Reducing False-Positives in Lung Nodules Detection Using Balanced Datasets

Abstract

Malignant pulmonary nodules are one of the main manifestations of lung cancer in early CT image screening. Since lung cancer may have no early obvious symptoms, it is important to develop a computer-aided detection (CAD) system to assist doctors to detect the malignant pulmonary nodules in the early stage of lung cancer CT diagnosis. Due to the recent successful applications of deep learning in image processing, more and more researchers have been trying to apply it to the diagnosis of pulmonary nodules. However, due to the ratio of nodules and non-nodules samples used in the training and testing datasets usually being different from the practical ratio of lung cancer, the CAD classification systems may easily produce higher false-positives while using this imbalanced dataset. This work introduces a filtering step to remove the irrelevant images from the dataset, and the results show that the false-positives can be reduced and the accuracy can be above 98%. There are two steps in nodule detection. Firstly, the images with pulmonary nodules are screened from the whole lung CT images of the patients. Secondly, the exact locations of pulmonary nodules will be detected using Faster R-CNN. Final results show that this method can effectively detect the pulmonary nodules in the CT images and hence potentially assist doctors in the early diagnosis of lung cancer.

AI Model for Detection of Abdominal Hemorrhage Lesions in Abdominal CT Images

“…The study in [ 30 ], which organized a pulmonary nodule detection system for false-positive reduction, consisted of a candidate detection phase to detect all suspicious pulmonary nodules and a processing phase to process the extracted patches for various candidate lesions using multiview ConvNets, showing a sensitivity of 90.1% on the Lung Image Database Consortium (LIDC-IDRI) [ 31 ] dataset. A study [ 32 ] on the problem of easily generating higher false positives while using an unbalanced dataset for lung nodule detection showed that introducing a filtering step to remove irrelevant images from the dataset can reduce false positives and increase the accuracy by more than 98%. The study in [ 33 ] proposed a 3D-CNN framework to extract more features by encoding richer spatial information compared with 2D to reduce false positives in automatic pulmonary nodule detection and obtained a sensitivity of 94.4% in the false-positive reduction track of the LUNA 16 Challenge using the LIDC dataset.…”

Deep Learning Applications in Computed Tomography Images for Pulmonary Nodule Detection and Diagnosis: A Review

“…Unsupervised learning approaches such as transfer learning techniques may be more suited in such situations. Balanced datasets, such as those used by [ 105 ], could be used to limit false positives.…”

A clinical-information-free method for early diagnosis of lung cancer from the patients with pulmonary nodules based on backpropagation neural network model

“…Due to its high incidence and mortality, lung cancer is leading cause of cancer deaths globally [3] , [4] . Lack of early obvious clinical symptoms contribute greatly to the high mortality [5] . Thus, early diagnosis and treatment are crucial [6] .…”