A multiscale and dual-loss network for pulmonary nodule classification

<p indent="0mm">Current diagnostic methods for pulmonary nodules are not ideal in clinical practice, especially for distinguishing benign from malignant nodules in high-risk populations. Therefore, there is an urgent need for precise and non-invasive diagnostic strategies. The oral microbiota, as an emerging biomarker, has received widespread attention, and advanced machine-learning algorithms provide new ideas for the development of non-invasive microbial markers for early identification of pulmonary nodules. In this study, saliva samples and information on traditional Chinese medicine constitution were collected from patients with pulmonary nodules. Based on histopathological detection of nodule tissue and the results of a blood stasis constitution scoring table, patients with benign (<italic>n</italic>=136) and malignant (<italic>n</italic>=73) pulmonary nodules of blood stasis constitution were selected as the study subjects. Using 16S rRNA sequencing technology and bioinformatics analysis, we revealed the oral microbial diversity, microbiota health index, microbiota dysbiosis index, species composition, differential bacterial genera, and the correlation between microorganisms and clinical characteristics of the subjects in patients with benign and malignant pulmonary nodules of blood stasis constitution. Based on the amplicon sequence variant of the salivary microbiota, seven machine learning techniques [logistic regression, support vector machine, multi-layer perceptron, naïve bayes, random forest, gradient boosting decision trees, light gradient boosting machine (lightGBM)] were used to construct a diagnostic model for pulmonary nodules of blood stasis constitution, and the predictive performance metrics were evaluated to determine the optimal performance model. The feature importance ranking based on the optimal model and the SHAP method was used to explain the contribution of features and screen for potential diagnostic biomarkers of pulmonary nodules in blood stasis constitution. Finally, based on PICRUSt2 functional prediction analysis of the microbiota, the potential mechanism of the oral microbiota in the malignant progression of pulmonary nodules was explored. The study suggested that salivary microbiota diversity of benign and malignant pulmonary nodules in blood stasis constitution was significantly different, with significantly reduced α and β diversity values in malignant pulmonary nodules (<italic>P</italic>​&lt;0.05). Simultaneously, compared to benign pulmonary nodules in blood stasis constitution, the microbiota health index was significantly reduced, and the microbiota dysbiosis index was significantly increased in patients with malignant pulmonary nodules (<italic>P</italic>​&lt;0.05). In terms of species composition, we found rich microbiota composition and high levels of differential microbiota in the saliva of patients with benign and malignant pulmonary nodules of blood stasis constitution, respectively. Additionally, Distance-based Redundancy Analysis showed a correlation between salivary microbiota composition and clinical characteristics in benign and malignant pulmonary nodules of blood stasis constitution. The seven machine learning models we developed were able to distinguish the malignancy probability of pulmonary nodules in blood stasis constitution. When validated on the model test dataset, the average performance AUC was 0.693–0.893, with LightGBM achieving high accuracy [area under the curve (AUC)=0.893, 95% CI: 0.866–0.925]. Based on the importance ranking of optimal model features and the SHAP method to explain feature contributions, potential biomarkers for predicting the risk of malignancy in pulmonary nodules of blood stasis constitution were identified as <italic>Granulicatella</italic>, <italic>Fusobacterium</italic>, <italic>Delftia</italic>, <italic>Streptococcus</italic>, <italic>Alloprevotella</italic>, and <italic>Veillonella</italic>. Finally, based on PICRUSt2 functional prediction analysis of the microbiota, we found interesting metabolic reprogramming that may exist in the oral microbiota during the malignant development of pulmonary nodules. Overall, machine learning models trained based on salivary microbiota features have good predictive performance. The oral microbiota can serve as a potential diagnostic biomarker for the differential diagnosis of pulmonary nodules in blood stasis constitution. This discovery opens new avenues for the development of non-invasive, highly accurate identification tools for high-risk populations with pulmonary nodules and deserves further research and clinical validation.

