Automatic pulmonary ground-glass opacity nodules detection and classification based on 3D neural network.
Medical physics
PURPOSE:Pulmonary ground-glass opacity (GGO) nodules are more likely to be malignant compared with solid solitary nodules. Due to indistinct boundaries of GGO nodules, the detection and diagnosis are challenging for doctors. Therefore, designing an automatic GGO nodule detection and classification scheme is significantly essential. METHODS:In this paper, we proposed a two-stage 3D GGO nodule detection and classification framework. First, we used a pretrained 3D U-Net to extract lung parenchyma. Second, we adapted the architecture of Mask region-based convolutional neural networks (RCNN) to handle 3D medical images. The 3D model was then applied to detect the locations of GGO nodules and classify lesions (benign or malignant). The class-balanced loss function was also used to balance the number of benign and malignant lesions. Finally, we employed a novel false positive elimination scheme called the feature-based weighted clustering (FWC) to promote the detection accuracy further. RESULTS:The experiments were conducted based on fivefold cross-validation with the imbalanced data set. Experimental results showed that the mean average precision could keep a high level (0.5182) in the phase of detection. Meanwhile, the false positive rate was effectively controlled, and the competition performance metric (CPM) reached 0.817 benefited from the FWC algorithm. The comparative statistical analyses with other deep learning methods also proved the effectiveness of our proposed method. CONCLUSIONS:We put forward an automatic pulmonary GGO nodules detection and classification framework based on deep learning. The proposed method locate and classify nodules accurately, which could be an effective tool to help doctors in clinical diagnoses.
10.1002/mp.15501
Pulmonary nodules detection assistant platform: An effective computer aided system for early pulmonary nodules detection in physical examination.
Han Yu,Qi Honggang,Wang Ling,Chen Chen,Miao Jun,Xu Hongbo,Wang Ziqi,Guo Zhijun,Xu Qian,Lin Qiang,Liu Haitao,Lu Junying,Liang Fei,Feng Wenqiu,Li Haiyan,Liu Yan
Computer methods and programs in biomedicine
BACKGROUND AND OBJECTIVE:Early detection of the pulmonary nodule from physical examination low-dose computer tomography (LDCT) images is an effective measure to reduce the mortality rate of lung cancer. Although there are many computer aided diagnosis (CAD) methods used for detecting pulmonary nodules, there are few CAD systems for small pulmonary nodule detection with a large amount of physical examination LDCT images. METHODS:In this work, we designed a CAD system called Pulmonary Nodules Detection Assistant Platform for early pulmonary nodules detection and classification based on the physical examination LDCT images. Based on the preprocessed physical examination CT images, the three-dimensional (3D) CNN-based model is presented to detect candidate pulmonary nodules and output detection results with quantitative parameters, the 3D ResNet is used to classify the detected nodules into intrapulmonary nodules and pleural nodules to reduce the physician workloads, and the Fully Connected Neural Network (FCNN) is used to classify ground-glass opacity (GGO) nodules and non-GGO nodules to help doctor pay more attention to those suspected early lung cancer nodules. RESULTS:Experiments are performed on our 1000 samples of physical examinations (LNPE1000) with an average diameter of 5.3 mm and LUNA16 dataset with an average diameter of 8.31 mm, which show that the designed CAD system is automatic and efficient for detecting smaller and larger nodules from different datasets, especially for the detection of smaller nodules with diameter between 3 mm and 6 mm in physical examinations. The accuracy of pulmonary nodule detection reaches 0.879 with an average of 1 false positive per CT in LNPE1000 dataset, which is comparable to the experienced physicians. The classification accuracy reaches 0.911 between intrapulmonary and pleural nodules, and 0.950 between GGO and non-GGO nodules, respectively. CONCLUSION:Experimental results show that the proposed pulmonary nodule detection model is robust for different datasets, which can successfully detect smaller and larger nodules in CT images obtained by physical examination. The interactive platform of the designed CAD system has been on trial in a hospital by combining with manual reading, which helps doctors analyze clinical data dynamically and improves the nodule detection efficiency in physical examination applications.
