Lijuan Yu

The method and efficacy of support vector machine classifiers based on texture features and multi-resolution histogram from 18F-FDG PET-CT images for the evaluation of mediastinal lymph nodes in patients with lung cancer

OBJECTIVES In clinical practice, image analysis is dependent on simply visual perception and the diagnostic efficacy of this analysis pattern is limited for mediastinal lymph nodes in patients with lung cancer. In order to improve diagnostic efficacy, we developed a new computer-based algorithm and tested its diagnostic efficacy. METHODS 132 consecutive patients with lung cancer underwent (18)F-FDG PET/CT examination before treatment. After all data were imported into the database of an on-line medical image analysis platform, the diagnostic efficacy of visual analysis was first evaluated without knowing pathological results, and the maximum short diameter and maximum standardized uptake value (SUVmax) were measured. Then lymph nodes were segmented manually. Three classifiers based on support vector machine (SVM) were constructed from CT, PET, and combined PET-CT images, respectively. The diagnostic efficacy of SVM classifiers was obtained and evaluated. RESULTS According to ROC curves, the areas under curves for maximum short diameter and SUVmax were 0.684 and 0.652, respectively. The areas under the ROC curve for SVM1, SVM2, and SVM3 were 0.689, 0.579, and 0.685, respectively. CONCLUSION The algorithm based on SVM was potential in the diagnosis of mediastinal lymph nodes.

Comparison of machine learning methods for classifying mediastinal lymph node metastasis of non-small cell lung cancer from 18F-FDG PET/CT images

BackgroundThis study aimed to compare one state-of-the-art deep learning method and four classical machine learning methods for classifying mediastinal lymph node metastasis of non-small cell lung cancer (NSCLC) from 18F-FDG PET/CT images. Another objective was to compare the discriminative power of the recently popular PET/CT texture features with the widely used diagnostic features such as tumor size, CT value, SUV, image contrast, and intensity standard deviation. The four classical machine learning methods included random forests, support vector machines, adaptive boosting, and artificial neural network. The deep learning method was the convolutional neural networks (CNN). The five methods were evaluated using 1397 lymph nodes collected from PET/CT images of 168 patients, with corresponding pathology analysis results as gold standard. The comparison was conducted using 10 times 10-fold cross-validation based on the criterion of sensitivity, specificity, accuracy (ACC), and area under the ROC curve (AUC). For each classical method, different input features were compared to select the optimal feature set. Based on the optimal feature set, the classical methods were compared with CNN, as well as with human doctors from our institute.ResultsFor the classical methods, the diagnostic features resulted in 81~85% ACC and 0.87~0.92 AUC, which were significantly higher than the results of texture features. CNN’s sensitivity, specificity, ACC, and AUC were 84, 88, 86, and 0.91, respectively. There was no significant difference between the results of CNN and the best classical method. The sensitivity, specificity, and ACC of human doctors were 73, 90, and 82, respectively. All the five machine learning methods had higher sensitivities but lower specificities than human doctors.ConclusionsThe present study shows that the performance of CNN is not significantly different from the best classical methods and human doctors for classifying mediastinal lymph node metastasis of NSCLC from PET/CT images. Because CNN does not need tumor segmentation or feature calculation, it is more convenient and more objective than the classical methods. However, CNN does not make use of the import diagnostic features, which have been proved more discriminative than the texture features for classifying small-sized lymph nodes. Therefore, incorporating the diagnostic features into CNN is a promising direction for future research.

The method and efficacy of 18F-fluorodeoxyglucose positron emission tomography/computed tomography for diagnosing the lymphatic metastasis of colorectal carcinoma.