Objective: To explore the value of detection of epidermal growth factor receptor (EGFR) gene amplification in peripheral blood rare cells in the assessment of benign and malignant pulmonary nodules. Methods: A total of 262 patients with pulmonary nodules were selected as the retrospectively study subjects from the Second Affiliated Hospital of Army Military Medical University and Peking Union Medical College Hospital from July 2022 to August 2023. There were 98 males and 164 females, with the age range from 16 to 79 (52.1±12.1) years. The EGFR gene amplification testing was performed on the rare cells enriched from patients' peripheral blood, and the clinical manifestations, CT imaging features, histopathological and/or pathological cytological confirmed results of patients were collected. The receiver operating characteristic (ROC) curve was used to determine the optimal cut-off value of the method of detection of EGFR gene amplification in peripheral blood rare cells, and its diagnostic efficacy was evaluated. Results: Among the 262 patients, 143 were malignant pulmonary nodules and 119 were benign pulmonary nodules. The differences between malignant pulmonary nodules and benign pulmonary nodules in nodule diameter and nodule density were statistically significant (both P<0.001), while the differences in age, gender and nodule number were not statistically significant (all P>0.05). The number [M (Q1, Q3)] of EGFR gene amplification positive rare cells in patients with malignant pulmonary nodule was 8 (6, 11), which was higher than that in patients with benign pulmonary nodule [2 (1, 4), P<0.001]. The ROC curve results showed that when the optimal cut-off value was 5 (that was, the number of EGFR gene amplification positive rare cells was>5), the area under the curve (AUC) of the detection of EGFR gene amplification in peripheral blood rare cells for discrimination of benign and malignant pulmonary lesions was 0.816 (95%CI: 0.761-0.870), with a sensitivity of 83.2%, a specificity of 80.7%, and an accuracy of 82.1%. Based on the analysis of the diameter of the nodules, the AUC for distinguishing between benign and malignant pulmonary nodules with diameter 5-9 mm and 10-30 mm was 0.797 (95%CI: 0.707-0.887) and 0.809 (95%CI: 0.669-0.949), respectively, with sensitivity, specificity and accuracy reached 75% or above. Based on the analysis of nodule density, the AUC for distinguishing between benign and malignant solid nodule and subsolid nodule was 0.845 (95%CI: 0.751-0.939) and 0.790 (95%CI: 0.701-0.880), respectively, with sensitivity, specificity and accuracy reached 75% or above. Based on the analysis of nodule number, the AUC for distinguishing between benign and malignant solitary pulmonary nodule and multiple pulmonary nodule was 0.830 (95%CI: 0.696-0.965) and 0.817 (95%CI: 0.758-0.877), respectively, with sensitivity, specificity and accuracy reached 80% or above. Conclusion: The detection of EGFR gene amplification in peripheral blood rare cells contributes to the evaluation of benign and malignant pulmonary nodules, and can be used in the auxiliary diagnosis of benign and malignant pulmonary nodules.

Background: LDCT screening can significantly lower the mortality rate of lung cancer among high-risk individuals. Nevertheless, the limitations of CT might lead to frequent follow-up examinations and false positive outcomes, thereby causing unnecessary interventions and overtreatment. Therefore, the development of reliable and convenient biomarkers to accurately differentiate between benign and malignant nodules and to assess the likelihood of cancerous transformation is essential. We attempted to provide meaningful biomarkers based on plasma proteomic studies. Methods: The participants in this study were chosen from individuals aged 40 to 74 years in the Chinese Colorectal, Breast, Lung, Liver, And Stomach Cancer Screening Trial (C-BLAST). We selected 10 patients with malignant lung nodules and 10 with benign lung nodules, matched for age and sex. Malignant lung nodules were defined as those with a LUNG-RADS diagnostic category of 4A, 4B, or 4X, accompanied by a biopsy confirming malignancy; benign nodules were those with a diagnostic category not exceeding 3. Plasma samples from two groups were collected and subjected to proteomic analysis using the Somascan Assay 11k detection platform. Paired t-tests were employed to identify the differential proteins between malignant and benign pulmonary nodules. The functional pathways enriched by these proteins were determined based on Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. Finally, the STRING was utilized to construct a protein-protein interaction network (PPI) and determine the key proteins related to malignant nodules. Results: The average age of both groups was 61.4 years. A comparison of the proteomics between the malignant pulmonary nodule group and the benign pulmonary nodule group identified 188 differentially expressed proteins (P &amp;lt; 0.05), among which 102 were up-regulated proteins and 86 were down-regulated proteins. GO analysis of the differential proteins indicated functional enrichment in pathways such as chemical carcinogenesis, fluid shear stress and atherosclerosis, and biosynthesis of cofactors. According to KEGG analysis, they were mainly enriched in pathways like chemical carcinogenesis-reactive oxygen species, fluid shear stress and atherosclerosis, and metabolism of xenobiotics by cytochrome P450. Through PPI analysis, ten key proteins were determined, including CRP, FCGR3B, CCL2, CYP3A5, GSTA3, GSTM1, GSTM3, GSTM5, CD163, and GSTM4. These molecules possess anti-atherosclerotic and anti-inflammatory activities, as well as chemotactic activity for monocytes and basophils, and play roles in hydrolyzing nucleotides and host defense. Conclusions: Our research results provide ten potential plasma protein biomarkers for the discrimination of benign and malignant pulmonary nodules, which might broaden our understanding of their characteristics. Citation Format: Ximin Gao, Zhangyan Lyu, Guojin Si, Fengju Song. Proteomics unveil characteristic proteins of patients with benign and malignant pulmonary nodules [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1887.

