Automated Lung Segmentation on Chest Computed Tomography Images with Extensive Lung Parenchymal Abnormalities Using a Deep Neural Network

Objective@#We aimed to develop a deep neural network for segmenting lung parenchyma with extensive pathological conditions on non-contrast chest computed tomography (CT) images. @*Materials and Methods@#Thin-section non-contrast chest CT images from 203 patients (115 males, 88 females; age range, 31–89 years) between January 2017 and May 2017 were included in the study, of which 150 cases had extensive lung parenchymal disease involving more than 40% of the parenchymal area. Parenchymal diseases included interstitial lung disease (ILD), emphysema, nontuberculous mycobacterial lung disease, tuberculous destroyed lung, pneumonia, lung cancer, and other diseases. Five experienced radiologists manually drew the margin of the lungs, slice by slice, on CT images. The dataset used to develop the network consisted of 157 cases for training, 20 cases for development, and 26 cases for internal validation. Two-dimensional (2D) U-Net and three-dimensional (3D) U-Net models were used for the task. The network was trained to segment the lung parenchyma as a whole and segment the right and left lung separately. The University Hospitals of Geneva ILD dataset, which contained high-resolution CT images of ILD, was used for external validation. @*Results@#The Dice similarity coefficients for internal validation were 99.6 ± 0.3% (2D U-Net whole lung model), 99.5 ± 0.3% (2D U-Net separate lung model), 99.4 ± 0.5% (3D U-Net whole lung model), and 99.4 ± 0.5% (3D U-Net separate lung model).The Dice similarity coefficients for the external validation dataset were 98.4 ± 1.0% (2D U-Net whole lung model) and 98.4 ± 1.0% (2D U-Net separate lung model). In 31 cases, where the extent of ILD was larger than 75% of the lung parenchymal area, the Dice similarity coefficients were 97.9 ± 1.3% (2D U-Net whole lung model) and 98.0 ± 1.2% (2D U-Net separate lung model). @*Conclusion@#The deep neural network achieved excellent performance in automatically delineating the boundaries of lung parenchyma with extensive pathological conditions on non-contrast chest CT images.

No-Touch Radiofrequency Ablation of VX2 Hepatic Tumors In Vivo in Rabbits: A Proof of Concept Study

OBJECTIVE: In a proof of concept study, we compared no-touch radiofrequency ablation (NtRFA) in bipolar mode with conventional direct tumor puncture (DTP) in terms of local tumor control (LTC), peritoneal seeding, and tumorigenic factors, in the rabbit VX2 subcapsular hepatic tumor model. MATERIALS AND METHODS: Sixty-two rabbits with VX2 subcapsular hepatic tumors were divided into three groups according to the procedure: DTP-RFA (n = 25); NtRFA (n = 25); and control (n = 12). Each of the three groups was subdivided into two sets for pathologic analysis (n = 24) or computed tomography (CT) follow-up for 6 weeks after RFA (n = 38). Ultrasonography-guided DTP-RFA and NtRFA were performed nine days after tumor implantation. LTC was defined by either achievement of complete tumor necrosis on histopathology or absence of local tumor progression on follow-up CT and autopsy. Development of peritoneal seeding was also compared among the groups. Serum hepatocyte growth factor (HGF), vascular endothelial growth factor (VEGF) and interleukin-6 (IL-6) were measured via ELISA (Elabscience Biotechnology Co.) after RFA for tumorigenic factor evaluation. RESULTS: Regarding LTC, there was a trend in NtRFA (80%, 20/25) toward better ablation than in DTP-RFA (56%, 14/25) (p = 0.069). Complete tumor necrosis was achieved in 54.5% of DTP-RFA (6/11) and 90.9% of NtRFA (10/11). Peritoneal seeding was significantly more common in DTP-RFA (71.4%, 10/14) than in NtRFA (21.4%, 3/14) (p = 0.021) or control (0%). Elevations of HGF, VEGF or IL-6 were not detected in any group. CONCLUSION: No-touch radiofrequency ablation led to lower rates of peritoneal seeding and showed a tendency toward better LTC than DTP-RFA.

Ultrasonographic features differentiating thyroglossal duct cysts from dermoid cysts

PURPOSE: The purpose of this study was to identify ultrasonographic features that can be used to differentiate between thyroglossal duct cysts (TGDCs) and dermoid cysts (DCs). METHODS: We searched surgical pathology reports completed between January 2004 and October 2015 and identified 66 patients with TGDCs or DCs who had undergone preoperative ultrasonography. The ultrasound images were reviewed by two radiologists who were blinded to the pathological diagnosis. They evaluated the following parameters: dimensions, shape, margin, location in relation to the midline, level in relation to the hyoid bone, attachment to the hyoid bone, the depth of the lesion in relation to the strap muscles, internal echogenicity, internal echogenic dots, multilocularity, the presence of a longitudinal extension into the tongue base, posterior acoustic enhancement, the presence of internal septae, and intralesional vascularity. RESULTS: There were 50 TGDCs and 16 DCs. TGDCs were significantly more likely than DCs to have an irregular shape, an ill-defined margin, attachment to the hyoid bone, an intramuscular location, heterogeneous internal echogenicity, multilocularity, and longitudinal extension into the tongue base. CONCLUSION: Ultrasound findings may inform the differential diagnosis between TGDCs and DCs.