Comparative Analysis of Hybrid 1D-CNN-LSTM and VGG16-1D-LSTM for Lung Lesion Classification

• Published in: JOURNAL OF ELECTRICAL ENGINEERING & TECHNOLOGY
• Main Authors: Jafery, Nurul Najiha; Sulaiman, Siti Noraini; Osman, Muhammad Khusairi; Karim, Noor Khairiah Abdul; Soh, Zainal Hisham Che; Isa, Nor Ashidi Mat
• Format: Article; Early Access
• Language: English
• Published: SPRINGER SINGAPORE PTE LTD 2025
• Subjects: Engineering
• Online Access: https://www-webofscience-com.uitm.idm.oclc.org/wos/woscc/full-record/WOS:001438966000001
• ISSN: 1975-0102; 2093-7423
• DOI: 10.1007/s42835-025-02182-w

Lung cancer accounts for approximately 20% of cancer deaths and is the second most common cancer among men and women, with an average age of diagnosis of 70 years. Early and accurate detection is crucial for improving patient outcomes through timely intervention and treatment. Enhancing lesion characterisation is key to advancing diagnostic accuracy. Deep learning provides a powerful tool for early diagnosis by enabling the development of sophisticated models that can accurately classify lung lesions in CT scans. This study investigates the effectiveness of two deep learning architectures for this purpose: a hybrid 1D-CNN-LSTM and a VGG16-1D-LSTM model. Both models classify lung lesions in CT scans using regression features and are evaluated with optimisers such as Adam, RMSprop, and SGD. Result reveal that the hybrid 1D-CNN-LSTM model with the Adam optimizer achieved 96% accuracy, 90% precision, 94.74% recall, and a 92.31% F1-score. The VGG16-1D-LSTM model with Adam also achieved 96% accuracy but with 85% precision, 100% recall, and a 91.89% F1-score. These findings suggest that the hybrid 1D-CNN-LSTM architecture with Adam optimisation offers a promising approach for accurate lung lesion classification, potentially improving early detection and patient outcomes in lung cancer cases.