload. Developed using modern web technologies and backed by a robust database, the system ensures scalability and data security. The research emphasizes the role of digital transformation in healthcare accessibility and efficiency. Experimental results demonstrate significant improvements in appointment accuracy and patient satisfaction. Future improvements may include AI-based appointment suggestions and integration with electronic health record (EHR) systems. The Skin Disease Detection System is a deep learning-based application designed to assist in the early diagnosis of common skin conditions using image processing techniques. Skin diseases often go undiagnosed due to a lack of immediate medical access or patient awareness. This system leverages convolutional neural networks (CNNs) to analyze dermatoscopic images and classify them into multiple categories such as eczema, psoriasis, melanoma, and others. The model is trained on publicly available dermatology datasets and achieves high accuracy in identifying skin abnormalities. Users can upload images through a web or mobile interface and receive instant predictions along with risk levels. The system offers a low-cost, accessible alternative for preliminary diagnosis and can be a valuable tool in teledermatology, especially in rural or underserved areas. The research outlines model architecture, training methodology, and performance evaluation metrics. Experimental results indicate the system’s potential in supporting dermatologists and improving early detection rates. Future work will involve expanding the dataset, improving classification accuracy, and incorporating expert feedback for continuous learning.

Skin concerns are a rising health issue globally, and accurate detection and diagnosis of these issues are key ‎in preventing serious consequences. We provide a complete overview of deep learning approaches to ‎dermatoscopic image classification, specifically focusing on the newly developed Vision Transformer (ViT) ‎approaches. We discuss the advantages of using ViT approaches in skin disease classification versus prior deep ‎learning approaches, specifically, convolutional neural networks (CNNs). We analyse four commonly reported ‎skin conditions: Basal Cell Carcinoma, Benign Keratosis, Eczema, and Melanoma. In doing so, we explore the ‎current literature and datasets available and summarize the advancements of artificial intelligence (AI) in ‎dermatology, identify potentially the most effective designs, and consider their incorporation into clinical ‎populations. This review is intended to provide insight into the current developments in the design of AI-‎AI-assisted automated skin disease diagnosis processes, including important trends, performance, and efficiencies ‎of models, and the current trends in skin disease diagnosis. In an ideal world, this review will provide a ‎foundation for the development of more accurate and ultimately less expensive diagnostics to enhance patient ‎care in dermatology.

Recently, Denoising Diffusion Models have achieved outstanding success in generative image modeling and attracted significant attention in the computer vision community. Although a substantial amount of diffusion-based research has focused on generative tasks, few studies apply diffusion models to medical diagnosis. In this paper, we propose a diffusion-based network (named DiffMIC-v2) to address general medical image classification by eliminating unexpected noise and perturbations in image representations. To achieve this goal, we first devise an improved dual-conditional guidance strategy that conditions each diffusion step with multiple granularities to enhance step-wise regional attention. Furthermore, we design a novel Heterologous diffusion process that achieves efficient visual representation learning in the latent space. We evaluate the effectiveness of our DiffMIC-v2 on four medical classification tasks with different image modalities, including thoracic diseases classification on chest X-ray, placental maturity grading on ultrasound images, skin lesion classification using dermatoscopic images, and diabetic retinopathy grading using fundus images. Experimental results demonstrate that our DiffMIC-v2 outperforms state-of-the-art methods by a significant margin, which indicates the universality and effectiveness of the proposed model on multi-class and multi-label classification tasks. DiffMIC-v2 can use fewer iterations than our previous DiffMIC to obtain accurate estimations, and also achieves greater runtime efficiency with superior results. The code will be publicly available at https://github.com/scott-yjyang/DiffMICv2.