Lung Nodule Classification Using Deep Features in CT Images

Abstract Objective: Robustness is an important aspect to consider, when developing methods for medical image analysis. This study investigated robustness properties of deep neural networks (DNNs) for a lung nodule classification problem based on CT images and proposed a solution to improve robustness. Approach: We firstly constructed a class of four DNNs with different widths, each predicting an output label (benign or malignant) for an input CT image cube containing a lung nodule. These networks were trained to achieve Area Under the Curve of 0.891-0.914 on a testing dataset. We then added to the input CT image cubes noise signals generated randomly using a realistic CT image noise model based on a noise power spectrum at 100 mAs, and monitored the DNN’s output change. We defined $SAR_{5} (\%)$ to quantify the robustness of the trained DNN model, indicating that for $5\%$ of CT image cubes, the noise can change the prediction results with a chance of at least $SAR_{5} (\%)$. To understand robustness, we viewed the information processing pipeline by the DNN as a two-step process, with the first step using all but the last layers to extract representations of the input CT image cubes in a latent space, and the second step employing the last fully-connected layer as a linear classifier to determine the position of the sample representations relative to a decision plane. To improve robustness, we proposed to retrain the last layer of the DNN with a Supporting Vector Machine (SVM) hinge loss function to enforce the desired position of the decision plane. Main results: $SAR_{5}$ ranged in $47.0\sim 62.0\%$ in different DNNs. The unrobustness behavior may be ascribed to the unfavorable placement of the decision plane in the latent representation space, which made some samples be perturbed to across the decision plane and hence susceptible to noise. The DNN-SVM model improved robustness over the DNN model and reduced $SAR_{5}$ by $8.8\sim 21.0\%$. Significance: This study provided insights about the potential reason for the unrobustness behavior of DNNs and the proposed DNN-SVM model improved model robustness.

Download Full-text

Shape and margin-aware lung nodule classification in low-dose CT images via soft activation mapping

Medical Image Analysis ◽

10.1016/j.media.2019.101628 ◽

2020 ◽

Vol 60 ◽

pp. 101628 ◽

Cited By ~ 5

Author(s):

Yiming Lei ◽

Yukun Tian ◽

Hongming Shan ◽

Junping Zhang ◽

Ge Wang ◽

...

Keyword(s):

Low Dose ◽

Lung Nodule ◽

Ct Images ◽

Low Dose Ct ◽

Nodule Classification ◽

Activation Mapping

Download Full-text

On the robustness of deep learning-based lung-nodule classification for CT images with respect to image noise

Physics in Medicine and Biology ◽

10.1088/1361-6560/abc812 ◽

2020 ◽

Vol 65 (24) ◽

pp. 245037

Author(s):

Chenyang Shen ◽

Min-Yu Tsai ◽

Liyuan Chen ◽

Shulong Li ◽

Dan Nguyen ◽

...

Keyword(s):

Deep Learning ◽

Lung Nodule ◽

Ct Images ◽

Image Noise ◽

Nodule Classification

Download Full-text

Lung Nodule Classification of CT Images Based on the Deep Learning Algorithms

10.1109/icispc53419.2021.00013 ◽

2021 ◽

Author(s):

Hassan Mkindu ◽

Longwen Wu ◽

Yaqin Zhao ◽

Liang Zhao

Keyword(s):

Deep Learning ◽

Learning Algorithms ◽

Lung Nodule ◽

Ct Images ◽

Nodule Classification

Download Full-text

OP605 Artificial Intelligence Assisted Diagnosis Technology For Benign-Malignant Lung Nodule Classification On Computerized Tomography Images: A Meta-Analysis

International Journal of Technology Assessment in Health Care ◽

10.1017/s0266462320001233 ◽

2020 ◽

Vol 36 (S1) ◽

pp. 14-15

Author(s):

Guo Huang ◽

Yizhong Zhang ◽

Di Xue

Keyword(s):

Artificial Intelligence ◽

Diagnostic Accuracy ◽

Likelihood Ratio ◽

Computerized Tomography ◽

Meta Analysis ◽

Lung Nodule ◽

Ct Images ◽

Lung Nodules ◽

Assisted Diagnosis ◽

Nodule Classification

IntroductionArtificial Intelligence (AI) is an important product of the rapid development of computer technology today. It has a far-reaching impact on the development of medical diagnostic technology especially in combination with medical imaging. The aim of this study was to analyze the diagnostic accuracy of AI-assisted diagnosis technology for classification of benign and malignant lung nodules on Computerized Tomography (CT) images.MethodsA meta-analysis was conducted of published research articles on diagnostic accuracy of AI-assisted diagnosis technology for lung nodules classification between 2010 and 2019 in the databases of PubMed, EMBASE, Cochrane Library, China National Knowledge Infrastructure, Wanfang Data Knowledge Service Platform and China Bio-medicine Database. Statistical analysis was performed with the software SAS 9.4 and Stata 12.0, and the summary receiver operating characteristic (SROC) curve was drawn to evaluate accuracy of the method.ResultsA total of 27 studies with 5,701 lung nodules were considered. The pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio and the area under the curve of SROC for AI-assisted diagnosis technology for lung nodules classification respectively were 0.892 (95% confidence interval [CI]: 0.854–0.920), 0.876 (95% CI: 0.833–0.909), 7.190 (95% CI: 5.194–9.955), 0.124 (95% CI: 0.089–0.171), 58.102 (95% CI: 32.391–104.219) and 0.95 (95%CI: 0.92–0.96).ConclusionsOf note, several limitations should be considered when interpreting the findings of this meta-analysis. Data acquisition is not comprehensive enough because the language of the literature search was limited to Chinese and English. Furthermore,heterogeneity caused due to the difference of lung nodule size affected the study results. Despite these limitations, our study suggests that AI-assisted diagnosis technology for benign-malignant lung nodule classification on CT images obtains high diagnostic accuracy, and it can be used as a novel method to differentiate benign and malignant pulmonary nodules.

Download Full-text