Early Skin Cancer Detection Using the SLICE-3D Dataset: A Transfer Learning Model and DCGAN Approach to Address Data Imbalance

Early detection of melanocytes can save lives from melanoma. Most individuals can’t be professionally diagnosed since it’s time-consuming, costly, and inconvenient. Smartphonebased early skin cancer diagnosis has emerged as a new approach. The existing computer-aided skin cancer diagnosis methods and mobile deep learning technology have been studied, and it is found that the existing smartphone-based skin cancer detection and identification methods rely on the support of background cloud services. Accuracy, reaction time, and patient data confidentiality are issues. A novel early detection and recognition model of melanoma skin cancer based on mobile deep learning, Melanlysis, is proposed. The model uses the EfficientNetLite-0 deep learning model to have low latency and considers the imbalance of the existing open-source skin image dataset. The proposed classification model is implemented and evaluated. Experimental results show that compared with the existing EfficientNetLite-0, MobileNet V2, and ResNet-50 models, the accuracy of correctly identifying malignant or non-melanoma is over 94%. At the same time, an Android application based on this mobile deep learning model was developed to diagnose potential malignant melanoma. Users can quickly obtain the classification results of melanoma through the application.

There are regional differences in skin cancer screening uptake in Germany. So far, it is unclear whether a high uptake of screening services leads to a reduction in mortality. This article presents study results on the investigation of spatiotemporal associations between skin cancer screening and mortality. The methods used are discussed regarding their suitability. The basis is ambulatory claims data on the utilization of early skin cancer detection as well as data on skin cancer mortality from the cause-of-death statistics of the years 2011-2015 at county level in Germany. In addition to a descriptive evaluation, spatiotemporal cluster analyses and regression models were used to investigate the relationship between the uptake of early detection and mortality. In addition to age, adjustments were also made for other selected socio-economic and socio-graphical variables. The descriptive results show striking spatial patterns of skin cancer screening and mortality. Cluster analyses identified regions with significantly higher and lower cases of early detection and skin cancer mortality. The spatiotemporal regression analyses show no clear association. Only early detection by a dermatologist, adjusted for age, shows an association with mortality. No clear association between early skin cancer detection and mortality can be derived from the results. However, the study design used with a spatiotemporal cluster and regression analysis has shown that these methods allow in-depth statements about the relationship between early skin cancer detection and mortality.

Cancer is a significant health problem due to the increase in patients worldwide and causing death. Recently, it has been seen that skin cancer is among the most common types of cancer. Prolonging the survival of individuals struggling with skin cancer and reducing treatment costs is possible with early diagnosis. However, techniques used for early diagnosis in today's health systems have limitations such as requiring extensive human resources, long-term results, and not having easy access to these services for everyone. For this reason, systems that are easy to apply, produce accurate results in the context of scientific methods and are accessible to everyone are needed for early skin cancer detection. Early detection of skin cancer is possible with the use of artificial intelligence techniques. This study aims to classify benign and malignant skin images and inform users of the results via mobile application. CNN, KNN and Decision Tree algorithms were used to classify the images. “Experiments ISIC: Skin Cancer: Malignant vs. It was carried out by applying the augmentation technique on the “Benign” data set. As a result of the experiments, the most successful results were obtained with the Transfer Learning algorithm, 94.89%. The study also compared the data and results obtained with other architectures. Experimental results show that it is possible to detect skin cancer early with artificial intelligence techniques and to notify the user of the results with a mobile application. We believe the study's results will shed light on new research for early skin cancer detection.

Skin cancer is the most common cancer in the United States, with about 1,000,000 people developing the disease each year. The incidence of melanoma has rapidly increased in young white women between the ages of 15-34 over the last three decades. While melanoma accounts for approximately 3% of skin cancers, it causes more than 75% of skin cancer deaths. Thorough skin assessments and awareness of new or changing appearance of skin lesions are critical to early detection and treatment of skin cancer, as well as teaching sun-protective behaviors. Educators created a novel approach to "bring to life" skin cancer assessment skills to promote awareness of prevention and early detection of skin cancer using moulage in a human patient simulation lab. Through the use of moulage-like lesions, simulated patients were evaluated and taught skin cancer prevention and early detection principles by baccalaureate nursing students. The average age of study participants (n = 104) was 26.50years. The majority of responders were female. At the end of the lab, students' average responses to an evaluation based on program goals were very positive. Anecdotal comments affirmed positive student perceptions of their simulation experience. Data analyses indicated item means were consistently higher for upper-division students. The age and gender of students who participated in this study align with the NCI statistics on age and gender of the population with increased incidence of melanoma.

