Determinants of Micro-entreprise Competitiveness: The Case of Northern Mindanao, Philippines

Abstract

Respiratory disorders, including pneumonia, COVID-19, and lung cancer, continue to pose significant challenges to global healthcare systems due to their high prevalence, mortality rates, and overlapping radiological manifestations. Accurate differentiation among these conditions remains difficult because conventional diagnostic procedures based on chest X-ray (CXR) and computed tomography (CT) imaging depend extensively on expert interpretation, which may lead to inter-observer variability, delayed diagnosis, and reduced detection sensitivity, particularly during the early stages of disease progression. These challenges are further amplified in resource-limited clinical environments where access to specialized expertise is restricted. To overcome these limitations, this study presents a novel Multi-Objective Optimized Hybrid Deep Learning (MUOPTDL) framework for automated respiratory disease classification using medical imaging modalities. While pneumonia serves as the primary target disease, additional categories, including COVID-19, lung cancer, and normal cases, are incorporated to enhance diagnostic discrimination and minimize classification errors caused by overlapping imaging characteristics. The proposed MUOPTDL framework combines three complementary deep convolutional neural network architectures, namely DenseNet121, MobileNetV3, and Xception, to extract robust and diverse feature representations from chest CT and X-ray images. The framework employs transfer learning, data augmentation, dropout-based regularization, and adaptive optimization techniques to improve model generalization and reduce overfitting. Experimental results demonstrate that the proposed weighted hybrid ensemble significantly outperforms individual baseline models, achieving an overall accuracy, precision, recall, and F1-score of 98%, along with an area under the receiver operating characteristic curve (AUC) of 0.9866. The developed framework effectively differentiates among clinically similar radiographic patterns, including COVID-19-associated ground-glass opacities, pneumonia-related consolidations, and early-stage pulmonary malignancies. Furthermore, implementation through a web-based clinical interface highlights the practical applicability of the proposed system for real-time screening and computer-aided diagnostic decision support in healthcare environments.