Skin Guard: Intelligent Detection and Management of Skin Cancer

Skin cancer continues to be one of the most prevalent types of cancer worldwide, with a steadily increasing number of cases reported annually. It is generally classified into two major types: benign lesions, which are harmless, and malignant lesions, which can be lifethreatening if not diagnosed and treated promptly. Early detection is critical, as it significantly improves the chances of successful treatment and can greatly reduce the burden on healthcare systems. However, traditional diagnostic methods such as visual inspection and biopsy often depend heavily on clinical expertise and may not be readily accessible in all settings. This highlights the need for efficient, automated systems that can aid in the early identification of skin cancer. Techniques such as dropout regularization and learning rate scheduling were also explored to further optimize the model. The findings suggest that with appropriate tuning and preprocessing, ANN models can serve as valuable tools in assisting dermatologists and healthcare providers in the early and reliable diagnosis of skin cancer. In conclusion, this work demonstrates the practical application of artificial neural networks in skin lesion classification and reinforces the growing role of AI in medical diagnostics. In response to this need, this project explores the use of Artificial Neural Networks (ANNs) for the classification of skin lesions using image- based data. The study leverages a publicly available Kaggle dataset composed of high-resolution dermoscopic images labelled as either benign or malignant. The primary aim is to evaluate the effectiveness of ANN models in classifying skin lesions by learning complex patterns from the visual data. A series of preprocessing steps were applied to the dataset to enhance model performance. These include image normalization, resizing, and augmentation techniques such as rotation, flipping, and contrast adjustment, which help in increasing data diversity and preventing overfitting. Feature extraction was performed directly from the pixel data, allowing the neural network to learn hierarchical features during training. The ANN architecture was designed and optimized by experimenting with different numbers of hidden layers, neurons, activation functions (such as ReLU and sigmoid), and optimizers (including Adam and SGD)The model was trained and validated using various evaluation metrics, including accuracy, precision, recall, F1-score, and a confusion matrix to assess its predictive capabilities. The results indicate that the ANN model is effective in distinguishing between benign and malignant lesions, achieving a high level of accuracy on the validation set. This confirms the potential of ANNs in medical image classification tasks, especially in dermatology. Additionally, the projectinvestigate how changes in model parameters and training strategies impactperformance. The model developed in this study provides a foundation for future research, including the use of more complex architectures like Convolutional Neural Networks (CNNs), integration with real-time diagnostic systems, and deployment in clinical environments.