.png){kind=icon}
.png){kind=icon}
Return
Multi-Modal Deep Learning for Lung Cancer Detection Using Attention-Based Inception-ResNet
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2025.3588407.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Lung cancer is one of the deadliest malignancies worldwide, demanding swift and accurate diagnosis for effective treatment. Traditional screening methods rely on manual interpretation of medical images. However, these methods are time-intensive and highly susceptible to human error. Deep learning (DL) has emerged as a powerful alternative to autonomously identify complex patterns within radiological and histopathological images. Nevertheless, existing DL models for lung cancer detection suffer from critical limitations, including reliance on single imaging modalities, insufficient datasets, suboptimal feature extraction techniques and constrained generalizability. Accordingly, we introduce a novel DL framework that blends diverse imaging modalities, including X-rays, computed tomography (CT) and histopathological images. The proposed framework employs Inception-ResNet module to extract multi-scale spatial features and refine deep feature representations through residual learning. This hybrid module combines the convolutional pathways of Inception architectures with the gradient optimization benefits of residual connections. Besides, the proposed architecture is embedded with sequential multi-scale convolutional fusion and efficient channel attention mechanisms to ameliorate feature diversity and optimize feature importance. These components aid the model to focus on highly discriminative regions within medical images. The proposed model attained an accuracy of 95.35%, 99.68%, 99.73% and 99.26% using X-ray, CT, histopathological and mega datasets, respectively. Comparative experiments unveiled that the proposed model outperformed conventional DL architectures in lung cancer detection. The proposed system, utilizing advanced attention mechanisms and multi-modal imaging capabilities, has the potential to revolutionize early lung cancer diagnosis and extend its impact to other critical diseases. This work represents a paradigm shift in medical image analysis through bridging the gap between DL and clinical applications. The proposed model is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/MohamedHosny90/Lung-Cancer-Detection.git</uri>
Keywords:
Multi-modal imaging
lung cancer detection
hierarchical feature extraction
efficient channel attention
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W