Deep Learning-Based Lung Cancer Detection & Classification

Deep Learning-Based Lung Cancer Detection & Classification

Ms.Sathya#1, J.Abarna#2

Assistant Professor(IT) Student(IT)

Francis Xavier Engineering College, Tirunelveli, India

sathya@francisxavier.ac.in

abarnaj.ug.21.it@francisxavier.ac.in

ABSTRACT:

One of the most common and deadly illnesses in the world, lung cancer requires early and precise detection in order to be effectively treated.  In this work, a convolutional neural network (CNN) with transfer learning is used to categorise lung cancer subtypes utilising a deep learning-based methodology.  Deep features were extracted from lung cancer histopathology images using the pre-trained Exception model, increasing classification accuracy while lowering computing expenses.  Image Data Generator was used to enrich the dataset, and the Adam optimiser and categorical cross-entropy loss were used to train the model. Techniques including dropout layers, early halting, and learning rate reduction were used to avoid overfitting. One of the most common and deadly illnesses in the world, lung cancer requires early and precise detection in order to be effectively treated. In this work, a convolutional neural network (CNN) with transfer learning is used to categorise lung cancer subtypes utilising a deep learning-based methodology. Deep features were extracted from lung cancer histopathology images using the pre-trained Exception model, increasing classification accuracy while lowering computing expenses. Image Data Generator was used to enrich the dataset, and the Adam optimiser and categorical cross-entropy loss were used to train the model. Techniques including dropout layers, early halting, and learning rate reduction were used to avoid overfitting. The disparity in validation performance indicates overfitting, despite the promising training accuracy, underscoring the need for a bigger and more varied dataset. Future work will focus on fine-tuning the pre-trained model by unfreezing selective layers, incorporating additional regularization techniques, and exploring advanced augmentation strategies to improve generalizability. Furthermore, integration with clinical workflows and real-world datasets will be explored to validate the model’s applicability in medical diagnostics. The findings of this work demonstrate the potential of deep learning in automating lung cancer diagnosis, establishing a platform for ongoing research in AI-assisted medical imaging. With continuous improvements, this approach could aid radiologists and pathologists in early cancer detection, ultimately improving patient outcomes.

KEYWORDS:

Lung cancer prediction, Lung cancer detection, medical image analysis, Lung tumour classification, Deep learning, machine learning, Transformer network.