The fight against cancer is one of the great challenges facing humanity. Specifically, for those dedicated to cancer research, both in investigating and understanding its existence and in developing effective treatments and drugs to combat its effects on health.
In fact, according to the Spanish Society of Medical Oncology (SEOM), they have increased deaths by 17.9% in 2020 compared to 2019. It is significant that tumors were the second cause of death with 22.8% of them being, infectious diseases responsible for 16.4% of deaths. Thus concluding that, cancer, in the year of the coronavirus pandemic (SARS-CoV-2), continued to cause more deaths than the disease caused by that virus, COVID-19.
In addition, it was estimated that 18.1 million new cases were diagnosed globally in 2020 and this figure is expected to increase by 49.2% by 2040, rising to 27 million cases of cancer [1].
Fine-tunning of VGG16 model for benign or malign melanoma images.
VGG16 architecture example
The architecture of the developed model is based on the VGGNet model, specifically the VGG16 model. VGGNet, a specific type of convolutional architecture highly recommended for image classification. The technique described above has been applied as transfer learning. Therefore, the original architecture of VGG16 has been modified architecture of VGG16 has been modified to adapt it to the problem at hand.
Link to the UPM digital archive where you can find the complete project report with additional technical information. --> Use of a convolutional neural network for melanoma cancer detection - UPM Digital Archive
Repository to host the code and the memory of the final project of the degree in software engineering at the Higher Technical School of Computer Systems Engineering ETSISI of the Polytechnic University of Madrid UPM.