A Cloud-Based Real-Time Skin Cancer Detection System Utilizing Artificial Intelligence

Abstract

One of the most serious types of cancer that can affect people is skin cancer. Skin cancer can be cured with early
discovery, and the patient's life may be saved with the appropriate care. Skin cancer disorders come in a variety
of forms, each with unique characteristics. Doctors utilize the ABCDET methodology, which is a classic
method, to detect skin lesions. Nevertheless, with the number of cases of skin cancer rising worldwide in the
present period, manual diagnosis of skin lesions is failing. To detect skin lesions more quickly and with fewer
diagnostic errors, doctors' workloads can be reduced by using automatic skin lesion detection. Combining
various deep learning and machine learning technologies can result in the development of an intelligent system
that can correctly diagnose skin lesions.One type of deep learning model used to extract and classify skin lesion
data is the neural network model. This work contains a real-time skin cancer detection simulation and compares
CNN and random-forest classifiers for skin tumour categorization. The controversy centres on the HAM10000
the dataset, which contains pictures of seven distinct kinds of skin diseases. Following picture preparation for
denoising and artifact removal, picture segmentation using Active Contours Without Edges (ACWE) and
extraction of features using the ABCDT technique are performed. Next, a textural analysis is carried out using
the Gray Level A combination Matrix (GLCM) and Fractal Dimensional Texture Analysis (FDTA).CNN's
classification accuracy is 91.97%, whereas Random Forest's classification accuracy is 89.82%. The CNN model
performed better than the Random Forest model for classification when the models that were trained were used
in a simulation in real time to identify skin cancer.