10.1016/j.cmpb.2022.106680
Hybrid Clinical-Radiomics Model for Precisely Predicting the Invasiveness of Lung Adenocarcinoma Manifesting as Pure Ground-Glass Nodule.
Song Lan,Xing Tongtong,Zhu Zhenchen,Han Wei,Fan Guangda,Li Ji,Du Huayang,Song Wei,Jin Zhengyu,Zhang Guanglei
Academic radiology
RATIONALE AND OBJECTIVES:To identify whether the radiomics features of computed tomography (CT) allowed for the preoperative discrimination of the invasiveness of lung adenocarcinomas manifesting as pure ground-glass nodules (pGGNs) and further to develop and compare different predictive models. MATERIALS AND METHODS:We retrospectively included 187 lung adenocarcinomas presenting as pGGNs (66 preinvasive lesions and 121 invasive lesions), which were randomly divided into the training and test sets (8:2). Radiomics features were extracted from non-enhanced CT images. Clinical features, including patient's demographic characteristics, smoking status, and conventional CT features that reflect tumor's morphology and surrounding information were also collected. Intraclass correlation coefficient and ℓ-norm minimization were used to identify influential feature subset which was then used to build three predictive models (clinical, radiomics, and clinical-radiomics models) with the gradient boosting regression tree classifier. The performances of the predictive models were evaluated using the area under the curve (AUC). RESULTS:Of the 1409 radiomics features and 27 clinical feature subtypes, 102 features were selected to construct the hybrid clinical-radiomics model, which achieved the best discriminative power (AUC = 0.934 and 0.929 in training and test set). The radiomics model showed comparable predictive performance (AUC = 0.911 and 0.901 in training and test set) compared to the clinical model (AUC = 0.911 and 0.894 in training and test set). CONCLUSION:The radiomics model showed good predictive performance in discriminating invasive lesions from preinvasive lesions for lung adenocarcinomas presenting as pGGNs. Its performance can be further improved by adding clinical features.
10.1016/j.acra.2020.05.004
3D deep learning based classification of pulmonary ground glass opacity nodules with automatic segmentation.
Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society
Classifying ground-glass lung nodules (GGNs) into atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IAC) on diagnostic CT images is important to evaluate the therapy options for lung cancer patients. In this paper, we propose a joint deep learning model where the segmentation can better facilitate the classification of pulmonary GGNs. Based on our observation that masking the nodule to train the model results in better lesion classification, we propose to build a cascade architecture with both segmentation and classification networks. The segmentation model works as a trainable preprocessing module to provide the classification-guided 'attention' weight map to the raw CT data to achieve better diagnosis performance. We evaluate our proposed model and compare with other baseline models for 4 clinically significant nodule classification tasks, defined by a combination of pathology types, using 4 classification metrics: Accuracy, Average F1 Score, Matthews Correlation Coefficient (MCC), and Area Under the Receiver Operating Characteristic Curve (AUC). Experimental results show that the proposed method outperforms other baseline models on all the diagnostic classification tasks.
10.1016/j.compmedimag.2020.101814
A deep residual learning network for predicting lung adenocarcinoma manifesting as ground-glass nodule on CT images.
