Mobile-Net Based Pneumonia Detection Model

Abstract

Pneumonia, a leading global cause of mortality, requires timely and accurate diagnosis to mitigate its impact. This work introduces an automated pneumonia detection system using MobileNet, a lightweight Convolutional Neural Network (CNN) optimized for mobile and embedded devices. By employing depth wise separable convolutions, Mobile Net reduces computational complexity while maintaining high accuracy. Trained on a labelled chest X-ray dataset with preprocessing techniques, it achieves 92.5% accuracy, 91.8% precision, and 93.2% recall, offering an efficient, accessible solution for resource-constrained healthcare settings.

Introduction

Pneumonia is a major global health issue, especially in low-resource areas where diagnosis is limited by lack of expert radiologists and diagnostic tools. Chest X-rays are standard for diagnosis, but interpreting them requires specialized skills often unavailable in underserved regions, causing delays and errors.

To address this, the paper proposes an automated pneumonia detection system using MobileNet—a lightweight convolutional neural network optimized for mobile devices. By leveraging transfer learning on a labeled chest X-ray dataset, the system achieves efficient and accurate pneumonia detection suitable for resource-constrained environments. The approach includes data preprocessing, augmentation, and uses a multi-class Support Vector Machine (SVM) for final classification to handle multiple pneumonia types.

The paper reviews recent deep learning models for pneumonia detection, including DenseNet-121, EfficientNetV2L, lightweight residual networks, and ensemble models, highlighting their strengths and limitations in terms of accuracy, computational demands, and interpretability.

The proposed hybrid model combines MobileNet and DenseNet features, reduces dimensionality with PCA, and classifies using multi-class SVM. Data augmentation and transfer learning improve generalization. For interpretability, Grad-CAM is integrated to visualize decision-influencing regions on X-rays.

Evaluation on benchmark datasets shows high accuracy (around 98.7%), precision, recall, and F1 scores, demonstrating the model’s robustness. The system is optimized for real-time deployment on mobile and embedded devices, aiming to improve timely pneumonia diagnosis and healthcare access in underserved regions.

Conclusion

This study introduces a hybrid deep learning model for pneumonia detection using MobileNet, a lightweight convolutional neural network (CNN) architecture.Essentialdataforpneumonia detection are processed by deep learning algorithms to extract key features. These extracted features are then concatenated and their dimensionality reduced through principal component analysis (PCA). Finally, the features are classified using a multiclasssupportvectormachine(SVM) to achieve improved performance in detection. Future enhancements to the proposed hybrid deep learning model could further improve its accuracy and utility. One promising avenue is the integration of additional medical imaging modalities,suchasCTscans,toprovidea more comprehensive understanding of pneumonia and enhance diagnostic capabilities.

References

[1] Singh, A., Gupta, R., & Kumar, S. Pneumonia detection using deep learning algorithms with CNNs. J. HealthInform.Technol.7(4),221–234(2021). [2] Patel, P., Singh, V., & Sharma, R. Deep learning-based pneumonia detection using chest X-rays and CNNs.J.Comput.VisionBiomed. 10(3), 156–168 (2022). [3] Zhao, H., Li, X., & Zhang, L. MobileNet for real-time pneumoniadetectioninmedical imaging. IEEE Trans. Biomed. Eng. 28(7), 1229–1238 (2023). [4] Wang, L., Li, H., & Zhao, X. A deep learning-based hybrid model forpneumoniaclassificationusing MobileNet. Comput. Biol. Med. 46, 87–97 (2021). [5] Kumar,M.,Yadav,S.,& Sharma,D. Pneumonia detection using MobileNet and X-ray images: A comparativestudy.J.Med.Imag. 11(2), 215–225 (2022). [6] Gupta,R.,Sharma,V.,& Kumar,P.OptimizingMobileNetforfaster pneumonia detection in chest radiographs. Comput. Methods Biomech. Biomed. Eng. 16(3), 391–400 (2023). [7] Meena,S.,Yadav,R.,&Patel,M. Pneumonia detection using CNN- based MobileNet and its variants. Int.J.Comput.Sci.9(5),349–360(2021). [8] Zhang, X., Liu, Y., & Li, C. Transferlearningapproachfor pneumonia detection using MobileNet. J. Health Inform. 18(4), 411–421 (2020). [9] Rani,S.,Kumar,V.,&Ramesh,S. CNN and MobileNet for automated pneumonia detection using chest X-ray images. J. Biomed. Eng. 45(8), 2250–2262(2022). [10] Zhao, L., Liang, C., & Xu, F. A hybrid approach combining MobileNet and SVM for pneumoniaclassification.Neural Comput. Appl. 34(2), 849–858(2023). [11] Sharma,P.,Bansal,P.,& Gupta,A. Pneumonia detection using MobileNetandlightweightneural networks. J. Neural Netw. 14(5), 189–201 (2021). [12] Yadav,M.,Kumar,K.,&Saini,P. AcomparativestudyofMobileNet and ResNet for pneumonia detection using X-ray images. Int.J.Artif.Intell.7(4),501–510(2022). [13] Kumawat,S.,Garg,A.,&Tripathi,R. Pneumonia detection using optimized MobileNet and deep learningmethods.Adv.Comput. Sci. 5(3), 112–122 (2023). [14] Patel,S.,Verma,D.,&Gupta,N. Real-time pneumonia detection using MobileNet for mobile applications.Int.J.Comput.App. 22(7), 130–142 (2022). [15] Sharma, A., Gupta, A., & Chauhan, S. MobileNet-based transferlearningforpneumonia detection from X-ray images. Comput. Vis. PatternRecognit. 34(1), 75–85 (2020).

Copyright

Copyright © 2025 Abinaya T, Aswini M, Renkhanayagi A R K, R Rakshitha, Dr. K. Rajakumari. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.