Automation of a Rule-based Workflow to Estimate Age from Brain MR Imaging of Infants and Children Up to 2 Years Old Using Stacked Deep Learning.

PURPOSE: Myelination-related MR signal changes in white matter are helpful for assessing normal development in infants and children. A rule-based myelination evaluation workflow regarding signal changes on T1-weighted images (T1WIs) and T2-weighted images (T2WIs) has been widely used in radiology. This study aimed to simulate a rule-based workflow using a stacked deep learning model and evaluate age estimation accuracy. METHODS: The age estimation system involved two stacked neural networks: a target network-to extract five myelination-related images from the whole brain, and an age estimation network from extracted T1- and T2WIs separately. A dataset was constructed from 119 children aged below 2 years with two MRI systems. A four-fold cross-validation method was adopted. The correlation coefficient (CC), mean absolute error (MAE), and root mean squared error (RMSE) of the corrected chronological age of full-term birth, as well as the mean difference and the upper and lower limits of 95% agreement, were measured. Generalization performance was assessed using datasets acquired from different MR images. Age estimation was performed in Sturge-Weber syndrome (SWS) cases. RESULTS: There was a strong correlation between estimated age and corrected chronological age (MAE: 0.98 months; RMSE: 1.27 months; and CC: 0.99). The mean difference and standard deviation (SD) were -0.15 and 1.26, respectively, and the upper and lower limits of 95% agreement were 2.33 and -2.63 months. Regarding generalization performance, the performance values on the external dataset were MAE of 1.85 months, RMSE of 2.59 months, and CC of 0.93. Among 13 SWS cases, 7 exceeded the limits of 95% agreement, and a proportional bias of age estimation based on myelination acceleration was exhibited below 12 months of age (P = 0.03). CONCLUSION: Stacked deep learning models automated the rule-based workflow in radiology and achieved highly accurate age estimation in infants and children up to 2 years of age.

Detection of residual metastatic tumor in the brain following Gamma Knife radiosurgery using a single or a series of magnetic resonance imaging scans: An autopsy study.

The aim of the present study was to investigate the usefulness of magnetic resonance image (MRI) for the detection of residual tumors following Gamma Knife radiosurgery (GKR) for brain metastases based on autopsy cases. The study investigated two hypotheses: i) Whether a single MRI may detect the existence of a tumor; and ii) whether a series of MRIs may detect the existence of a tumor. The study is a retrospective case series in a single institution. A total of 11 brain metastases in 6 patients were treated with GKR between 2002 and 2011. Histopathological specimens from autopsy were compared with reconstructed follow-up MRIs. The maximum diameters of the lesions on MRI series were measured, and the size changes classified. The primary sites in the patients were the kidneys (n=2), lung (n=1), breast (n=1) and colon (n=1), as well as 1 adenocarcinoma of unknown origin. The median prescribed dose for radiosurgery was 20 Gy (range, 18-20 Gy), and median time interval between GKR and autopsy was 10 months (range, 1.6-20 months). The pathological outcomes included 7 remissions and 4 failures. Enhanced areas on gadolinium-enhanced MRI contained various components: Viable tumor cells, tumor necrosis, hemorrhage, inflammation and vessels. Regarding the first hypothesis, it was impossible to distinguish pathological failure from remission with a single MRI scan due to the presence of various components. Conversely, in treatment response (remission or failure), on time-volume curves of MRI scans were in agreement with pathological findings, with the exception of progressive disease in the acute phase (0-3 months). Thus, regarding the second hypothesis, time-volume curves were useful for predicting treatment responses. In conclusion, it was difficult to predict treatment response using a single MRI, and a series of MRI scans were required to detect the existence of a tumor.

A clinical model to identify patients with high-risk plaque by coronary computed tomography angiography.

OBJECTIVES: Current clinical models predict the pre-test probability of obstructive coronary artery disease, but these models do not predict the presence of high-risk plaques. Thus the objective of this study was to propose a model to predict high-risk plaque assessed by coronary computed tomography (CT) angiography. METHODS: This study was a retrospective cross-sectional study. A clinical model was derived from 2392 patients and verified by 733 patients who underwent coronary CT suspected of coronary artery disease. High-risk plaque was defined as a plaque with positive remodeling (remodeling index>1.1), low attenuation (<30Hounsfield units) and napkin-ring sign. The risk score was calculated from the following 6 variables with a maximum of 24 points: age, sex, hemoglobin A1c, systolic blood pressure, high-density lipoprotein and smoking status. RESULTS: The proportion of patients with high-risk plaque was 11% and 17% in the derivation and validation cohort, respectively. The area under the receiver operation characteristic curve was 0.71 (95% confidence interval (CI): 0.68 to 0.74) in the derivation cohort and 0.75 (95% CI: 0.70 to 0.79) in the validation cohort. The frequency of high-risk plaques was 4% in the low-risk group (≤8 points) while it was 53% in the high-risk group (≥17 points) of the derivation cohort. CONCLUSIONS: We propose a scoring system to detect high-risk plaque assessed by coronary CT. Patients in the high-risk group have a high prevalence of high-risk plaque and might benefit from lipid lowering therapy.

Patient setup error and day-to-day esophageal motion error analyzed by cone-beam computed tomography in radiation therapy.

UNLABELLED: Little has been reported on the errors of setup and daily organ motion that occur during radiation therapy (RT) for esophageal cancer. The purpose of this paper was to determine the margins of esophageal motion during RT. METHODS AND MATERIALS: The shift of the esophagus was analyzed in 20 consecutive patients treated with RT for esophageal cancer from November 2007. CT images for RT planning were used as the primary image series. Computed tomography (CT) images were acquired using an Elekta Synergy System, equipped with a kilovoltage-based cone-beam CT (CBCT) unit. The subsequent CBCT image series used for daily RT setup were compared with the primary image series to analyze esophageal motion. CBCT was performed before treatment sessions a total of 10 times in each patient twice a week. The outer esophageal wall was contoured on the CBCT images of all 200 sets. RESULTS: In the 200 sets of CBCT images, the mean (absolute) +/- standard deviation (SD) of setup errors were 2 +/- 2 mm (max, 8 mm) in the lateral direction, 4 +/- 3 mm (max, 11 mm) in the longitudinal direction, and 4 +/- 3 mm (max, 13 mm) in the vertical direction. Additionally, the mean +/- SD values of daily esophageal motion comparing the CBCT with RT planning CT were 5 +/- 3 mm (max, 15 mm) in the lateral direction and 5 +/- 3 mm (max, 15 mm) in the vertical direction. CONCLUSIONS: Our data support the use of target margins (between the clinical target volume and planning target volume) of 9 mm for day-to-day esophageal motion and 8 mm for patient setup in all directions, respectively.