European radiology
OBJECTIVE:To develop a deep learning-based artificial intelligence (AI) scheme for predicting the likelihood of the ground-glass nodule (GGN) detected on CT images being invasive adenocarcinoma (IA) and also compare the accuracy of this AI scheme with that of two radiologists. METHODS:First, we retrospectively collected 828 histopathologically confirmed GGNs of 644 patients from two centers. Among them, 209 GGNs are confirmed IA and 619 are non-IA, including 409 adenocarcinomas in situ and 210 minimally invasive adenocarcinomas. Second, we applied a series of pre-preprocessing techniques, such as image resampling, rescaling and cropping, and data augmentation, to process original CT images and generate new training and testing images. Third, we built an AI scheme based on a deep convolutional neural network by using a residual learning architecture and batch normalization technique. Finally, we conducted an observer study and compared the prediction performance of the AI scheme with that of two radiologists using an independent dataset with 102 GGNs. RESULTS:The new AI scheme yielded an area under the receiver operating characteristic curve (AUC) of 0.92 ± 0.03 in classifying between IA and non-IA GGNs, which is equivalent to the senior radiologist's performance (AUC 0.92 ± 0.03) and higher than the score of the junior radiologist (AUC 0.90 ± 0.03). The Kappa value of two sets of subjective prediction scores generated by two radiologists is 0.6. CONCLUSIONS:The study result demonstrates using an AI scheme to improve the performance in predicting IA, which can help improve the development of a more effective personalized cancer treatment paradigm. KEY POINTS:• The feasibility of using a deep learning method to predict the likelihood of the ground-glass nodule being invasive adenocarcinoma. • Residual learning-based CNN model improves the performance in classifying between IA and non-IA nodules. • Artificial intelligence (AI) scheme yields higher performance than radiologists in predicting invasive adenocarcinoma.
10.1007/s00330-019-06533-w
HRCT features between lepidic-predominant type and other pathological subtypes in early-stage invasive pulmonary adenocarcinoma appearing as a ground-glass nodule.
Zhang Pengju,Li Tianran,Tao Xuemin,Jin Xin,Zhao Shaohong
BMC cancer
BACKGROUND:Different pathological subtypes of invasive pulmonary adenocarcinoma (IPA) have different surgical methods and heterogeneous prognosis. It is essential to clarify IPA subtypes before operation and high-resolution computed tomography (HRCT) plays a very important role in this regard. We aimed to investigate the HRCT features of lepidic-predominant type and other pathological subtypes of early-stage (T1N0M0) IPA appearing as a ground-glass nodule (GGN). METHODS:We performed a retrospective analysis on clinical data and HRCT features of 630 lesions in 589 patients with pathologically confirmed IPA (invasive foci > 5 mm) appearing as pure GGN (pGGN) and mixed GGN (mGGN) with consolidation-to-tumor ratio (CTR) ≤0.5 from January to December 2019. All GGNs were classified as lepidic-predominant adenocarcinoma (LPA) and nonlepidic-predominant adenocarcinoma (n-LPA) groups. Univariate analysis was performed to analyze the differences of clinical data and HRCT features between the LPA and n-LPA groups. Multivariate analysis was conducted to determine the variables to distinguish the LPA from n-LPA group independently. The diagnostic performance of different parameters was compared using receiver operating characteristic curves. RESULTS:In total, 367 GGNs in the LPA group and 263 GGNs in the n-LPA group were identified. In the univariate analysis, the CTR, mean CT values, and mean diameters as well as mixed GGN, deep lobulation, spiculation, vascular change, bronchial change, and tumor-lung interface were smaller in the LPA group than in the n-LPA group (P < 0.05). Logistic regression model was reconstructed including the mean CT value, CTR, deep lobulation, spiculation, vascular change, and bronchial change (P < 0.05). Area under the curve of the logistic regression model for differentiating LPA and n-LPA was 0.840 (76.4% sensitivity, 78.7% specificity), which was significantly higher than that of the mean CT value or CTR. CONCLUSIONS:Deep lobulation, spiculation, vascular change, and bronchial change, CT value > - 472.5 HU and CTR > 27.4% may indicate nonlepidic predominant invasive pulmonary adenocarcinoma in GGNs.
10.1186/s12885-021-08821-5
Computer-aided diagnosis of ground glass pulmonary nodule by fusing deep learning and radiomics features.
