Lung Cancer Prediction Using Deep Learning Techniques

Abstract

The introduction of deep learning has had a profound impact on medical research, providing priceless insights and disease-specific prognostic skills. By improving the detection of lung nodules, deep learning techniques, in particular Convolutional Neural Networks (CNNs), have revolutionized the field of medical sciences. Our study aims to take advantage of deep learning's amazing potential for detecting malignant lung nodules in CT scan pictures. We have adopted an ensemble strategy to efficiently address this difficulty, integrating many CNN models to improve their performance and forecast accuracy. The cancerous CT scan dataset, readily accessible through their website, has proven instrumental in our efforts. The CT scan pictures in this dataset, which have been enhanced with important annotations, act as an important training set for our deep learning model. Deep learning uses Artificial Neural Networks to uncover patterns in complicated and nuanced data, drawing inspiration from the intricate operation of neurons in the human brain.We carefully selected a sizable CT scan dataset and used it to train our model. Convolutional Neural Networks (CNNs) are trained using this dataset to distinguish between malignant and non- malignant photos for use in medical diagnosis. In order to conduct a thorough evaluation, we have thoughtfully partitioned the dataset into distinct sets for training, validation, and testing. These meticulously crafted datasets play a crucial role in training and assessing our LungNet, which encompasses three distinct CNNs featuring varying layers, kernels, and pooling techniques. The created LungNet model exhibits training and validation accuracy of 88.13% and 84.85%, respectively. In order to find the best method, deep architectures like InceptionNet, MobileNet, ResNet, VGG, and XceptionNet were thoroughly evaluated both with and without transfer learning. © 2023 IEEE.