吉川 武

Background Deep learning may help to improve computer-aided detection of volume (CADv) measurement of pulmonary nodules at chest CT. Purpose To determine the efficacy of a deep learning method for improving CADv for measuring the solid and ground-glass opacity (GGO) volumes of a nodule, doubling time (DT), and the change in volume at chest CT. Materials and Methods From January 2014 to December 2016, patients with pulmonary nodules at CT were retrospectively reviewed. CADv without and with a convolutional neural network (CNN) automatically determined total nodule volume change per day and DT. Area under the curves (AUCs) on a per-nodule basis and diagnostic accuracy on a per-patient basis were compared among all indexes from CADv with and without CNN for differentiating benign from malignant nodules. Results The CNN training set was 294 nodules in 217 patients, the validation set was 41 nodules in 32 validation patients, and the test set was 290 nodules in 188 patients. A total of 170 patients had 290 nodules (mean size ± standard deviation, 11 mm ± 5; range, 4-29 mm) diagnosed as 132 malignant nodules and 158 benign nodules. There were 132 solid nodules (46%), 106 part-solid nodules (36%), and 52 ground-glass nodules (18%). The test set results showed that the diagnostic performance of the CNN with CADv for total nodule volume change per day was larger than DT of CADv with CNN (AUC, 0.94 [95% confidence interval {CI}: 0.90, 0.96] vs 0.67 [95% CI: 0.60, 0.74]; P < .001) and CADv without CNN (total nodule volume change per day: AUC, 0.69 [95% CI: 0.62, 0.75]; P < .001; DT: AUC, 0.58 [95% CI: 0.51, 0.65]; P < .001). The accuracy of total nodule volume change per day of CADv with CNN was significantly higher than that of CADv without CNN (P < .001) and DT of both methods (P < .001). Conclusion Convolutional neural network is useful for improving accuracy of computer-aided detection of volume measurement and nodule differentiation capability at CT for patients with pulmonary nodules. © RSNA, 2020 Online supplemental material is available for this article.

PURPOSE: To prospectively compare the capability of dynamic first-pass contrast-enhanced (CE) perfusion MR imaging with ultra-short TE and area-detector CT (ADCT), analyzed with the same mathematical methods, and that of FDG-PET/CT for diagnosis and management of solitary pulmonary nodules (SPNs). METHODS AND MATERIALS: Our institutional review board approved this study and written informed consent was obtained from all subjects. A total 57 consecutive patients with 71 nodules prospectively underwent dynamic CE-perfusion ADCT and MR imaging with ultra-short TE, FDG-PET/CT, as well as microbacterial and/or pathological examinations. The nodules were classified into malignant nodules (n = 45) and benign nodules (n = 26). Pulmonary arterial, systemic arterial and total perfusions were determined by means of dual-input maximum slope models on ADCT and MR imaging and maximum values of standard uptake values (SUVmax) on PET/CT. Receiver operating characteristic (ROC) analysis was performed for each index, and sensitivity, specificity and accuracy were compared by McNemar's test. RESULTS: Areas under the curve (Azs) of total perfusion on ADCT (Az = 0.89) and MR imaging (Az = 0.88) were significantly larger than those of systemic arterial perfusion and MR imaging (p<0.05). Accuracy of total perfusion on ADCT (87.3% [62/71]) and MR imaging (87.3% [62/71]) was significantly higher than that of systemic arterial perfusion for both methods (77.5% [55/71] p = 0.02) and SUVmax (78.9% [56/71], p = 0.03). CONCLUSION: Dynamic CE-perfusion MR imaging with ultra-short TE and ADCT and have similar potential capabilities, and are superior to FDG-PET/CT in this setting.

OBJECTIVE: Ultrashort TE (UTE) MRI has been shown to deliver high-resolution images comparable to CT images. Here we evaluate the potential of UTE-MRI for precise lung nodule characterization. SUBJECTS AND METHODS: Fifty-one patients (mean [± SD] age, 68.7 ± 10.8 years) with 119 nodules or masses (mean size, 17.4 ± 16.3 mm; range, 4-88 mm) prospectively underwent CT (1-mm slice thickness) and UTE-MRI (TE, 192 μs; 1 mm3 resolution). Two radiologists assessed nodule dimensions and morphologic features (i.e., attenuation, margins, and internal lucencies), in consensus for CT and in a blinded fashion for UTE-MRI. Sensitivity, specificity, and kappa statistics were calculated in reference to CT. RESULTS: Readers 1 and 2 underestimated the nodules' long axial diameter with UTEMRI by 1.2 ± 3.4 and 2.1 ± 4.2 mm, respectively (p < 0.001). The sensitivity and specificity of UTE-MRI for subsolid attenuation were 95.9% and 70.3%, respectively, for reader 1 and 97.1% and 71.4%, respectively, for reader 2 (κ = 0.71 and 0.68). With regard to margin characteristics, for lobulation, sensitivity was 70.6% and 54.9%, and specificity was 93.2% and 96.3% for readers 1 and 2, respectively; for spiculation, sensitivity was 61.5% and 48.0%, and specificity was 95.2% and 95.0%; and for pleural tags, sensitivity was 87.0% and 73.3%, and specificity was 93.8% and 95.0%. Finally, for internal lucencies, sensitivity was 72.7% and 61.3%, and specificity was 96.1% and 97.3% for readers 1 and 2, respectively (κ = 0.64-0.81 for reader 1 and 0.48-0.72 for reader 2). Interreader agreement for attenuation, margin characteristics, and lucencies was substantial to almost perfect with few exceptions (κ = 0.51-0.90). CONCLUSION: UTE-MRI systematically underestimated dimension measurements by approximately 1-2 mm but otherwise showed high diagnostic properties and interreader agreement, yet unprecedented by MRI, for nodule morphologic assessment.