Convolutional Neural Networks for the Effectiveness of the First HPV Diagnosis in Piura 2025

Abstract

This research involves the development of a mobile application that integrates the MobileNet V2 model to detect lesions caused by the Human Papillomavirus (HPV), contributing to SDG 9, which focuses on building resilient infrastructure, promoting sustainable industrialization, and fostering innovation. The objective was to evaluate the effectiveness of a convolutional neural network (CNN) as a support tool in the initial diagnostic phase. A mixed-method approach with an experimental design was used to measure the effectiveness of a pre-trained CNN for the initial diagnosis of HPV. The MobileNet V2 model was adapted using transfer learning and optimized with regularization techniques such as dropout and fine-tuning. Effectiveness was determined using performance metrics: accuracy (98.63%), sensitivity (100%), specificity (97.33%), PPV (97.26%), NPV (100%), and F1-Score (98.60%). Training was crucial, as the use of regularization strategies allowed the model to extract relevant features for lesion identification. Although the testing period was short, the results were promising in a simulated clinical setting. In conclusion, the CNN proved to be a valid diagnostic support tool, paving the way for the use of artificial intelligence in early disease detection.