Abstract:
Lung cancer is one of the most prevalent causes of carcinoma-related fatalities globally; precise and effective diagnostic
instruments are desperately needed. Researchers and physicians continue to encounter difficulties when aiming to employ deep
learning models in healthcare environments for recognizing lung cancer, in order achieve higher sensitivity and accuracy on large
data sets, despite a number of advancements in this field. In this work, we enhance the recognition and classification of lung cancer
from CT images using a distinct deep learning approach. Here, hybrid deep learning model that blends the potency of CNNs with the
cutting-edge design of capsule networks is proposed. Inspired by recent advancements, particularly the VggCapNet model (VGG16
and Capsule network), our model integrates the VGG19 and Capsule Network architectures to address orientation-related challenges
often encountered in traditional CNN-based approaches. The creation, training, and assessment of this hybrid model for lung nodule
identification and categorization are our main research goals. We focus on achieving superior Performance indicators, such as high
Accuracy, Specificity, F1 score, and Sensitivity with a significant emphasis on reducing false positive rates. We compare the
outcomes of our suggested model with the most recent findings in the field to confirm its efficacy. We anticipate that our research
has yield several valuable outcomes, including improved nodule classification accuracy of 99.20%, reduced false positive rates, and
minimized training times. This research could lead to more widespread utilization in cancer diagnosis by enhancing early lung cancer
detection and developing the field of medical image analysis.
