Desarrollo de una herramienta educativa móvil basada en inteligencia artificial para la enseñanza de la interpretación de radiografías de tórax en el diagnóstico de neumonía

Abstract

The objective of this project is to develop a mobile educational tool named NeumoLens, based on artificial intelligence, to strengthen the teaching of chest X-ray interpretation for pneumonia diagnosis. The system addresses the complexity of visual identification of radiological patterns through the use of Convolutional Neural Networks (CNN), providing an accessible diagnostic guide for students and non-experts. For the implementation of this system, the DenseNet121 architecture was employed and optimized using Transfer Learning and Fine-Tuning techniques to classify images into five categories: Normal, Bronchopneumonia, Lobar Pneumonia, Interstitial Pneumonia, and Noise. The model was natively integrated into a mobile application developed in Android Studio using the Kotlin language. Through the use of the TensorFlow Lite engine, the application operates under the Edge AI paradigm, allowing local inference and diagnosis on the device without the need for an internet connection, thus ensuring an immediate response and data privacy. Finally, the tool incorporates an interactive educational module that utilizes contrastive learning (Active Recall) and a gamified evaluation system (Quiz). The artificial intelligence model achieved an accuracy of 80% in the training environment and a 70% success rate in field tests with previously unseen images. The results demonstrate that NeumoLens is not only an accurate classifier but also an effective pedagogical support that facilitates the learning curve in the detection of lung pathologies through computer vision.