Hu Xianfang,Gong Jing,Zhou Wei,Li Haiming,Wang Shengping,Wei Meng,Peng Weijun,Gu Yajia
Physics in medicine and biology
OBJECTIVES:This study aims to develop a computer-aided diagnosis (CADx) scheme to classify between benign and malignant ground glass nodules (GGNs), and fuse deep leaning and radiomics imaging features to improve the classification performance. METHODS:We first retrospectively collected 513 surgery histopathology confirmed GGNs from two centers. Among these GGNs, 100 were benign and 413 were malignant. All malignant tumors were stage I lung adenocarcinoma. To segment GGNs, we applied a deep convolutional neural network and residual architecture to train and build a 3D U-Net. Then, based on the pre-trained U-Net, we used a transfer learning approach to build a deep neural network (DNN) to classify between benign and malignant GGNs. With the GGN segmentation results generated by 3D U-Net, we also developed a CT radiomics model by adopting a series of image processing techniques, i.e. radiomics feature extraction, feature selection, synthetic minority over-sampling technique, and support vector machine classifier training/testing, etc. Finally, we applied an information fusion method to fuse the prediction scores generated by DNN based CADx model and CT-radiomics based model. To evaluate the proposed model performance, we conducted a comparison experiment by testing on an independent testing dataset. RESULTS:Comparing with DNN model and radiomics model, our fusion model yielded a significant higher area under a receiver operating characteristic curve (AUC) value of 0.73 ± 0.06 (P < 0.01). The fusion model generated an accuracy of 75.6%, F1 score of 84.6%, weighted average F1 score of 70.3%, and Matthews correlation coefficient of 43.6%, which were higher than the DNN model and radiomics model individually. CONCLUSIONS:Our experimental results demonstrated that (1) applying a CADx scheme was feasible to diagnosis of early-stage lung adenocarcinoma, (2) deep image features and radiomics features provided complementary information in classifying benign and malignant GGNs, and (3) it was an effective way to build DNN model with limited dataset by using transfer learning. Thus, to build a robust image analysis based CADx model, one can combine different types of image features to decode the imaging phenotypes of GGN.
10.1088/1361-6560/abe735
Evaluation of Computer-Aided Nodule Assessment and Risk Yield (CANARY) in Korean patients for prediction of invasiveness of ground-glass opacity nodule.
Lee Juyoung,Bartholmai Brian,Peikert Tobias,Chun Jaehee,Kim Hojin,Kim Jin Sung,Park Seong Yong
PloS one
Differentiating the invasiveness of ground-glass nodules (GGN) is clinically important, and several institutions have attempted to develop their own solutions by using computed tomography images. The purpose of this study is to evaluate Computer-Aided Analysis of Risk Yield (CANARY), a validated virtual biopsy and risk-stratification machine-learning tool for lung adenocarcinomas, in a Korean patient population. To this end, a total of 380 GGNs from 360 patients who underwent pulmonary resection in a single institution were reviewed. Based on the Score Indicative of Lung Cancer Aggression (SILA), a quantitative indicator of CANARY analysis results, all of the GGNs were classified as "indolent" (atypical adenomatous hyperplasia, adenocarcinomas in situ, or minimally invasive adenocarcinoma) or "invasive" (invasive adenocarcinoma) and compared with the pathology reports. By considering the possibility of uneven class distribution, statistical analysis was performed on the 1) entire cohort and 2) randomly extracted six sets of class-balanced samples. For each trial, the optimal cutoff SILA was obtained from the receiver operating characteristic curve. The classification results were evaluated using several binary classification metrics. Of a total of 380 GGNs, the mean SILA for 65 (17.1%) indolent and 315 (82.9%) invasive lesions were 0.195±0.124 and 0.391±0.208 (p < 0.0001). The area under the curve (AUC) of each trial was 0.814 and 0.809, with an optimal threshold SILA of 0.229 for both. The macro F1-score and geometric mean were found to be 0.675 and 0.745 for the entire cohort, while both scored 0.741 in the class-equalized dataset. From these results, CANARY could be confirmed acceptable in classifying GGN for Korean patients after the cutoff SILA was calibrated. We found that adjusting the cutoff SILA is needed to use CANARY in other countries or races, and geometric mean could be more objective than F1-score or AUC in the binary classification of imbalanced data.
10.1371/journal.pone.0253204
Risk Factors for The Growth of Residual Nodule in Surgical Patients with Adenocarcinoma Presenting as Multifocal Ground-glass Nodules.
