Few-Shot Learning for Medical Image Segmentation Using 3D U-Net and Model-Agnostic Meta-Learning (MAML).

Deep learning has attained state-of-the-art results in general image segmentation problems; however, it requires a substantial number of annotated images to achieve the desired outcomes. In the medical field, the availability of annotated images is often limited. To address this challenge, few-shot learning techniques have been successfully adapted to rapidly generalize to new tasks with only a few samples, leveraging prior knowledge. In this paper, we employ a gradient-based method known as Model-Agnostic Meta-Learning (MAML) for medical image segmentation. MAML is a meta-learning algorithm that quickly adapts to new tasks by updating a model's parameters based on a limited set of training samples. Additionally, we use an enhanced 3D U-Net as the foundational network for our models. The enhanced 3D U-Net is a convolutional neural network specifically designed for medical image segmentation. We evaluate our approach on the TotalSegmentator dataset, considering a few annotated images for four tasks: liver, spleen, right kidney, and left kidney. The results demonstrate that our approach facilitates rapid adaptation to new tasks using only a few annotated images. In 10-shot settings, our approach achieved mean dice coefficients of 93.70%, 85.98%, 81.20%, and 89.58% for liver, spleen, right kidney, and left kidney segmentation, respectively. In five-shot sittings, the approach attained mean Dice coefficients of 90.27%, 83.89%, 77.53%, and 87.01% for liver, spleen, right kidney, and left kidney segmentation, respectively. Finally, we assess the effectiveness of our proposed approach on a dataset collected from a local hospital. Employing five-shot sittings, we achieve mean Dice coefficients of 90.62%, 79.86%, 79.87%, and 78.21% for liver, spleen, right kidney, and left kidney segmentation, respectively.

Psoas muscle area and paraspinal muscle fat in children and young adults with or without obesity and fatty liver.

BACKGROUND: Little is known about the muscle condition in children with obesity. OBJECTIVES: To investigate the effect of obesity and fatty liver on muscle area and muscle fat in children and young adults. MATERIALS AND METHODS: We evaluated consecutive liver fat quantification MRIs in children and young adults between June 2015 and April 2019. We obtained hepatic fat and paraspinal muscle fat at mid L2 from the fat map, psoas muscle area (PMA) at mid L3, and z-score of PMA. The patient's age, height and weight at the time of the MRI were recorded. Body mass index (BMI) z-score was also calculated. Spearman correlation and partial correlation analyses were performed. Univariate and multivariate regression analyses were also performed using significant variables. RESULTS: A total of 132 patients (97 male) were included with a median age of 13.0 years (interquartile range 11-16 years). The median BMI was 23.7 kg/m2 (interquartile range 21.2-27.7 kg/m2). The weight, BMI, liver fat, and z-score of PMA were all higher in male patients than they were in female patients. The amount of liver fat had no correlation with muscle fat or PMA z-score after adjusting BMI. However, the BMI z-score was positively correlated with the PMA z-score (ρ = 0.432, p<0.001) even after adjusting for liver fat. On regression analyses, the BMI z-score had linear positive relationship with PMA z-score (ß = 0.289, p<0.001) and muscle fat (ß = 0.218, p = 0.016). CONCLUSIONS: Male children and young adults have greater PMA than do female children and young adults. Obesity is associated with higher PMA and paraspinal muscle fat. However, liver fat is not related with the muscle condition in children and young adults.

Quantitative MRI Assessment of Pancreatic Steatosis Using Proton Density Fat Fraction in Pediatric Obesity.