Diagnostic Efficacy of PET-FDG Imaging in Solitary Pulmonary Nodules: Potential Role in Evaluation and Management

The clinical significance of a pulmonary nodule that is detected incidentally on CT studies in children is unknown. In addition, there is limited information regarding the management of incidentally detected pulmonary nodules discovered on abdominal CT studies in children. The purpose of this study was to investigate the clinical significance of incidental pulmonary nodules detected on abdominal CT studies in children. This was a retrospective study performed following institutional review board approval. Abdominal CT reports in patients younger than 18years of age from July 2004 to June 2011 were reviewed for the terms "nodule," "nodular" or "mass" in reference to the lung bases. The study population included those pediatric patients in whom pulmonary nodules were initially detected on abdominal CT studies. The largest pulmonary nodules detected on CT studies were evaluated for their features (size, shape, margin, attenuation, location, and presence of calcification and cavitation). Follow-up CT studies and clinical records were reviewed for demographic information, history of underlying malignancies and the clinical outcome of the incidental pulmonary nodules. Comparison of malignant versus benign pulmonary nodules was performed with respect to the size of the nodule, imaging features on CT, and patient history of malignancy using the Student's t-test and Fisher exact test. Youden J-index in receiver operating characteristic (ROC) analysis was used to determine the optimal cut-off size for suggesting a high risk of malignancy of incidentally detected pulmonary nodules. Pulmonary nodules meeting inclusion criteria were detected in 62 (1.2%) of 5,234 patients. The mean age of patients with nodules was 11.2years (range: 5months-18years). Thirty-one patients (50%) had follow-up CT studies and two of these patients (6%) were subsequently found to have malignant pulmonary nodules. Both of these patients had a history of malignancy. Of the remaining 31 patients without follow-up CT studies, none had a history of malignancy. Clinical follow-up data was available in 26 of these 31 patients (84%) and none had any evidence of malignant pulmonary nodule development. There was a significant association between history of malignancy and incidentally detected pulmonary nodules on abdominal CT studies subsequently found to be malignant (P = 0.036). The size was significantly larger for the malignant pulmonary nodules compared to the benign pulmonary nodules with a size ≥7mm in diameter being the optimal cut-off for suggesting a high risk of malignancy (11.5 ± 6.4mm vs. 4.7 ± 3.0mm, P = 0.003). The incidence of pulmonary nodules found on pediatric abdominal CT studies is 1.2%. The incidence of malignancy in such pulmonary nodules is low (3%) and only seen in the setting of pulmonary nodules ≥7mm in diameter in children with a history of malignancy. Therefore, further investigation is warranted for pulmonary nodules that are ≥7mm in children with a history of malignancy while further imaging work-up may not be necessary in the remaining patients in this pediatric patient population.

BackgroundIncidental indeterminate solitary solid pulmonary nodule is a progressively common finding on CT worldwide. Once detected, there are a number of imaging modalities that can be done to help in nodule characterization and differentiating benign from malignant nodules. Through these imaging modalities, there are PET CT, SPECT and dynamic CE-CT. Dynamic CE-CT is a functional test that help in assessment of the vascularity of the nodule which reverb the degree of angiogenesis of that nodule so can help in differentiating benign from malignant pulmonary nodules. The purpose of this study was to evaluate the role of Dynamic CE-CT in characterization of solitary pulmonary nodules. Detect what are the important parameters on dynamic CE-CT to differentiate benign from malignant nodules and detect their cutoff values.ResultsThe pre-enhancement value shows cutoff point of 26.50 HU with sensitivity 93.8% and specificity 75% with accuracy rate 90% in differentiating benign from malignant pulmonary nodules. Peak enhancement value (at 2 min) shows cutoff point of 40.00 HU with sensitivity 96.9% and specificity 87.5% with accuracy rate 95% in differentiating benign from malignant pulmonary nodules. Net enhancement value shows cutoff point of 19.00 HU with sensitivity 96.9% and specificity 87.5% with accuracy rate 95% in differentiating benign from malignant pulmonary nodules.ConclusionDynamic CE-CT is a useful tool in differentiating benign from malignant pulmonary nodules. Peak and net enhancement values are important parameters with high sensitivity and specificity in differentiating benign from malignant pulmonary nodules.