The early and accurate detection of skin cancer can reduce mortality rates and improve patient outcomes, but requires advanced diagnostics. The integration of artificial intelligence (AI) into healthcare enables the precise and timely detection of skin cancer. However, significant challenges remain including the difficulty in differentiating visually similar skin conditions and the limitations of diverse, representative datasets. In this study, we proposed DCAN-Net, a novel deep-learning framework designed for the early detection of skin cancer. The model leverages an efficient backbone architecture optimized for capturing diverse skin patterns, utilizing carefully tuned parameters to enhance the discrimination capabilities and refine the extracted features using modified attention modules, thereby prioritizing relevant foreground information while minimizing background noise. Furthermore, the Grad-CAM explainable AI method was employed, highlighting the most salient features within dermatoscopic images. The fused optimal feature representations significantly enhanced the dermatoscopic image analysis. When evaluated on the HAM10000 dataset, DCAN-Net achieved a precision, recall, F1-score, and accuracy of 97.00%, 97.57%, 97.10%, and 97.57%, respectively. Moreover, the application of advanced data augmentation techniques mitigated data imbalance issues and reduced false-positive and false-negative rates across the original and augmented datasets. These findings demonstrate the potential of DCAN-Net for improving clinical outcomes and advancing AI-driven skin cancer diagnostics.

Age standardised skin cancers (melanoma and non-melanoma) continue to increase in Australia, although they are stabilising for those under age 40. People living in regional, rural and remote Australia have higher rates of skin cancer and challenges accessing care. Better targeting of skin protection measures and improved opportunistic screening have been promoted as ways to improve care for these populations as have increased use of information technology and upskilling of primary health care nurses. The Australian Government supports that Optimal Cancer Care Pathways for skin cancers be used as a key resource in exploring best practice models of care for skin cancer patients both for the development of digital platforms and face to face multidisciplinary teams (MDTs). Better use of technology has been a core recommendation of national health and skin cancer organisations for improving prevention and early detection of skin cancer. Skin cancers, as a primarily visual diagnosis are considered one of the prime areas for technological health interventions. The harnessing of artificial intelligence (AI) technology as a tool for early detection and disease management of skin cancers has great potential to reduce the burden of health care costs to the regional, rural and remote community and improve health outcomes.

Skin cancer remains one of the most common and serious global health problems, with cases continuing to increase annually. Early and accurate detection is essential for improving patient survival; however, conventional diagnostic methods often depend on manual visual assessment, which can be subjective and inconsistent. Hence, the development of an automated and reliable detection system is vital to support healthcare professionals in early diagnosis. This study proposes an intelligent diagnostic model for early skin cancer detection using dermatoscopic images, integrating transfer learning with Convolutional Neural Network (CNN) techniques. The model employs the HAM10000 dataset from the International Skin Imaging Collaboration (ISIC), which contains high-resolution dermatoscopic images classified into three malignant skin cancer types: Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC), and Malignant Melanoma (MM). The CNN framework was built using pre-trained models optimized to enhance classification accuracy. Experimental results showed that the model achieved an accuracy of 96.67% and an F1-score of 0.97, demonstrating strong capability in identifying multiple malignant lesions. These findings indicate that the model can assist dermatologists and clinicians in improving diagnostic precision and reducing examination time in clinical practice. In conclusion, integrating transfer learning within a CNN architecture significantly improves classification efficiency even with limited data, and with further validation, the model shows strong potential for real-world implementation as an accurate, efficient, and accessible computer-aided diagnostic tool for early skin cancer detection.

Terahertz imaging offers significant potential for the early detection of skin cancer. This study introduces a metamaterial unit cell designed to operate in the terahertz (THz) band for non-invasive contact-based skin cancer detection. The sensor relies exclusively on the reflection coefficient response, providing high sensitivity to subtle changes in tissue properties without requiring complex signal processing. This simplicity may result in a cost-effective and straightforward implementation for early cancer detection.Simulations were conducted using 3D models representing various skin types, including normal skin, basal cell carcinoma (BCC), and melanoma. The dielectric characteristics of the samples were determined using the Double Debye model. The simulations revealed that the metamaterial design exhibited double negative material properties at a specific frequency of 1.15 THz. Upon skin contact and detection of malignancy, the reflection coefficient showed a shift toward lower frequencies. Notably, the melanoma sample exhibited the most significant shift, indicating a more severe form of cancer compared to BCC. Furthermore, it was observed that the difference in resonance frequencies between normal and malignant skin increased with the thickness of the sample. The sensor demonstrated high sensitivity in detecting cancer thickness, with a sensitivity of 9.25 GHz/μm for basal cell carcinoma (BCC) and 10.2 GHz/μm for melanoma. Furthermore, linear regression analysis revealed a robust correlation between the resonance frequency and the variation in cancer thickness, with R2 values of 0.9948 and 0.9947 for BCC and melanoma, respectively. These findings underscore the sensor's ability to detect skin cancer at its earliest stages, regardless of its severity.

