Prediction of Pathological Subtypes of Lung Adenocarcinoma with Pure Ground Glass Nodules by Deep Learning Model / 中国医学科学院学报

To make a preliminary pathological classification of lung adenocarcinoma with pure ground glass nodules(pGGN)on CT by using a deep learning model. CT images and pathological data of 219 patients(240 lesions in total)with pGGN on CT and pathologically confirmed adenocarcinoma were collected.According to pathological subtypes,the lesions were divided into non-invasive lung adenocarcinoma group(which included atypical adenomatous hyperplasia and adenocarcinoma in situ and micro-invasive adenocarcinoma)and invasive lung adenocarcinoma group.First,the lesions were outlined and labeled by two young radiologists,and then the labeled data were randomly divided into two datasets:the training set(80%)and the test set(20%).The prediction Results of deep learning were compared with those of two experienced radiologists by using the test dataset. The deep learning model achieved high performance in predicting the pathological types(non-invasive and invasive)of pGGN lung adenocarcinoma.The accuracy rate in pGGN diagnosis was 0.8330(95% =0.7016-0.9157)for of deep learning model,0.5000(95% =0.3639-0.6361)for expert 1,0.5625(95% =0.4227-0.6931)for expert 2,and 0.5417(95% =0.4029-0.6743)for both two experts.Thus,the accuracy of the deep learning model was significantly higher than those of the experienced radiologists(=0.002).The intra-observer agreements were good(Kappa values:0.939 and 0.799,respectively).The inter-observer agreement was general(Kappa value:0.667)(=0.000). The deep learning model showed better performance in predicting the pathological types of pGGN lung adenocarcinoma compared with experienced radiologists.

Correlations between Vessel Changes and the Histopathologic Subtypes of Lung Adenocarcinoma with Pure Ground-glass Nodule on Computed Tomography / 中国医学科学院学报

<p><b>OBJECTIVE</b>To investigate the correlations between vessel changes and the histopathologic subtypes of lung adenocarcinoma with pure ground-glass nodule (pGGN) on computed tomography (CT).</p><p><b>METHODS</b>Totally 107 patients (116 lesions) with lung adenocarcinomas with pGGN who had undergone curative resection were included. Vessel changes included vascular convergence and/or vessel dilation or distortion within the pGGN. According to the vessel appearances within the pGGN, all patients were categorized into two groups: no change group and change group. Pearson chi-square test was used to analyze the relationships between vessel changes and histopathologic subtypes. Mann-Whitney rank test and t-test were used to identify the relationship of vessels changes with pGGN density and diameter.</p><p><b>RESULTS</b>Among these 116 pGGNs, there were 21 without vessel changes and 4 with vessel changes in 25 preinvasive lesions; 14 without vessel changes and 15 with vessel changes in 29 minimally invasive adenocacinomas; 16 without vessel changes and 46 with vessel changes in 62 invasive lung adenocarcinomas. There were statistically significant differences of vessel changes (P=0.000) among histopathologic subtypes. The lesion diameter was significantly different between these two groups (P=0.000), while the lesion density showed no significant difference (P=0.826).</p><p><b>CONCLUSION</b>Vessel changes may indicate the invasiveness of lung adenocarcinoma with pGGNs and are related with the lesion diameter.</p>

Value of Baseline Clinical and CT Characteristics for Predicting the Progression of Persistent Pure Ground-glass Nodule 10 mm or Less in Diameter / 中国医学科学院学报

Objective To explore the risk factors of the progression of persistent pure ground-glass nodule (pGGN) and make the risk stratification for pGGN 10 mm or less in diameter. Methods From June 2008 to April 2015,100 patients (108 lesions) with persistent pGGN≤10 mm in diameter were included in this study. Patients were followed up at least 1 year using thin-section computed tomography (CT). Patients' baseline clinical data and CT characteristics of pGGN were compared between progression group (size increased or/and solid component appeared) and non-progression group. Cox regression analysis was used to assess the relationship between clinical data,CT characteristics of pGGN,and lesion progression. The risk indices of lesion progression were calculated according to the results of Cox regression analysis and the relative factors of lung adenocarcinoma in previous studies. Logistic regression analysis was used to assess the relationship between risk indices and lesion progression. The optimal cutoff value was decided on receiver operating characteristic curve of risk indices and verified for predicting lesion progression. Results Fifteen of 108 lesions showed progression. The mean follow-up duration was (1016.36±486.00) days. There were statistically significant differences of lesion size,air bronchogram,and vessel changes between progression group and non-progression group (P=0.040,P=0.003,P=0.030,respectively).Lesion density (CT value≥-542.5 HU) and air bronchogram were the risk factors of lesion progression (P=0.003,P=0.021,respectively). The optimal cutoff value of total risk indices on predicting lesion progression was 4.25,with the sensitivity of 46.7%,specificity of 89.2%,and consistency of 83.3%. Conclusions CT value ≥-542.5 HU of pGGN and air bronchogram within lesion may predict lesion progression in persistent pGGN 10 mm or less in diameter. A risk index of less than 4.25 often suggests small probability of disease progression and thus a longer follow-up interval is recommended.