BackgroundAccurate differentiation between malignant and benign pulmonary nodules, especially those measuring 5–10 mm in diameter, continues to pose a significant diagnostic challenge. This study introduces a novel, precise approach by integrating circulating cell-free DNA (cfDNA) methylation patterns, protein profiling, and computed tomography (CT) imaging features to enhance the classification of pulmonary nodules.MethodsBlood samples were collected from 419 participants diagnosed with pulmonary nodules ranging from 5 to 30 mm in size, before any disease-altering procedures such as treatment or surgical intervention. High-throughput bisulfite sequencing was used to conduct DNA methylation profiling, while protein profiling was performed utilizing the Olink proximity extension assay. The dataset was divided into a training set and an independent test set. The training set included 162 matched cases of benign and malignant nodules, balanced for sex and age. In contrast, the test set consisted of 46 benign and 49 malignant nodules. By effectively integrating both molecular (DNA methylation and protein profiling) and CT imaging parameters, a sophisticated deep learning-based classifier was developed to accurately distinguish between benign and malignant pulmonary nodules.ResultsOur results demonstrate that the integrated model is both accurate and robust in distinguishing between benign and malignant pulmonary nodules. It achieved an AUC score 0.925 (sensitivity = 83.7%, specificity = 82.6%) in classifying test set. The performance of the integrated model was significantly higher than that of individual methylation (AUC = 0.799, P = 0.004), protein (AUC = 0.846, P = 0.009), and imaging models (AUC = 0.866, P = 0.01). Importantly, the integrated model achieved a higher AUC of 0.951 (sensitivity = 83.9%, specificity = 89.7%) in 5–10 mm small nodules. These results collectively confirm the accuracy and robustness of our model in detecting malignant nodules from benign ones.ConclusionsOur study presents a promising noninvasive approach to distinguish the malignancy of pulmonary nodules using multiple molecular and imaging features, which has the potential to assist in clinical decision-making.Trial registration: This study was registered on ClinicalTrials.gov on 01/01/2020 (NCT05432128). https://classic.clinicaltrials.gov/ct2/show/NCT05432128.

Classification of malignant and benign pulmonary nodules is important for further treatment plan. The present work focuses on the classification of benign and malignant pulmonary nodules using support vector machine. The pulmonary nodules are segmented using a semi-automated technique, which requires only a seed point from the end user. Several shape-based, margin-based, and texture-based features are computed to represent the pulmonary nodules. A set of relevant features is determined for the efficient representation of nodules in the feature space. The proposed classification scheme is validated on a data set of 891 nodules of Lung Image Database Consortium and Image Database Resource Initiative public database. The proposed classification scheme is evaluated for three configurations such as configuration 1 (composite rank of malignancy "1" and "2" as benign and "4" and "5" as malignant), configuration 2 (composite rank of malignancy "1","2", and "3" as benign and "4" and "5" as malignant), and configuration 3 (composite rank of malignancy "1" and "2" as benign and "3","4" and "5" as malignant). The performance of the classification is evaluated in terms of area (A z) under the receiver operating characteristic curve. The A z achieved by the proposed method for configuration-1, configuration-2, and configuration-3 are 0.9505, 0.8822, and 0.8488, respectively. The proposed method outperforms the most recent technique, which depends on the manual segmentation of pulmonary nodules by a trained radiologist.

The aim of this prospective study was not to describe individual morphological findings in benign and malignant solitary intrapulmonary nodules; it was instead to examine in a critical manner the indications for differentiation found in the literature in order to facilitate safe differential diagnosis of benign and malignant nodules. A total of 64 solitary pulmonary nodules were examined with high-resolution computed tomography and correlated with histological findings. Only lesions that had been removed by surgery were used. No lesion was excluded on the grounds of size. Useful characteristics for the differentiation of benign from malignant pulmonary nodules were: diameter and density of the lesion, air inclusion, unsharp and dystelectatic margin, the presence of spicules, length of spicules, spicules extending to the visceral pleura, pleural tail sign and cirumscribed pleural thickening. For the differentiation of benign and malignant solitary pulmonary nodules meticulous assessment of the margin of the nodule is necessary. Using the criteria mentioned, a sensitivity of 85% and a specifity of 78% can be achieved for the identification of malignant pulmonary nodules. Since it was not possible to differentiate between benign and malignant nodules with certainty using imaging methods, the chance of patient survival could only be promoted by early surgery.