Xia Tianyi,Cai Mengting,Zhuang Yuandi,Ji Xiaowei,Huang Dingpin,Lin Liaoyi,Liu Jinjin,Yang Yunjun,Fu Gangze
European journal of radiology
PURPOSE:We aim to investigate the risk factors influencing the growth of residual nodule (RN) in surgical patients with adenocarcinoma presenting as multifocal ground-glass nodules (GGNs). METHOD:From January 2014 to June 2018, we enrolled 238 patients with multiple GGNs in a retrospective review. Patients were categorized into growth group 63 (26.5%), and non-growth group 175 (73.5%). The median follow-up time was 28.2 months (range, 6.3-73.0 months). To obtain the time of RN growth and find the risk factors for growth, data such as age, gender, history of smoking, history of malignancy, type of surgery, pathology and radiological characteristics were analyzed to use Kaplan-Meier method with the log-rank test and Cox regression analysis. RESULTS:The median growth time of RN was 56.0 months (95% CI, 45.0-67.0 months) in all 238 patients. Roundness (HR 4.62, 95% CI 2.20-9.68), part-solid nodule (CTR ≥ 50%) (HR 4.39, 95% CI 2.29-8.45), vascular convergence sign (HR 2.32, 95% CI 1.36-3.96) of RN, and age (HR 1.04, 95% CI 1.01-1.07) were independent predictors of further nodule growth. However, radiological characteristics and pathology of domain tumour (DT) cannot be used as indicators to predict RN growth. CONCLUSIONS:RN showed an indolent growth pattern in surgical patients with multifocal GGNs. RN with a higher roundness, presence of vascular convergence sign, more solid component, and in the elder was likely to grow. However, the growth of RN showed no association with the radiological features and pathology of DT.
10.1016/j.ejrad.2020.109332
Computed tomography density is not associated with pathological tumor invasion for pure ground-glass nodules.
Fu Fangqiu,Zhang Yang,Wang Shengping,Li Yuan,Wang Zezhou,Hu Hong,Chen Haiquan
The Journal of thoracic and cardiovascular surgery
OBJECTIVE:Pure ground-glass nodules are considered to be radiologically noninvasive in lung adenocarcinoma. However, some pure ground-glass nodules are found to be invasive adenocarcinoma pathologically. This study aims to identify the computed tomography parameters distinguishing invasive adenocarcinoma from adenocarcinoma in situ and minimally invasive adenocarcinoma. METHODS:From May 2011 to December 2015, patients with completely resected adenocarcinoma appearing as pure ground-glass nodules were reviewed. To evaluate the association between computed tomography features and the invasiveness of pure ground-glass nodules, logistic regression analyses were conducted. RESULTS:Among 432 enrolled patients, 118 (27.3%) were classified as adenocarcinoma in situ, 213 (49.3%) were classified as minimally invasive adenocarcinoma, 101 (23.4%) were classified as invasive adenocarcinoma. There was no postoperative recurrence for patients with pure ground-glass nodules. Logistic regression analyses demonstrated that computed tomography size was the only independent radiographic factor associated with adenocarcinoma in situ (odds ratio, 47.165; 95% confidence interval, 19.279-115.390; P < .001), whereas computed tomography density was not (odds ratio, 1.002; 95% confidence interval, 0.999-1.005; P = .127). Further analyses revealed that there was no distributional difference in computed tomography density among 3 groups (P = .173). Even after propensity score matching for adenocarcinoma in situ/minimally invasive adenocarcinoma and invasive adenocarcinoma, no significant difference in computed tomography density was observed (P = .741). The subanalyses for pure ground-glass nodules with 1 cm or more in size also indicated similar results. CONCLUSIONS:In patients with pure ground-glass nodules, computed tomography size was the only radiographic parameter associated with tumor invasion. Measuring computed tomography density provided no advantage in differentiating invasive adenocarcinoma from adenocarcinoma in situ and minimally invasive adenocarcinoma.
10.1016/j.jtcvs.2020.04.169