Correlations between pathologic subtypes/immunohistochemical implication and CT characteristics of lung adenocarcinoma ≤ 1 cm with ground-glass opacity / 中国医学科学院学报

<p><b>OBJECTIVE</b>To discuss the correlation of pathologic subtypes and immunohistochemical implication with CT features of lung adenocarcinoma 1 cm or less in diameter with focal ground-glass opacity (fGGO).</p><p><b>METHODS</b>CT appearances of 59 patients who underwent curative resection of lung adenocarcinoma ≤ 1 cm with fGGO were analyzed in terms of lesion location, size, density, shape (round, oval, polygonal, irregular), margin (smooth, lobular, spiculated, lobular and spiculated), bubble-like sign, air bronchogram, pleural tag, and tumor-lung interface. Histopathologic subtypes were classified according to International Association for the Study of Lung Cancer/ American Thoracic Society/European Respiratory Society classification of lung adenocarcinoma. Common molecular markers in immunohistochemical study included human epidermal growth factor receptor (HER)-1,HER-2,Ki-67, vascular endothelial growth factor (VEGF) and DNA topoisomerase 2Α.Patients' age and lesions' size and density were compared with pathologic subtypes using analysis of variance or nonparametric Wilcoxon tests. Patients' gender, lesion location, shape and margin, bubble-like sign, air bronchogram, pleural tag, and tumor-lung interface were compared with histopathologic subtypes and immunohistochemical implication using ψ² test or Fisher's exact test.</p><p><b>RESULTS</b>The patients' gender, age, lesion location, shape, air bronchogram, pleural tag, and tumor-lung interface were not significantly different among different histopathologic subtypes (P=0.194, 0.126, 0.609, 0.678, 0.091, 0.374, and 0.339, respectively), whereas the lesion size,density,bubble-like sign, and margin showed significant differences (P=0.028, 0.002, 0.003, 0.046, respectively). The expression of Ki-67 significantly differed among nodules with different shapes(P=0.015). Statistically significant difference also existed between tumor-lung interface and HER-1 expression (P=0.019) and between bubble sign and HER-2 expression (P=0.049).</p><p><b>CONCLUSIONS</b>Of lung adenocarcinoma ≤ 1 cm with fGGO,bubble-like sign occurs more frequently in invasive pulmonary adenocarcinoma and less frequently in atypical adenomatous hyperplasia. In addition, preinvasive lesions (atypical adenomatous hyperplasia and adenocarcinoma in situ) more frequently demonstrates smooth margin,while invasive lesions (minimally invasive adenocarcinoma and invasive pulmonary adenocarcinoma) more frequently demonstrates lobular and spiculated margin. Some CT features are associated with immunohistochemical implication of lung adenocarcinoma ≤ 1 cm with fGGO.</p>

CT characteristics of primary salivary gland-type lung cancer / 中华肿瘤杂志

<p><b>OBJECTIVE</b>To improve the diagnostic accuracy of primary salivary gland-type lung cancer on CT.</p><p><b>METHODS</b>The CT findings of 13 pathologically proven primary salivary gland-type lung cancers (mucoepidermoid carcinoma, n = 8, adenoid cystic carcinoma, n = 5) were retrospectively analyzed.</p><p><b>RESULTS</b>Three mucoepidermoid carcinomas were located in the main bronchus, 4 in segmental bronchus, and 1 in peripheral lung. Intrabronchial nodule or mass with smooth or lobulated margin and punctuate or strip calcification (n = 2) was the main CT feature. The tumor showed moderate to significant enhancement after the administration of contrast medium. Three adenoid cystic carcinomas involved trachea, and 2 involved the main and lobular bronchi. The main CT features were diffuse or circumferential irregular thickness of the wall, distorted lumen, and nodule protruding into the lumen, and the longitudinal extent of the tumor was greater than its transverse axis.</p><p><b>CONCLUSION</b>The CT findings of primary salivary gland-type lung cancer are rather specific and may provide helpful information for the clinical diagnosis and treatment.</p>

Assessment of the anterior spinal artery and the artery of Adamkiewicz using multi-detector CT angiography / 中华医学杂志(英文版)