Background: Advancements in technology have significantly improved the diagnosis of solitary pulmonary nodules in the lungs. Various computed tomography (CT) imaging techniques, including modern dual-energy computed tomography (DECT), have enhanced the ability to accurately classify pulmonary nodules as benign or malignant. In this study, three different dual-energy parameters — iodine load, contrast load, and visual assessment — were evaluated for their potential in characterizing pulmonary nodules. Objectives: The aim of this study was to assess the reliability and effectiveness of DECT in distinguishing benign from malignant pulmonary nodules using different parameters, including visual assessment, iodine concentration, and contrast load. Patients and Methods: This prospective study included patients who underwent contrast-enhanced thoracic DECT for solitary pulmonary nodules, had histopathological examination results, or had at least a two-year follow-up CT scan. Patients with nodules smaller than 6 mm or completely calcified nodules were excluded. Patients diagnosed with a suspicious solitary pulmonary nodule on chest radiography and subsequently underwent contrast-enhanced DECT, or those diagnosed with a lung nodule on routine non-contrast CT scans and later evaluated using DECT, were included in the study. Benign and malignant nodules were compared based on gender, age, contrast load, iodine load, and color map assessment. Nodule images were obtained 40 seconds after intravenous contrast administration using single-source DECT (120 kV split filter) with twin-beam technology. The visual enhancement and color map evaluation, including contrast and iodine load measurements, were separately calculated and recorded for each lung nodule. Results: A total of 59 patients [30 males (50.8%) and 29 females (49.2%)] with a solitary pulmonary nodule met the inclusion criteria. Among the 59 pulmonary nodules, 16 (27.1%) were malignant, and 43 (72.9%) were benign. Of the benign lesions, 23 (53.5%) were found in males and 20 (46.5%) in females. The mean age of patients with benign nodules was 53.5 ± 12 years (range: 25 - 73 years), while for those with malignant nodules, it was 69.2 ± 5.59 years (range: 57 - 75 years). There was no statistically significant difference in age between the two groups (P = 0.506). The median contrast load was 0.0 Hounsfield units (HU) [interquartile range (IQR: 64)] in benign nodules and 63 HU (IQR: 154) in malignant nodules. Malignant nodules had a significantly higher contrast load than benign nodules (P = 0.003). Using a cut-off value of 22 HU for contrast load in malignancy diagnosis, the sensitivity was 100%, specificity was 58.14%, positive predictive value (PPV) was 47.06%, and negative predictive value (NPV) was 100%. The area under the curve (AUC) was 0.746. The median iodine load was 0.0 mg/dL (IQR: 4.5) in benign nodules and 4.5 mg/dL (IQR: 11.8) in malignant nodules. Malignant nodules had a significantly higher iodine load than benign nodules (P &lt; 0.001). Using a cut-off value of 1 mg/mL for malignancy diagnosis, the sensitivity was 100%, specificity was 62.79%, PPV was 50%, and NPV was 100% (AUC: 0.768). Conclusion: Dual-energy computed tomography provides valuable contributions in differentiating benign and malignant pulmonary nodules. In this study, the diagnostic value of three different approaches — visual iodine coverage color map, iodine concentration, and contrast load — was demonstrated in distinguishing these lesions.

Under the background of urgent need for computer-aided technology to provide physicians with objective decision support, aiming at reducing the false positive rate of nodule CT detection in pulmonary nodules detection and improving the accuracy of lung nodule recognition, this paper puts forward a method based on ensemble learning to distinguish between malignant and benign pulmonary nodules. Firstly, trained on a public data set, a multi-layer feature fusion YOLOv3 network is used to detect lung nodules. Secondly, a CNN was trained to differentiate benign from malignant pulmonary nodules. Then, based on the idea of ensemble learning, the confidence probability of the above two models and the label of the training set are taken as data features to build a Logistic regression model. Finally, two test sets (public data set and private data set) were tested, and the confidence probability output by the two models was fused into the established logistic regression model to determine benign and malignant pulmonary nodules. The YOLOv3 network was trained to detect chest CT images of the test set. The number of pulmonary nodules detected in the public and private test sets was 356 and 314, respectively. The accuracy, sensitivity and specificity of the two test sets were 80.97%, 81.63%, 78.75% and 79.69%, 86.59%, 72.16%, respectively. With CNN training pulmonary nodules benign and malignant discriminant model analysis of two kinds of test set, the result of accuracy, sensitivity and specificity were 90.12%, 90.66%, 89.47% and 88.57%, 85.62%, 90.87%, respectively. Fused model based on YOLOv3 network and CNN is tested on two test sets, and the result of accuracy, sensitivity and specificity were 93.82%, 94.85%, 92.59% and 92.31%, 92.68%, 91.89%, respectively. The ensemble learning model is more effective than YOLOv3 network and CNN in removing false positives, and the accuracy of the ensemble. Learning model is higher than the other two networks in identifying pulmonary nodules.

