IoT-Based Signal Processing for Lung Nodule Detection using 3D CT Images with 3D Convolutional Neural Networks and Feature Selection

Abstract

The recognition of lung nodules through 3D CT imaging is an important and highly precise task that facilitates early diagnosis of lung cancer.  In this paper, an IoT-based signal processing framework has been suggested that integrates 3D convolutional neural networks (3D CNNs) using advanced feature selection methods for the detection of lung nodules.  The system processes the medical images in steps: dimensionality reduction, contrast enhancement, and noise removal, after which the derived features go to the 3D CNN for classification.  In this process, for improved performance of the model, feature selection techniques like wrappers and hybrid filters are used in such a way as to ensure that the most relevant features support the detection of abnormalities. By means of the insightful clinical timelines of faster decision-making and real-time image transmission and processing thanks to IoT integration, the performance evaluation of the system gave great study-like recall, accuracy, precision, and AUC-ROC values indicative of promise in lung nodule diagnosis in an automated and real-time framework.  With the aforementioned, this study provides great insight into how IoT, deep learning, and feature selection can be synergistically brought together to complement lung nodule diagnosis in medical imaging.