OBJECTIVE: To assess the feasibility of quantitatively assessing pancreatic steatosis using magnetic resonance imaging (MRI) and its correlation with obesity and metabolic risk factors in pediatric patients. MATERIALS AND METHODS: Pediatric patients (≤ 18 years) who underwent liver fat quantification MRI between January 2016 and June 2019 were retrospectively included and divided into the obesity and control groups. Pancreatic proton density fat fraction (P-PDFF) was measured as the average value for three circular regions of interest (ROIs) drawn in the pancreatic head, body, and tail. Age, weight, laboratory results, and mean liver MRI values including liver PDFF (L-PDFF), stiffness on MR elastography, and T2* values were assessed for their correlation with P-PDFF using linear regression analysis. The associations between P-PDFF and metabolic risk factors, including obesity, hypertension, diabetes mellitus (DM), and dyslipidemia, were assessed using logistic regression analysis. RESULTS: A total of 172 patients (male:female = 125:47; mean ± standard deviation [SD], 13.2 ± 3.1 years) were included. The mean P-PDFF was significantly higher in the obesity group than in the control group (mean ± SD, 4.2 ± 2.5% vs. 3.4 ± 2.4%; p = 0.037). L-PDFF and liver stiffness values showed no significant correlation with P-PDFF (p = 0.235 and p = 0.567, respectively). P-PDFF was significantly associated with obesity (odds ratio 1.146, 95% confidence interval 1.006-1.307, p = 0.041), but there was no significant association with hypertension, DM, and dyslipidemia. CONCLUSION: MRI can be used to quantitatively measure pancreatic steatosis in children. P-PDFF is significantly associated with obesity in pediatric patients.

Determining the optimal timing of screening spinal cord ultrasonography to detect filum terminale lipoma in infants.

PURPOSE: The purpose of this study was to identify the optimal timing for screening spinal cord ultrasonography (US) to detect filum terminale lipoma in infants. METHODS: We retrospectively reviewed infants (<12 months old) who underwent repeated spinal cord US between April 2011 and January 2019. We excluded infants if they only had one US examination, or if they had lesions other than filum terminale lipoma. Infants with filum terminale lipoma on magnetic resonance imaging were included in the lipoma group and the others in the control group. A linear mixed model was used to assess differences in the growth pattern of filum terminale thickness by age and group. The cutoff thickness on US and its diagnostic performance were assessed according to age. RESULTS: Among 442 infants with 901 US examinations, 46 were included in the lipoma group and 58 in the control group. Sixty-seven infants had unmeasurable filum terminale thickness on initial US, including 55 neonates (82.1%) before 1 month of age. The lipoma group had significantly greater filum terminale thickness than the control group (P<0.001). Thickness increased with age in the lipoma group (P=0.027). The sensitivity of US was 87.5% and the area under the receiver operating characteristic curve was 0.949 (95% confidence interval, 0.849 to 0.991) with a cutoff value of 1.1 mm in 4- to 6-month-old infants. CONCLUSION: Screening spinal cord US could effectively diagnose filum terminale lipoma in 4- to 6-month-old infants with a cutoff thickness of 1.1 mm. Spinal cord US can be used to screen young infants with intraspinal abnormalities.

Bone marrow fat change in pediatric patients with non-alcoholic fatty liver disease.

OBJECTIVES: To investigate changes of fat in bone marrow (BM) and paraspinal muscle (PSM) associated with the degree of fatty liver in pediatric patients with non-alcoholic fatty liver disease (NAFLD) in consideration of age and body mass index (BMI). METHODS: Hepatic fat, BM fat, and PSM fat from proton density fat fraction of liver MRI between June 2015 and April 2019 were quantitatively evaluated on axial images of the fat map at the mid-level of T11-L2 vertebral bodies for BM fat and at the mid-level of L2 for PSM fat. Age, height, and weight at the time of MRI were recorded and BMI was calculated. Correlation analysis was performed. RESULTS: A total of 147 patients (114 male) were included with a mean age of 13.3 ± 2.9 years (range 7-18 years). The mean fat fractions were 24.3 ± 13.0% (2-53%) in liver, 37.4 ± 8.6% (17.3-56%) in vertebral BM, and 2.7 ± 1.1% (1.0-6.9%) in PSM. Age, height, weight, and BMI were not correlated with liver fat or BM fat. However, weight (ρ = 0.174, p = 0.035) and BMI (ρ = 0.247, p = 0.003) were positively correlated with PSM fat. Liver fat showed positive correlation with BM fat when adjusting age and BMI (ρ = 0.309, p<0.001), but not with PSM fat. CONCLUSIONS: BM fat positively correlates with liver fat, but not with age or BMI in pediatric NAFLD patients.

Accelerated metallic artifact reduction imaging using spectral bin modulation of multiacquisition variable-resonance image combination selective imaging.