Objective To analyze retrospectively the clinical data of solitary pulmonary nodules and summarize the clinical characteristics to provide evidences for the diagnosis of benign and malignant solitary pulmonary nodules. Methods 148 cases of solitary pulmonary nodules were selected in the Northern Jiangsu People′s Hospital from January 2016 to June 2018.All the lesions of these patients were confirmed by surgery, bronchoscopy or lung puncture and had definite pathological diagnosis.The clinical data were collected to judge benign and malignant solitary pulmonary nodules, including gender, age, smoking history, cancer history, family history of cancer, maximum diameter of nodule, location, lobulation, spicule, pleural indentation, vascular convergence sign, tumor markers.Single factor binary regression was performed in the univariate analysis and logistic regression in multivariate analysis of these clincial data. Results 148 cases collected included 40 benign nodules and 108 malignant nodules.The univariate analysis showed that there were statistical differences in gender, age, cancer history, maximum diameter of nodule, location, lobulation, spicule, pleural indentation, vascular convergence sign between benign and malignant pulmonary nodules (all P<0.05). Multivariate logistic regression analysis showed that there were statistical differences in gender, age, maximum diameter of nodule and vascular convergence sign between benign and malignant pulmonary nodules (all P<0.05). Conclusions Patient gender, age, maximum diameter of nodule, vascular convergence sign are independent predictors of malignance in patients with solitary pulmonary nodules. Key words: Solitary pulmonary nodule; Adenocarcinoma; Risk factors; Logistic regression analysis

The ability to differentiate between benign, suspicious, and malignant pulmonary nodules is imperative for definitive intervention in patients with early stage lung cancers. Here, we report that plasma protein functional effector sncRNAs (pfeRNAs) serve as non-invasive biomarkers for determining both the existence and the nature of pulmonary nodules in a three-stage study that included the healthy group, patients with benign pulmonary nodules, patients with suspicious nodules, and patients with malignant nodules. Following the standards required for a clinical laboratory improvement amendments (CLIA)-compliant laboratory-developed test (LDT), we identified a pfeRNA classifier containing 8 pfeRNAs in 108 biospecimens from 60 patients by sncRNA deep sequencing, deduced prediction rules using a separate training cohort of 198 plasma specimens, and then applied the prediction rules to another 230 plasma specimens in an independent validation cohort. The pfeRNA classifier could (1) differentiate patients with or without pulmonary nodules with an average sensitivity and specificity of 96.2% and 97.35% and (2) differentiate malignant versus benign pulmonary nodules with an average sensitivity and specificity of 77.1% and 74.25%. Our biomarkers are cost-effective, non-invasive, sensitive, and specific, and the qPCR-based method provides the possibility for automatic testing of robotic applications.

ES 02.02 The Fleischner Guideline / Lung-RADs

Classification of benign and malignant pulmonary nodules can provide useful indicators for estimating the risk of lung cancer. In this study, an improved random forest (RF) algorithm is proposed for classification of benign and malignant pulmonary nodules in thoracic computed tomography images. First, an improved random walk algorithm is proposed to automatically segment pulmonary nodules. Then, intensity, geometric and texture features based on the grey‐level co‐occurrence matrix, rotation invariant uniform local binary pattern and Gabor filter methods are combined to generate an effective and discriminative feature vector. Mutual information is employed to reduce the dimensionality. Finally, an improved RF classifier is trained to classify benign and malignant nodules. An appropriate feature subset is selected by the bootstrap method and an effective combination method is introduced to predict a class label. The proposed classification method on the lung images dataset consortium dataset achieves a sensitivity of 0.92 and the area under the receiver‐operating‐characteristic curve of 0.95. An additional evaluation is performed on another dataset coming from General Hospital of Guangzhou Military Command. A mean sensitivity and a mean specificity of the proposed method are 0.85 and 0.82, respectively. Experimental results demonstrate that the proposed method achieves the satisfactory classification performance.