RATIONALE AND OBJECTIVES The aim of this study was to analyze the morphologic and functional features of metastatic lymph nodes of colorectal carcinoma on (18)F-fluorodeoxyglucose positron emission tomographic (PET)/computed tomographic (CT) images and correlate these with pathologic results to explore the best diagnostic performance. METHODS Sixty-eight patients without any previous treatment underwent (18)F-fluorodeoxyglucose PET/CT examinations and subsequent operations. All lymph nodes were evaluated by recording short diameter and maximum standardized uptake value (SUVmax) on axial images and were carefully verified on the surgically resected specimens. The radiologic diagnostic efficacies on the basis of different diagnostic criteria were compared and evaluated with pathologic results. RESULTS There was a significant difference for SUVmax between metastatic and benign juxtaintestinal lymph nodes (F = 96.836, P = .000) and a correlation between size and SUVmax in metastatic juxtaintestinal lymph nodes (r = 0.352, P = .038). Diagnosing according to short diameter ≥ 10 mm and SUVmax ≥ 2.5, the sensitivity, specificity, positivity prediction value (PPV), and negative prediction value (NPV) were 10.00%, 98.26%, 66.67%, and 75.84% and 82.50%, 90.43%, 75.00%, and 93.69%, respectively. Considered together, the sensitivity, specificity, PPV, and NPV were 10.00%, 99.13%, 80.00%, and 76.00%, respectively. Receiver-operating characteristic curves showed that the best cutoff values for SUVmax and short diameter were 2.0 and 4.85 mm, respectively; the corresponding sensitivity, specificity, PPV, and NPV, were 91.43%, 87.83%, 69.57%, and 97.12% and 85.71%, 60.87%, 40.00%, and 98.2%, respectively. Considered together, the sensitivity, specificity, PPV, and NPV were 95.00%, 86.96%, 71.70%, and 98.04%, respectively. CONCLUSIONS Fluorine-18-fluorodeoxyglucose uptake was a more reliable indicator in diagnosing malignant juxtaintestinal lymph node of colorectal carcinoma. The optimal diagnostic efficacy could be reached by considering morphologic and functional features together.

Evaluation of the diagnostic performance of magnetic resonance spectroscopy in brain tumors: a systematic review and meta-analysis.

Object The aim of this study was to determine the suitability of magnetic resonance spectroscopy (MRS) for screening brain tumors, based on a systematic review and meta-analysis of published data on the diagnostic performance of MRS. Methods The PubMed and PHMC databases were systematically searched for relevant studies up to December 2013. The sensitivities and specificities of MRS in individual studies were calculated and the pooled diagnostic accuracies, with 95% confidence intervals (CI), were assessed under a fixed-effects model. Results Twenty-four studies were included, comprising a total of 1013 participants. Overall, no heterogeneity of diagnostic effects was observed between studies. The pooled sensitivity and specificity of MRS were 80.05% (95% CI = 75.97%–83.59%) and 78.46% (95% CI: 73.40%–82.78%), respectively. The area under the summary receiver operating characteristic curve was 0.78. Stratified meta analysis showed higher sensitivity and specificity in child than adult. CSI had higher sensitivity and SV had higher specificity. Higher sensitivity and specificity were obtained in short TE value. Conclusion Although the qualities of the studies included in the meta-analysis were moderate, current evidence suggests that MRS may be a valuable adjunct to magnetic resonance imaging for diagnosing brain tumors, but requires selection of suitable technique and TE value.

Correlations between SUVmax and Expression of GLUT1 and Growth Factors Inducing Lymphangiogenesis

