AI-Based Medical Assistance via Mobile Devices and Enhanced CNN for Diagnostic Imaging
- Autores: Ali Yousuf Khan
- Directores de la Tesis: Miguel Angel Luque Nieto (dir. tes.), Pablo Otero Roth (dir. tes.)
- Lectura: En la Universidad de Málaga ( España ) en 2025
- Idioma: inglés
- Tribunal Calificador de la Tesis: Alfonso Ariza Quintana (presid.), Sandra Sendra Compte (secret.), Muhammad Yousuf Irfan Zia (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería de Telecomunicación por la Universidad de Málaga
- Materias:
- Enlaces
- Tesis en acceso abierto en: RIUMA
- Resumen
The COVID-19 pandemic underscored critical flaws in global healthcare diagnostics and emergency response systems, revealing the urgent need for practical solutions. AI-driven technologies can function under real-world pressures, and this dissertation responds to those challenges through three applied contributions that incorporate artificial intelligence to enhance medical diagnosis of diseases (COVID-19 and brain tumors) by image classification, and crisis communication in both routine and emergency settings.
The first work of research, discussed in Chapter 3, involves the creation of an Android-based mobile application named Help Pro: SOS Application. Using a variety of programming applications such as Android Studio, Java, Firebase, Google Maps API and Laravel, this tool allows users to send essential information including their geographic location, identification, and the nature of the emergency (vehicle mechanical problem, ambulance, blood donation, police or firefighter) to preselected contacts or responders with a single touch. The app is further optimized to operate effectively in limited connectivity environments, with testing focused on usability, response time, and system stability. The major goal is to minimize delays during emergency situations and to support timely response and intervention on site.
Chapter 4 presents the second contribution, a deep learning model called CXR-DNN, which was built using the EfficientNetB7 architecture to detect COVID-19 in chest X-ray images. The model underwent training and validation on several datasets (e.g. COVID-19 Radiography Database from Kaggle, or Actualmed COVID-19 Chest X-ray Dataset Initiative), followed by performance evaluation using essential metrics such as accuracy, precision, recall, and F1-score. Also included are suggestions for enhancing the fairness and robustness of models, as well as issues like class imbalance, hardware restrictions, and the openness of AI decision-making.
The third and final contribution, presented in Chapter 5, introduces a CNN-based joint diagnostic framework for classifying brain tumors using CT and MR imaging. The system uses MobileNetV3 to assess a six-category dataset, four of them using MR images (tumors: Glioma, Meningioma, Pituitary, and No Tumor) and other two classes considering CT images (tumor, non-tumor). The experimental results suggest that this strategy could be valuable in neuro-oncology, where the combination of the two imaging modalities significantly enhances classification accuracy.
Chapter 6 concludes the dissertation by summarizing the results and proposing directions for further research, including expanding the diagnostic tools to detect additional illnesses and improving the functionality of the SOS application. This work combines artificial intelligence with healthcare and emergency management to deliver practical solutions capable of making a meaningful impact in real-world scenarios.