<p><b>BACKGROUND</b>Damage to the spinal cord after the treatment of the descending thoracic and thoracoabdominal aortic aneurysms is an uncommon but devastating complication. The artery of Adamkiewicz (AKA) is the principal arterial supply of the anterior spinal artery (ASA) in the lower thoracic and lumbar level. The purpose of this study was to evaluate the visualization of the anterior spinal artery and the artery of Adamkiewicz, the affecting factors for the detection rate using multi-detector row CT (MDCT).</p><p><b>METHODS</b>Ninety-nine consecutive patients (31 women and 68 men; age range, 25 - 90 years; average age 61.3 years), with suspicion for thoracic aortic lesions necessitating surgical intervention (31 aortic aneurysm, 45 dissection, 5 intramural hematoma, and 18 normal), underwent CT angiography from the aortic arch to the aortic bifurcation. Transverse sections, multiplanar reformations and thin maximum intensity projections were used to assess the ASA and AKA. The level of the ASA and AKA origins and CT acquisition parameters were recorded. The contrast-to-noise ratio of the image, an index of the mass of the T11 body (vertebral mass index), the subcutaneous fat thickness, and the CT value within the aortic arch and at the T11 level were measured. The detection of the ASA and AKA were evaluated relative to the acquisition parameters, scan characteristics, and aortic lesion type. Differences were assessed with the Wilcoxon rank-sum and t tests.</p><p><b>RESULTS</b>The ASA was visualized in 51 patients (52%) and the AKA in 18 patients (18%). The ASA was identified in 36/67 patients (54%) with 1.25 mm thickness and in 15/32 patients (47%) with 2.5 - 3.0 mm thickness. This difference did not achieve significance (P = 0.13). The detection rate of the ASA and the AKA was influenced by the vertebral mass index and the contrast-to-noise ratio (P < 0.05). The amount of subcutaneous fat affected the detection rate of the ASA (P < 0.05) but not the AKA. In CT scans of ASA detection, the mean CT values in the aorta at the arch and at T11 were 360 and 358 HU, respectively, whereas in CT scans without ASA detection, the CT values in the aorta at the arch and at T11 were lower (P < 0.05), 297 and 317 HU, respectively.</p><p><b>CONCLUSIONS</b>The ASA and AKA were less frequently detected in our cohort than previous reports. The visualization of the ASA and AKA was significantly affected by aortic enhancement, the "vertebral mass index", and the contrast-to-noise ratio.</p>

Evaluation of coronary artery in-stent restenosis using 64-slice CT / 中华放射学杂志

Objective To evaluate the accuracy of 64-slice CT in the diagnosis of coronary in-stent restenosis (ISR).Methods Fourteen patients with 37 implanted coronary stents were examined by both 64-slice CT angiography (CTA) and conventional coronary angiography(CCA).The diagnosis of ISR was evaluated by two methods(visual inspection and the measurement of the in-stent contrast attenuation)on CTA.The accuracy of the two methods in the diagnosis of ISR was compared with reference to CCA.Results ISR(＞50%)was found on CCA in 11 stents.CTA with visual inspection and with measurement of in-stent CT attenuation correctly detected ISR in 2 and 3 stents respectively with reference to CCA.The sensitivity, specificity,positive predictive value and negative predicitive value of the two methods were 18%,69%, 20%,67% and 27%,81%,38%,72%,respectively.There was no significant difference (P＞0.05) between the accuracy of two methods.Conclusion The ISR was very difficult to diagnose by 64 slice CT, but the high specificity of 64-slice CT study implied an important role in excluding ISR.

CT perfusion of pulmonary carcinoma:the correlative study with fluoro-deoxyglucose positron emission tomography and tumor microvessel density / 中华放射学杂志

Objective To investigate the correlation between CT perfusion parameters of pulmonary carcinoma and standardized uptake values(SUV)derived from ~(18)F-fluoro-deoxyglucose positron emission tomography(~8F-FDG PET)and tumor microvessel density(MVD),and to determine the validity of CT perfusion in assessing tumor angiagenic activity of pulmonary carcinoma.Methods Fifty patients(mean age 57.5,17 females)with pulmonary carcinoma underwent CT perfusion using 16-slice helical CT.Blood flow(BF,ml?100g~(-1)?min~(-1)),blood volume(BV,ml?100g~(-1)),mean transmit time(MTF,s)and permeability surface area product(PS,ml?100g~(-1)?min~(-1))were analyzed.SUV of PET was calculated in 14 patients.The CD34 immunohistochemical staining was used for tumor microvessel counting.CT perfusion parameters of pulmonary carcinoma were correlatively studied with SUV and tumor MVD.Pearson's correlation analysis was performed to evaluate the association between CT perfusion parameters and SUV and MVD.Results The average values of BF,BV,MTT and PS were 97.30 ml?100g~(-1)?min~(-1), 8.86 ml?100g~(-1),6.75 s and 34.52 ml?100g~(-1)?min~(-1),respectively.The average value of MVD was 61.82/FOV.The mean value of SUV was 5.96.There was positive correlation between BF and SUV(r= 0.727,P