Medical picture categorization has been greatly enhanced by the use of deep learning, especially in the timely identification of skin lesions. Still, predictions from a single model remain unreliable due to their susceptibility to variations in the dataset, complicating their application to diverse clinical scenarios. This study introduces a unique CNN, PHMBCNN, designed to enhance classification accuracy using a progressive learning strategy. We propose the R3MV three-tier decision fusion system, which integrates predictions from (i) individual CNN model predictions, (ii) a feature fusion classification architecture, and (iii) a meta-classifier trained on the outputs of the CNN models. The final forecast is reached using a majority voting procedure, which enhances the reliability of the decision. The study utilizes two datasets for skin cancer: PAD_UFES_20 and HAM10000. Incorporating a GRU into the PHMBCNN model results in the PHMBCNN-GRU. The classification accuracy enhanced from 75.70% for the PAD_UFES_20 dataset to 80.69%, and from 92.07% for the HAM10000 dataset to 96.01%. The R3MV system design achieves 81.78% for PAD_UFES_20 and 99.3% for HAM10000.

Skin cancer is widely acknowledged as a particularly dangerous type of cancer, and there has been a large increase in death rates due to a lack of awareness about its symptoms and preventive measures. Hence, it is crucial to identify cancer in its initial phase to minimize its advancement. The shortage of competent physicians, insufficient medical equipment, and arduous diagnostic procedures for identifying malignant skin lesions pose a severe obstacle to timely life-saving interventions. The utilization of image processing and deep learning models can effectively address these challenges by accurately identifying the malignancy of skin cancer lesions. Integrating an optimizer into the deep learning classifier model will enhance the extraction of pertinent features from visual images by dynamically modifying the model’s hyperparameters. An optimizer enhances the acquisition of features during the training stage of the deep learning model in order to achieve the global optimum more quickly. Moreover, this modification will enhance the training accuracy and improve other performance indicators, hence facilitating the evaluation of the model. This study compares the performances of different optimizers, such as Teaching-Learning-Based Optimizer (TLBO), Grey Wolf Optimizer (GWO), Dragonfly Algorithm (DA), and Wildebeest Herd Optimizer (WHO), when combined with various deep learning models like GoogLeNet and ResNet50. The objective is to enhance the exploration of parameter space and achieve global optima for a classification problem. The deep learning model and optimizer were applied to the PAD_UFES_20 and HAM10000 datasets, which are well-known datasets for skin cancer. The deep learning model, in conjunction with the optimizer, has been assessed using particular criteria like precision, recall, F1-score, and accuracy.

Dermatological conditions pose a substantial worldwide health risk, exerting a significant negative influence on the overall physical and mental well-being of individuals. The potential exists to infer an individual’s susceptibility to dermatological conditions by considering their skin pigmentation. Skin features, such as variations in colour, are commonly assessed using the Fitzpatrick skin type scale. However, it is observed that these features are frequently lacking in adequate representation within publicly available dermatological databases. The field of automated visual evaluation of skin has experienced significant progress due to recent advancements in computer vision and artificial intelligence techniques. This study presents a novel approach, namely the Optimal Image Pattern to Identify Fitzpatrick Skin Tone (OIPFST), for accurately determining the Fitzpatrick skin tone using Individual Typology Angle(ITA) of skin lesion images. The ITA is determined by analysing four image patterns: Whole, Healthy, Quadrant, and Octant. These patterns are utilised to ascertain the Fitzpatrick skin tone of a patient based on their skin lesion images. This study utilises skin lesion images from the PAD_UFES_20 dataset. The patterns efficiency is assessed by considering metrics such as accuracy, F1 score, root mean square error (RMSE), and mean absolute error (MAE). The Octant pattern image proposed in this study demonstrates superior accuracy, as indicated by the highest f1_score and the lowest RMSE and MAE values of 71.07%, 68.89, 0.72, 0.35 respectively, when compared to other state-of-the-art methods. The proposed pattern has the potential to improve the accuracy of annotating Fitzpatrick skin tone using skin lesion images. This could be beneficial for various skin disease treatments and surgeries, as it would reduce errors and increase precision.

INTRODUCTION: Skin cancer is emerging as a significant health risk. Melanoma, a perilous kind of skin cancer, prominently manifests asymmetry in its morphological characteristics. OBJECTIVE: The objective of the study is to classify the asymmetry of the skin lesion shape accurately and to find the number of symmetric lines and the angles of formation of symmetric lines. METHOD: This study introduces a unique methodology known as Shape Asymmetry Analysis (SAA). The SAA incorporates a comprehensive framework including image pre-processing, segmentation along with the computation of mean deviation error and the subsequent categorization of data into symmetric and asymmetric forms using a classification model. RESULT: The PH2 dataset is used in this study, where the three labels are consolidated into two categories. Specifically, the labels "symmetric" and "symmetric with one axis" are merged and classified as "symmetric," while the label "asymmetric" is unchanged and classified as "asymmetric". The model demonstrates superior performance compared to conventional methodologies, achieving a noteworthy accuracy rate of 90%. Additionally, it exhibits a weighted F1-score, precision, and recall of 0.89,0.91,0.90 respectively. CONCLUSION: The SAA model accurately classifies skin lesion shapes compared to state-of-the-art methods. The model can be applied to the shapes, irrespective of irregularity, to find symmetric lines and angles.

A major public health issue on a global scale is skin cancer. During the early stages, a small percentage of skin malignancies resemble typical moles but eventually develop into deadly cancer. Dermatologist may occasionally need a biopsy to diagnose the issue, which is a time-consuming and painful treatment. Using cutting-edge deep learning and CNN models can detect and categorise skin malignancies using skin lesion pictures, which solves the problem of most rural areas of the world being inaccessible to skin experts to do biopsy. In-depth analysis of more modern, accurate methods for early skin cancer detection is provided in this publication.