Lung Cancer Detection Using Nodule Based Methods and Machine Learning Algorithms

Overview

Lung cancer is the leading cause of cancer-related deaths worldwide. Early and accurate detection is key to effective treatment. However, there is a need to develop an innovative diagnostic and detection method to promptly detect the presence and type of cancer. The aim of this project was to develop an innovative lung cancer detection algorithm, LCDetect. It uses complex machine learning methods and nodule-based analysis to detect the presence of cancer, malignancy, type and stage of cancer. A lung CT scan is used to provide a thorough diagnosis, reducing the need for and delay of a PET scan and biopsy, and reducing the cost of diagnosis significantly. The algorithm works in three modules; image preprocessing, image detection, and a convolutional neural network model (CNN). Preprocessing includes noise removal, normalization, image filters, segmentation and augmentation. Extracted regions of interest are passed to image detection for feature extraction of the nodules after pixel calculations and conversions. Based on the extracted features, the layers of CNN categorize the lung cancer using location-based analysis. After several optimization techniques such as back propagation and regression, a thorough statistical analysis was performed to evaluate metrics such as confusion matrix, ROC curve, and accuracy. Then, the algorithm was deployed on Azure Web Services. The system was trained and tested across 50,000 datasets and patient CT scans from local radiologists. The final algorithm passed with an accuracy of 98%. LCDetect is an innovative and fully functional solution to accurately detect for adenocarcinoma, squamous cell, non-small cell, and small cell lung cancers and to predict the stage. Using low-level computational techniques and the input CT scan, LCDetect satisfies the goals of accuracy and performance to reduce detection time.