Improvement of semantic segmentation through transfer learning of multi-class regions with convolutional neural networks on supine and prone breast MRI images.

Semantic segmentation of breast and surrounding tissues in supine and prone breast magnetic resonance imaging (MRI) is required for various kinds of computer-assisted diagnoses for surgical applications. Variability of breast shape in supine and prone poses along with various MRI artifacts makes it difficult to determine robust breast and surrounding tissue segmentation. Therefore, we evaluated semantic segmentation with transfer learning of convolutional neural networks to create robust breast segmentation in supine breast MRI without considering supine or prone positions. Total 29 patients with T1-weighted contrast-enhanced images were collected at Asan Medical Center and two types of breast MRI were performed in the prone position and the supine position. The four classes, including lungs and heart, muscles and bones, parenchyma with cancer, and skin and fat, were manually drawn by an expert. Semantic segmentation on breast MRI scans with supine, prone, transferred from prone to supine, and pooled supine and prone MRI were trained and compared using 2D U-Net, 3D U-Net, 2D nnU-Net and 3D nnU-Net. The best performance was 2D models with transfer learning. Our results showed excellent performance and could be used for clinical purposes such as breast registration and computer-aided diagnosis.

Breast tumor movements analysis using MRI scans in prone and supine positions.

We quantitatively evaluated breast tumor movement and volume changes between magnetic resonance imaging (MRI) scans in prone and supine positions. Twenty-seven breast tumor patients who received neoadjuvant systemic therapy (NST) for breast-conserving surgery were studied. Before and after NST, MRI scans in prone and supine positions were performed immediately. Tumor segmentation, volume, and position of tumors were evaluated in both positions. Average tumor volumes in prone and supine positions did not significantly differ (p = 0.877). Tumor movement from prone to supine positions from the origin of the bottom center of the sternum was strongly correlated with the distance from the tumor center to the chest wall (r = 0.669; p < 0.05). Tumor changes from prone to supine positions measured from the origin of the nipple depended on the location of the tumor in the breast. The prone-to-supine movement of all tumors from the origin of the bottom center of the sternum tended to move outward from the sagittal centerline of the body on the coronal plane, to the inside of the body on the sagittal plane, and outward and downward close to the body on the axial plane, which might help in planning operations using prone MRI in supine-position breast cancer surgery.

Effect of laterality, gender, age and body mass index on the fat fraction of salivary glands in healthy volunteers: assessed using iterative decomposition of water and fat with echo asymmetry and least-squares estimation method.

OBJECTIVES:: To evaluate the effect of laterality, gender, age and body mass index (BMI) on fat fraction (FF) measurements of both parotid glands (PGs) and submandibular glands (SMGs) by using: Iterative decomposition of water and fat with echo asymmetry and least-squares estimation method (IDEAL-IQ). METHODS:: A total of 87 healthy participants were enrolled in our study. IDEAL-IQ image was scanned using a 3.0 T scanner. Paired t test was performed to compare the difference on FF of both PGs and SMGs between left and right side. The FF of two glands between male and female healthy participants were compared using an unpaired t-test. The correlation between the FF of two glands and participant age or BMI were analyzed by Pearson's correlation. RESULTS:: Excellent inter- and intrareader agreements were obtained during the measurements of FF by IDEAL-IQ method (ICC, 0.952-0.981). FF values correlated positively with the age and BMI in both left and right PGs and SMGs (p < 0.05). No significant difference was found on FF between left and right PGs and SMGs (p > 0.05). There was also no difference on FF between male and female healthy participants (p > 0.05). CONCLUSIONS:: FFs of PGs and SMGs were age- and BMI- dependent, but not laterality- and gender-dependent. The effect of age and BMI need to be considered in further studies using Ideal-IQ technology to evaluate FFs of salivary gland diseases.

Evaluation of MR issues for the latest standard brands of orthopedic metal implants: plates and screws.

PURPOSE: The study was performed to evaluate magnetic resonance (MR) issues for the latest standard brands of plates and screws used in orthopedic surgery at a 1.5-T MR system, including the safety and metallic artifacts. METHODS: The plates and screws (made of titanium alloy and stainless steel materials, according to the latest standard brands) were assessed for displacement in degrees, MRI-related heating and artifacts at a 1.5-T MR system. The displacement in degrees of the plates and screws was evaluated on an angel-measurement instrument at the entrance of the MR scanner. The MRI-related heating was assessed on a swine leg fixed with a plate by using a "worst-case" pulse sequence. A rectangular water phantom was designed to evaluate metallic artifacts of a screw on different sequences (T1/T2-weighted FSE, STIR, T2-FSE fat saturation, GRE, DWI) and then artifacts were evaluated on T2-weighted FSE sequence by modifying the scanning parameters including field of view (FOV), echo train length (ETL) and bandwidth to identify the influence of parameters on metallic artifacts. 15 volunteers with internal vertebral fixation (titanium alloy materials) were scanned with MR using axial and sagittal T2-FSE, sagittal T2-FSE fat suppression and STIR with conventional and optimized parameters, respectively. Then all images were graded by two experienced radiologists having the experience of more than 7 years under double-blind studies that is neither of them knew which was conventional parameter group and optimized parameter group. RESULTS: The average deflection angle of titanium alloy and stainless steel implants were 4.3° and 7.7°, respectively, (less than 45°) which indicated that the magnetically induced force was less than the weight of the object. The deflection angle of the titanium alloy implants was less than the stainless steel one (t=9.69, P<0.001). The average temperature changes of titanium alloy before and after the scan was 0.48°C and stainless steel implants was 0.74°C, respectively, with the background temperature changes of 0.24°C. The water phantom test indicated that the DWI sequence produced largest artifacts, while FSE pulse sequence produced smallest artifacts. And T2-weighted FSE fat saturation sequence produced larger artifacts than STIR sequence. The influence of the scanning parameters on metallic artifacts was verified that metallic artifacts increased with longer echo train length and bigger FOV, while decreased with larger bandwidth. The interreader agreement was good or excellent for each set of images graded with Cohen's Kappa statistic. Image grading of axial and sagittal T2-FSE with optimized parameters were significantly superior to that with conventional parameters (grade, 3.3±0.5 vs 2.7±0.6, P=0.003; 3.2±0.4 vs 1.9±0.7, P=0.001) and image of STIR sequence received a better grade than T2-FSE FS sequence (grade, 3.4±0.5 vs 1.7±0.6, P<0.001). CONCLUSIONS: The latest standard plates and screws used in orthopedic surgery do not pose an additional hazard or risk to patients undergoing MR imaging at 1.5-T or less. Though artifacts caused by them cannot be ignored because of their relatively large size, it is possible to be minimized by choosing appropriate pulse sequences and optimizing scanning parameters, such as FSE and STIR sequence with large bandwidth, small FOV and appropriate echo train length.

[Investigation of short-term therapy results for radiofrequency ablation by positron emission tomography].

OBJECTIVE: To explore the possibility of evaluating the short-term therapeulic effect of radiofrequency ablation (RFA) in cancer treatment by positron emission tomography. METHODS: The radioactivity intensity of the tumor was detected by PET before and after RFA. RESULTS: Radioactivity blank was observed in all 33 cases with 54 lesions, indicating the elimination of the tumor was destroyed. CONCLUSION: PET is of great significance in evaluating the short-term effect of RFA.