OBJECTIVES: To assess the clinical utility of a prototype sequence for metal artifact reduction, the multiacquisition variable-resonance image combination selective (MAVRIC-SL) at 3 T. This sequence allows a surgical prosthesis-dependent reduction in the number of spectral bins. We compared the prototype MAVRIC SL to the conventional two-dimensional fast spin-echo (FSE) sequences and MAVRIC SL images acquired with all spectral bins to those acquired with the optimized number of spectral bins. METHODS: MAVRIC SL images were acquired in 25 image sets from August 2017 to April 2018. For each subject, the optimized number of spectral bins was determined using a short spectral calibration scan. The image sets obtained with magnetic resonance imaging that were used for the analysis consisted of MAVRIC-SL proton density (PD)-weighted or short inversion time inversion recovery (STIR) images acquired with all 24 spectral bins, the corresponding images with the optimized number of spectral bins, and the conventional two-dimensional FSE or STIR PD-weighted images. A musculoskeletal radiologist reviewed and scored the images using a five-point scale for artifact reduction around the prosthesis and visualization of the prosthesis and peri-prosthetic tissues. Quantitative evaluation of the peri-prosthetic tissues was also performed. The Wilcoxon rank-sum test was used to test for significance. RESULTS: The MAVRIC SL images enabled a significantly improved reduction in metallic artifacts compared to the conventional two-dimensional FSE sequences. The optimized number of spectral bins ranged from 6 to 20, depending on the prosthesis susceptibility difference, size, and orientation to the B0 field. The scan times significantly decreased with a reduced number of spectral bins (354.0 ± 139.1 versus 283.0 ± 89.6 s; 20% reduced scan time; p < .05). Compared to the MAVRIC SL images acquired with all 24 bins, the artifact reduction and visualization of the prosthesis and peri-prosthetic tissues on the MAVRIC SL images acquired with calibrated bins were not significantly different. CONCLUSIONS: Compared to the MAVRIC SL images acquired with all 24 spectral bins, those acquired with an optimized number of spectral bins can reduce metallic artifacts with no significant image quality degradation while providing reduced scan time.

Fast isotropic volumetric magnetic resonance imaging of the ankle: Acceleration of the three-dimensional fast spin echo sequence using compressed sensing combined with parallel imaging.

OBJECTIVES: To investigate the feasibility of three-dimensional fast spin echo (3D-FSE) imaging with compressed sensing (CS) and parallel imaging (PI) compared to 3D-FSE imaging with only PI in evaluating ankle joint pathologies. MATERIALS AND METHODS: Twenty consecutive patients underwent ankle magnetic resonance imaging (MRI), including acquisition of image sets of 2D-FSE sequences, and 3D-FSE sequences without and with CS, between June 2016 and November 2017. Three MR image sets were independently rated by two radiologists for the presence/absence of ankle pathology. Quantitative image similarity and subjective image quality were evaluated using 3D-FSE images without CS and those with CS-PI. Inter-sequence agreement between 3D-FSE sequences without CS and with CS-PI in both readers was evaluated. RESULTS: Interobserver agreements were nearly perfect for sprain of the anterior talofibular ligament (ATFL, κ=0.77), osteochondral lesion of the talus (OLT, κ=0.76-0.88), osteochondral lesion of the distal tibia (OLTi, κ=0.74) and os subfibulare (OSF, κ=0.62-0.64). The structural similarity index (mean, 0.996; range, 0.990-0.997) between the 3D-FSE sequences without CS and with CS-PI was acceptable. There was no significant difference in subjective image quality between the two imaging sequences (ATFL, p = 0.317; bone marrow, p = 0.083; cartilage, p = 1.000, tendon, p = 1.000). Intersequence agreement between the 3D-FSE sequences with and without CS was nearly perfect (ATFL and OLTi, κ=1.00; OLT, κ=0.87-0.96; OSF, κ=0.62-0.64) in both readers. CONCLUSIONS: Isotropic 3D-FSE ankle MRI with CS provides acceptable diagnostic performance with reduced scan time. Compressed sensing-related artifacts could be minimized with CS reconstruction enhancement, allowing for better image quality for evaluating ankle joint pathologies.