RATIONALE AND OBJECTIVES The purpose of this study was to assess the correlations between the maximum standardized uptake value (SUVmax) of colorectal carcinoma and hepatocyte growth factor (HGF), vascular endothelial growth factor C (VEGF-C), and their respective receptors using (18)F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT). METHODS Fluorine-18-FDG PET/CT scans were performed on 33 patients with colorectal carcinoma before any treatment. The SUVmax of colorectal carcinoma and the clinicopathologic data associated with lymphatic metastases were analyzed. The expression of glucose transporter 1 (GLUT1), HGF, c-Met, VEGF-C, and vascular endothelial growth factor receptor 3 (VEGFR-3) in tumor tissues was analyzed using immunohistochemical methods. Lymphatic endothelial cells were marked with D2-40, and lymphatic vessel density (LVD) was recorded. The correlations were analyzed among the SUVmax of colorectal carcinoma, LVD, and the expression of GLUT1, HGF, c-Met, VEGF-C, and VEGFR-3 in tumor tissues. RESULTS SUVmax and LVD in 15 patients with lymphatic metastases were 13.00 ± 4.51 and 6.25 ± 1.54, respectively, whereas in 18 patients with nonmetastatic nodes, SUVmax and LVD were 9.66 ± 4.82 and 4.54 ± 1.02, respectively. The differences of SUVmax and LVD between metastatic and nonmetastatic patients were statistically significant (F = 4.153, P = .025, and F = 14.501, P = .001, respectively). There were no statistical differences of SUVmax and LVD in variably differentiated colorectal carcinoma (F = 0.708, P = .502, and F = 0.311, P = .735, respectively). The expression rates of GLUT1 in neoplastic and normal tissue were 72.7% (24 of 33) and 21.2% (seven of 33), respectively (P = .001). Moreover, the expression rates of GLUT1 in metastatic and nonmetastatic tissue were 93.33% (14 of 15) and 61.11% (11 of 18), respectively (P = .038). LVD and the integrated optical density of GLUT1 were 5.31 ± 1.53 and 8.21 × 10(4) ± 4.30 × 10(4), respectively, in tumor tissue, and there were linear correlations between SUVmax and LVD (r = 0.373, P = .033) and between SUVmax and expression of GLUT1 (r = 0.428, P = .013). The differences of SUVmax in HGF, c-Met, and VEGF-C groups with different expressions were statistically significant (P = .007, P = .009, and P = .030, respectively). No correlation was found between the expression of VEGFR-3 and SUVmax. The expression of GLUT1 and HGF as well as of GLUT1 and VEGF-C was rank correlated (r = 0.521, P = .002, and r = 0.505, P = .003, respectively). No rank correlations were found between the expression of GLUT1 and c-Met, GLUT1, and VEGFR-3. CONCLUSIONS The SUVmax of colorectal carcinoma was significantly higher in metastatic patients; the uptake of colorectal carcinoma was associated with LVD and the expression of HGF and VEGF-C but not with the expression of VEGFR-3.

Evaluation of18F-FDG PET/CT in Early-Stage Cervical Carcinoma

Introduction:The purpose is to evaluate 18F-fluorodeoxy glucose-positron emission tomography (PET)/computed tomography (CT) for the detection of cervical carcinoma of the International Federation of Gynecology and Obstetrics stages Ib and IIa. Methods:Twenty-two patients with cervical carcinoma were included in this study. Eleven of them were at the International Federation of Gynecology and Obstetrics stage Ib, and the remaining patients were at stage IIa. All the patients underwent 18F-fluorodeoxy glucose-PET/CT scan, and in 15 patients, additional delayed imaging of pelvis was performed. The PET/CT findings were compared with histopathologic and follow-up results. Results:Of the 22 patients, 20 had cervical cancer with abnormal uptake that were detected directly using PET/CT without additional information from delayed imaging of pelvis; the other 2 patients showed lesions with no abnormal uptake. The use of delayed imaging of the pelvis with PET/CT corrected 1 false-negative case. Thus, the accuracy of PET/CT for early-stage cervical carcinoma was 95.45% (21/22). The uptake of early-stage cervical carcinoma showed no statistical significance between the group of stage Ib and the group of stage IIa (P > 0.05). The false-negative rate and negative predictive value of PET/CT imaging in predicting pelvic nodal involvement were 6.25% (1/16) and 93.75%, respectively. Conclusions:PET/CT showed a high accuracy for the diagnosis of early-stage cervical cancer and a high-negative predictive value for predicting the pelvic nodal status in early-stage cervical cancer.

Prognostic Value of Baseline 18F-FDG PET/CT Functional Parameters in Patients with Advanced Lung Adenocarcinoma Stratified by EGFR Mutation Status.