The unrepaired DNA in skin cells causes skin cancer disease which leads to mutations or genetic defects in the skin. This disease will be widely spread in other parts of the body which can be cured in the initial stages. So, the early deterrence of this disease is a vital factor. This skin cancer has increased the mortality rate, and the treatment is highly expensive. Researchers have undergone several techniques for skin lesion detection based upon parameters of skin such as symmetry, color, size, and shape. These parameters are useful for distinguishing non-melanoma cancer from melanoma cancer. This paper gives a detailed study of the early detection of skin cancer with the aid of image processing, machine learning, and deep learning techniques. KeywordsSkin cancer detectionConvolutional neural networksDeep learningTransfer learning

Melanoma is the most lethal of the three primary skin cancers, including also basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), which are less lethal. The accepted diagnosis process involves manually observing a suspicious lesion through a Dermascope (i.e., a magnifying glass), followed by a biopsy. This process relies on the skill and the experience of a dermatologist. However, to the best of our knowledge, there is no accepted automatic, noninvasive, and rapid method for the early detection of the three types of skin cancer, distinguishing between them and noncancerous lesions, and identifying each of them. It is our aim to develop such a system. We developed a fiber-optic evanescent wave spectroscopy (FEWS) system based on middle infrared (mid-IR) transmitting AgClBr fibers and a Fourier-transform infrared spectrometer (FTIR). We used the system to perform mid-IR spectral measurements on suspicious lesions in 90 patients, before biopsy, in situ, and in real time. The lesions were then biopsied and sent for pathology. The spectra were analyzed and the differences between pathological and healthy tissues were found and correlated. Five of the lesions measured were identified as melanomas, seven as BCC, and three as SCC. Using mathematical analyses of the spectra of these lesions we were able to tell that all were skin cancers and we found specific and easily identifiable differences between them. This FEWS method lends itself to rapid, automatic and noninvasive early detection and characterization of skin cancers. It will be easily implemented in community clinics and has the potential to greatly simplify the diagnosis process.

Artificial intelligence (AI) has rapidly emerged as a transformative technology in dermatology, offering new possibilities for the early detection of skin cancer. Over the past decade, advances in deep learning, multimodal imaging, and explainable AI have achieved diagnostic performance comparable to, and in some cases exceeding, that of expert dermatologists. This review analyzes scientific evidence published between 2017 and 2025, focusing on diagnostic accuracy, algorithmic diversity, ethical implications, and regional adoption in Latin America, with particular emphasis on Mexico, Colombia, and Ecuador. Using the DMAIC methodological framework (Define–Measure–Analyze–Improve–Control), twenty peer-reviewed studies from high-impact journals such as Nature, The Lancet Digital Health, and npj Digital Medicine were examined through descriptive and comparative analysis. Results demonstrate that AI-based systems exhibit high diagnostic sensitivity (93%), specificity (89%), and accuracy (91%), outperforming dermatologists in controlled settings. Convolutional neural networks (CNNs) remain the dominant architecture, though hybrid, multimodal, and explainable models are gaining clinical relevance. A persistent dataset bias was identified, with light skin tones (I–II) representing 62% of images and darker tones (IV–VI) only 15%, revealing a structural inequity that limits global applicability. Regional initiatives in Mexico, Colombia, and Ecuador are addressing these disparities through the creation of diverse datasets, pilot projects in teledermatology, and ethically guided digital health strategies. The evidence confirms that artificial intelligence, when applied responsibly, has the potential to enhance diagnostic precision, promote healthcare equity, and transform dermatologic practice through human–machine collaboration. AI should not replace clinical expertise but rather expand it—turning early skin cancer detection into a more accurate, inclusive, and globally equitable discipline.

Narrative review. Building on the discussions from the Melanoma Screening Summit held in Brisbane, Australia, in 2019, we reviewed evidence related to current approaches and new opportunities for early detection of melanoma and other skin cancers. Population-based melanoma screening is not currently recommended due to insufficient evidence that screening reduces melanoma mortality. Instead, in most countries including Australia, early detection of melanoma and keratinocyte carcinomas is undertaken opportunistically, by either the patient presenting for a routine skin check or with a lesion of concern, or by the doctor detecting a lesion incidentally. Several concerns about the current unstructured approach to skin cancer early detection have been identified, including variable quality of care, sociodemographic inequalities in medical access and health outcomes, excision of many benign lesions, overdiagnosis, gaps in workforce training, and health system inefficiencies. There has also been renewed interest in melanoma screening in Australia, driven by a changing landscape of skin cancer early detection. These changes include increasing health system costs for adjuvant therapies, advances in diagnostic technologies and artificial intelligence, the availability of validated risk-stratification tools, and consumer-driven digital technologies. The future of skin cancer early detection in Australia and internationally may incorporate features such as a more structured approach to skin cancer risk assessment using online risk calculators and invitations to screen, consumer-driven melanoma surveillance, and new technologies for diagnosis and monitoring of lesions. High-quality research evidence is being generated across multiple research programs, and is essential to underpin any changes to policy and practice in skin cancer early detection.

