Abstract: Thyroid cancer is the deadliest cancer worldwide. It has been shown that early detection using low-dose computer tomography (LDCT) scans can reduce deaths caused by this disease. We aim to present a general framework for the detection of thyroid cancer in chest LDCT images. Our method consists of a nodule detector followed by a cancer predictor method. Our candidate extraction approach is expected to produce higher accuracy on the images of each subject and is also expected to increase precision for all recall values using convolutional neural networks. Our model insists over 3D CNN than other methods as they include different planes in detecting the nodules. In addition, our false positive reduction stage aims to successfully classify the candidates and is expected to increase precision. Our cancer predictor's ROC AUC curve is expected to determine how well our model can classify the nodules from the non-nodules based on the features and their properties.
Keywords: Pulmonary Nodule detection, Thyroid Cancer, Machine Learning, Deep Learning, KNN, Image processing, ANN, Random Forest, Logistic Regression.
|
DOI:
10.17148/IJARCCE.2025.14158
