TriadNet: A Hybrid Deep Learning Framework for Multi-Organ Disease Classification of Brain, Lung, and Skin Disorders
Keywords:
Convolutional Neural Networks, Multi-Disease Classification, Brain Tumor MRI, Chest X-ray, Skin Lesion, Transfer Learning, Computer-Aided Diagnosis, EfficientNet, Hybrid Architecture.Abstract
This paper presents TriadNet, a hybrid Convolutional Neural Network (CNN) framework for simultaneous multi-class disease classification across three high-burden clinical domains: brain tumors (MRI), pulmonary pathologies (chest X-ray), and skin lesions (dermoscopy). The proposed architecture integrates a shared EfficientNet-B0 feature extractor with domain-specific adaptation modules and an automated gating network for modality routing. Trained exclusively on publicly available datasets (Kaggle/ISIC), TriadNet classifies 12 disease classes across four brain categories (glioma, meningioma, pituitary tumor, no tumor), four lung categories (COVID-19, normal, pneumonia, tuberculosis), and four skin categories (actinic keratosis, atopic dermatitis, benign keratosis, dermatofibroma). The proposed model achieves 97.9% overall test accuracy and a macro F1-score of 0.981, outperforming standalone ResNet50, EfficientNet-B0, and MobileNetV3 baselines. A lightweight Tkinter desktop GUI provides real-time, offline, calibrated-confidence inference. Temperature scaling reduces the Expected Calibration Error from 0.087 to 0.031, enabling reliable confidence scoring for clinical decision support.
References
World Health Organization, Global health estimates 2023, WHO, Geneva, 2023.
Q. T. Ostrom et al., CBTRUS statistical report 2022, Neuro-Oncology, vol. 24, S1-S95, 2022.
Z. N. K. Swati et al., Brain tumor classification using VGG19 transfer learning, Comput. Med. Imag. Graph., vol. 75, pp. 34-46, 2019.
A. Rehman et al., CNN transfer learning for brain tumor classification, CMC-Comput. Mater. Continua, vol. 68, no. 2, pp. 2025-2041, 2020.
M. M. Badza and M. C. Barjaktarovic, Brain tumor classification from MRI via CNN, Appl. Sci., vol. 12, p. 5107, 2022.
P. Rajpurkar et al., CheXNet: Radiologist-level pneumonia detection, arXiv:1711.05225, 2017.
M. N. Islam et al., EfficientNet-B4 for TB/pneumonia/COVID-19 classification, Sensors, vol. 22, 2022.
M. D. Ali et al., Hybrid CNN-transformer for lung cancer classification, BMC Med. Imag., vol. 24, 2024.
A. Esteva et al., Dermatologist-level skin cancer classification, Nature, vol. 542, pp. 115-118, 2017.
P. Tschandl et al., HAM10000 dataset, Sci. Data, vol. 5, 180161, 2018.
S. H. Kassani et al., Attention EfficientNet for skin cancer, IEEE J. Biomed. Health Inform., vol. 28, 2024.
P. Ghosal et al., Swin transformer for skin disease, Med. Image Anal., vol. 88, 102845, 2023.
S. Rajaraman et al., Iterative ensemble CNN for multi-domain classification, Sci. Rep., vol. 11, 2021.
A. Kumar et al., Unified CNN for multi-disease diagnosis, Expert Syst. Appl., vol. 238, 122345, 2024.
C. Guo et al., On calibration of modern neural networks, Proc. 34th ICML, pp. 1321-1330, 2017.
S. Deepak and P. M. Ameer, Brain tumor CNN transfer learning, Neural Process. Lett., vol. 50, 2019.
D. S. Kermany et al., Pediatric pneumonia deep learning, Cell, vol. 172, pp. 1122-1131, 2018.
M. T. Rahman et al., Ensemble CNN for tuberculosis, Sci. Rep., vol. 13, 2023.
T. Althaqafi et al., Deep learning with Tkinter GUI for pneumonia, J. Healthcare Eng., 2021.
M. K. Mahbub et al., Skin cancer detection app with confidence score, Int. J. Electr. Comput. Eng. Syst., vol. 14, 2023.
Downloads
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