The study objective was to retrospectively analyze the metabolic variables derived from 18 F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) as predictors of progression-free survival (PFS) and overall survival (OS) in advanced lung adenocarcinoma stratified by epidermal growth factor receptor (EGFR) mutation status. A total of 176 patients (91, EGFR mutation; 85, wild-type EGFR) who underwent 18F-FDG PET/CT before treatment were enrolled. The main 18F-FDG PET/CT-derived variables: primary tumor maximum standardized uptake value (SUVmaxT), primary tumor total lesion glycolysis (TLGT), the maximum SUVmax of all selected lesions in whole body determined using the Response Evaluation Criteria In Solid Tumors (RECIST) 1.1 criteria (SUVmaxWBR), and whole-body total TLG determined using the RECIST 1.1 criteria (TLGWBR) were measured. Survival analysis regarding TLGWBR, and other factors in advanced lung adenocarcinoma patients stratified using EGFR mutation status, were evaluated. The results indicated that high TLGWBR (≥259.85), EGFR wild-type, and high serum LDH were independent predictors of worse PFS and OS in all patients with advanced lung adenocarcinoma. Among patients with wild-type EGFR, only TLGWBR retained significance as an independent predictor of both PFS and OS. Among patients with the EGFR mutation, high serum LDH level was an independent predictor of worse PFS and OS, and high TLGWBR (≥259.85) was an independent predictor of worse PFS but not worse OS. In conclusion, TLGWBR is a promising parameter for prognostic stratification of patients with advanced lung adenocarcinoma and EGFR status; however, it cannot be used to further stratify the risk of worse OS for patients with the EGFR mutation. Further prospective studies are needed to validate our findings.

Dynamics of angiogenesis and cellularity in rabbit VX2 tumors using contrast-enhanced magnetic resonance imaging and diffusion-weighted imaging

A number of studies have demonstrated that dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) may be used to evaluate microvessel density (MVD), and may quantitatively reflect tumor angiogenesis. To investigate the dynamics, including angiogenesis and tumor cellularity, of rabbit VX2 tumors during the 4 weeks following tumor implantation, the present study used DCE-MRI combined with diffusion-weighted imaging (DWI) to scan the tumors at 3 days, and then at 1, 2, 3 and 4-week intervals, following tumor implantation. The dynamics, volume transfer coefficient (Ktrans) and apparent diffusion coefficient (ADC) of the tumor parenchyma were analyzed. Furthermore, the associations between Ktrans and MVD at 4 weeks after tumor implantation were analyzed. Tumor Ktrans was positively correlated with MVD at 4 weeks (r=0.674, P<0.001). Following tumor implantation, the tumor Ktrans level rose for 2 weeks and then began to decline, reaching its lowest point at 4 weeks (P<0.001). ADC values at 1 week were higher than at 3 days, but declined thereafter (P<0.001). Tumor necrosis appeared by 1 week after tumor implantation. The necrosis degree of tumor was gradually increased from the occurrence of necrosis within the 4-week time span of the present study (1 vs. 2 weeks, P=0.008; 2 vs. 3 weeks, P<0.001; 3 vs. 4 weeks, P<0.001). The present study identified that tumor angiogenesis is a dynamic process that serves a function in tumor growth, and that DCE-MRI may reflect tumor parenchymal MVD and be useful in evaluating angiogenesis.

Predictive Role of the Number of 18F-FDG-Positive Lymph Nodes Detected by PET/CT for Pre-Treatment Evaluation of Locally Advanced Gastric Cancer.

Objectives The aim of this study was to investigate the predictive value of the numbers of metabolically positive lymph nodes (MPLN) detected by 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET)/computed tomography (CT) in patients with locally advanced gastric cancer (LAGC). Methods We retrospectively analyzed the records of 50 patients with LAGC (stage T2-T4) who had undergone pre-operative PET/CT examination and laparotomy (total gastrectomy, n = 11; subtotal gastrectomy, n = 13; distal gastrectomy, n = 22; and bypass with gastrojejunstomy, n = 4). The numbers of MPLN were determined by combining visual observations with semi-quantitative measurements of the maximized standardized uptake value (SUVmax). Performance was investigated in terms of predicting post-surgical overall survival (OS). Results The median post-surgical OS was 32.57 months (range 3.0-94 months). The numbers of MPLN were moderately correlated with the numbers of histological positive LN (r = 0.694, p = 0.001). In univariate analyses, the numbers of MPLN (≤ 2 vs. ≥3), PET/CT LN (positivity vs. negativity), SUVmax of LN (< 2.8 vs. ≥ 2.8), TNM stage (I, II vs. III, IV), and surgery type (R0 vs. non-R0) were significantly associated with OS. In multivariate analysis, surgery type (R0 vs. non-R0) and numbers of MPLN (≤ 2 vs. ≥ 3) were both independent factors for poor OS. Conclusions This explored study indicates that the number of MPLN could provide additional information for LAGC prognosis. Patients with MPLNs ≥ 3 may be at the risk of the more bad outcomes, and the further clinical trials are needed.