the early detection and successful treatment of skin cancers, a potent form of cancer, calls for the use of sophisticated diagnostic instruments. This study delves into the use of support vector machines (SVMs), to cope with the inconsistencies occurring among skin lesions, by merging them with feature fusion techniques. SVMs are preferred for this situation, as they are highly effective when it comes to the management of exceedingly dimensional data. Initially, in order to train and enhance the diagnostic capacity of the SVM classifier, a single and all-inclusive single dataset was generated through the analysis, identification and extraction of a wide variety of explanatory features (including colour, texture and shape) from a dataset comprising 10000 dermatoscope skin lesion representations. This was followed by the use of early and late fusion approaches, to generate an extensive dataset of descriptions, for assessing the reliability of the SVM classifier. Finally, the accuracy, precision and recall of the SVM classifier were ascertained by way of an objective dataset, comprising 25 dermatoscope representations of malignant and benign lesions. The accuracy, precision and recall of the SVM classifier are supported by its capacity to distinguish 10 true positives, 12 true negatives, three false positives and zero false negatives. As such, the SVM classifier can be considered effective, for the early detection of skin cancers. The results from this investigation verify that the capacity of SVMs, in terms of skin cancer diagnosis, is greatly improved with the utilization of feature fusion techniques. Also verified through this undertaking, is the effectiveness of innovative computational procedures, for the delivery of dependable medical diagnoses.

PurposeThe mortality rate due to skin cancers has been increasing over the past decades. Early detection and treatment of skin cancers can save lives. However, due to visual resemblance of normal skin and lesion and blurred lesion borders, skin cancer diagnosis has become a challenging task even for skilled dermatologists. Hence, the purpose of this study is to present an image-based automatic approach for multiclass skin lesion classification and compare the performance of various models.Design/methodology/approachIn this paper, the authors have presented a multiclass skin lesion classification approach based on transfer learning of deep convolutional neural network. The following pre-trained models have been used: VGG16, VGG19, ResNet50, ResNet101, ResNet152, Xception, MobileNet and compared their performances on skin cancer classification.FindingsThe experiments have been performed on HAM10000 dataset, which contains 10,015 dermoscopic images of seven skin lesion classes. The categorical accuracy of 83.69%, Top2 accuracy of 91.48% and Top3 accuracy of 96.19% has been obtained.Originality/valueEarly detection and treatment of skin cancer can save millions of lives. This work demonstrates that the transfer learning can be an effective way to classify skin cancer images, providing adequate performance with less computational complexity.

Skin cancer, particularly malignant melanoma, poses a significant health threat, underscoring the critical need for accurate and timely detection. This project focuses on developing a skin cancer detection system using Convolutional Neural Networks (CNNs), specifically tailored to differentiate between benign and malignant melanomas. The utilization of artificial intelligence in medical image analysis aims to enhance early diagnosis, reduce manual examination reliance, and contribute to improved patient outcomes. By focusing on the differentiationbetween benign and malignant melanomas, theproject aimsto contributetopersonalized treatment plans, early intervention strategies, and improved prognostic outcomes for patients. Skin cancer is one of the most prevalent types of cancer worldwide, with early detection being crucial for effective treatment. we propose a comprehensive approach for the automated detection of skin cancer lesions leveraging image processing techniques and Convolutional Neural Network (CNN) classification. The proposed methodology comprises several stages. Firstly, preprocessing techniques are applied to enhance the quality of input images into RBG. Subsequently,Gray LevelCooccurrence Matrix (GLCM) features are extracted from the preprocessed images and converted into binary representations to capture texture information effectively. These binary GLCM features are then fed into a CNN-based classification algorithm. The CNN model is trained on a large dataset of annotated skin lesion images, allowing it to learn discriminative features indicative of malignant or benign characteristics. The trained CNN model is capable of classifying unseen skin lesion images accurately. The proposed approach offers several advantages, including automation, which reduces the dependence on manual inspection by dermatologists, thereby potentially increasing the efficiency and accuracy of skin cancer diagnosis. Moreover, by integrating both preprocessing techniques and advanced classification algorithms, the proposed system demonstrates robustness and effectiveness in detecting skin cancer lesions across diverse datasets. Experimental results on benchmark datasets demonstrate the efficacy of the proposed approach, achieving high accuracy rates in distinguishingbetween malignantandbenign skinlesions. The proposed methodologyholdspromise for aidinghealthcare professionals in early skin cancer detection, ultimately improving patient outcomes and reducing the burden on healthcare systems.