Improved Robustness for Deep Learning-based Segmentation of Multi-Center Myocardial Perfusion MRI Datasets Using Data Adaptive Uncertainty-guided Space-time Analysis.

BACKGROUND: Fully automatic analysis of myocardial perfusion MRI datasets enables rapid and objective reporting of stress/rest studies in patients with suspected ischemic heart disease. Developing deep learning techniques that can analyze multi-center datasets despite limited training data and variations in software (pulse sequence) and hardware (scanner vendor) is an ongoing challenge. METHODS: Datasets from 3 medical centers acquired at 3T (n = 150 subjects; 21,150 first-pass images) were included: an internal dataset (inD; n = 95) and two external datasets (exDs; n = 55) used for evaluating the robustness of the trained deep neural network (DNN) models against differences in pulse sequence (exD-1) and scanner vendor (exD-2). A subset of inD (n = 85) was used for training/validation of a pool of DNNs for segmentation, all using the same spatiotemporal U-Net architecture and hyperparameters but with different parameter initializations. We employed a space-time sliding-patch analysis approach that automatically yields a pixel-wise "uncertainty map" as a byproduct of the segmentation process. In our approach, dubbed Data Adaptive Uncertainty-Guided Space-time (DAUGS) analysis, a given test case is segmented by all members of the DNN pool and the resulting uncertainty maps are leveraged to automatically select the "best" one among the pool of solutions. For comparison, we also trained a DNN using the established approach with the same settings (hyperparameters, data augmentation, etc.). RESULTS: The proposed DAUGS analysis approach performed similarly to the established approach on the internal dataset (Dice score for the testing subset of inD: 0.896 ± 0.050 vs. 0.890 ± 0.049; p = n.s.) whereas it significantly outperformed on the external datasets (Dice for exD-1: 0.885 ± 0.040 vs. 0.849 ± 0.065, p < 0.005; Dice for exD-2: 0.811 ± 0.070 vs. 0.728 ± 0.149, p < 0.005). Moreover, the number of image series with "failed" segmentation (defined as having myocardial contours that include bloodpool or are noncontiguous in ≥1 segment) was significantly lower for the proposed vs. the established approach (4.3% vs. 17.1%, p < 0.0005). CONCLUSIONS: The proposed DAUGS analysis approach has the potential to improve the robustness of deep learning methods for segmentation of multi-center stress perfusion datasets with variations in the choice of pulse sequence, site location or scanner vendor.

Improved Robustness for Deep Learning-based Segmentation of Multi-Center Myocardial Perfusion MRI Datasets Using Data Adaptive Uncertainty-guided Space-time Analysis.

Background: Fully automatic analysis of myocardial perfusion MRI datasets enables rapid and objective reporting of stress/rest studies in patients with suspected ischemic heart disease. Developing deep learning techniques that can analyze multi-center datasets despite limited training data and variations in software (pulse sequence) and hardware (scanner vendor) is an ongoing challenge. Methods: Datasets from 3 medical centers acquired at 3T (n = 150 subjects; 21,150 first-pass images) were included: an internal dataset (inD; n = 95) and two external datasets (exDs; n = 55) used for evaluating the robustness of the trained deep neural network (DNN) models against differences in pulse sequence (exD-1) and scanner vendor (exD-2). A subset of inD (n = 85) was used for training/validation of a pool of DNNs for segmentation, all using the same spatiotemporal U-Net architecture and hyperparameters but with different parameter initializations. We employed a space-time sliding-patch analysis approach that automatically yields a pixel-wise "uncertainty map" as a byproduct of the segmentation process. In our approach, dubbed Data Adaptive Uncertainty-Guided Space-time (DAUGS) analysis, a given test case is segmented by all members of the DNN pool and the resulting uncertainty maps are leveraged to automatically select the "best" one among the pool of solutions. For comparison, we also trained a DNN using the established approach with the same settings (hyperparameters, data augmentation, etc.). Results: The proposed DAUGS analysis approach performed similarly to the established approach on the internal dataset (Dice score for the testing subset of inD: 0.896 ± 0.050 vs. 0.890 ± 0.049; p = n.s.) whereas it significantly outperformed on the external datasets (Dice for exD-1: 0.885 ± 0.040 vs. 0.849 ± 0.065, p < 0.005; Dice for exD-2: 0.811 ± 0.070 vs. 0.728 ± 0.149, p < 0.005). Moreover, the number of image series with "failed" segmentation (defined as having myocardial contours that include bloodpool or are noncontiguous in ≥1 segment) was significantly lower for the proposed vs. the established approach (4.3% vs. 17.1%, p < 0.0005). Conclusions: The proposed DAUGS analysis approach has the potential to improve the robustness of deep learning methods for segmentation of multi-center stress perfusion datasets with variations in the choice of pulse sequence, site location or scanner vendor.

Temporal Uncertainty Localization to Enable Human-in-the-loop Analysis of Dynamic Contrast-enhanced Cardiac MRI Datasets.

Dynamic contrast-enhanced (DCE) cardiac magnetic resonance imaging (CMRI) is a widely used modality for diagnosing myocardial blood flow (perfusion) abnormalities. During a typical free-breathing DCE-CMRI scan, close to 300 time-resolved images of myocardial perfusion are acquired at various contrast "wash in/out" phases. Manual segmentation of myocardial contours in each time-frame of a DCE image series can be tedious and time-consuming, particularly when non-rigid motion correction has failed or is unavailable. While deep neural networks (DNNs) have shown promise for analyzing DCE-CMRI datasets, a "dynamic quality control" (dQC) technique for reliably detecting failed segmentations is lacking. Here we propose a new space-time uncertainty metric as a dQC tool for DNN-based segmentation of free-breathing DCE-CMRI datasets by validating the proposed metric on an external dataset and establishing a human-in-the-loop framework to improve the segmentation results. In the proposed approach, we referred the top 10% most uncertain segmentations as detected by our dQC tool to the human expert for refinement. This approach resulted in a significant increase in the Dice score (p<0.001) and a notable decrease in the number of images with failed segmentation (16.2% to 11.3%) whereas the alternative approach of randomly selecting the same number of segmentations for human referral did not achieve any significant improvement. Our results suggest that the proposed dQC framework has the potential to accurately identify poor-quality segmentations and may enable efficient DNN-based analysis of DCE-CMRI in a human-in-the-loop pipeline for clinical interpretation and reporting of dynamic CMRI datasets.

Temporal Uncertainty Localization to Enable Human-in-the-Loop Analysis of Dynamic Contrast-Enhanced Cardiac MRI Datasets.

Dynamic contrast-enhanced (DCE) cardiac magnetic resonance imaging (CMRI) is a widely used modality for diagnosing myocardial blood flow (perfusion) abnormalities. During a typical free-breathing DCE-CMRI scan, close to 300 time-resolved images of myocardial perfusion are acquired at various contrast "wash in/out" phases. Manual segmentation of myocardial contours in each time-frame of a DCE image series can be tedious and time-consuming, particularly when non-rigid motion correction has failed or is unavailable. While deep neural networks (DNNs) have shown promise for analyzing DCE-CMRI datasets, a "dynamic quality control" (dQC) technique for reliably detecting failed segmentations is lacking. Here we propose a new space-time uncertainty metric as a dQC tool for DNN-based segmentation of free-breathing DCE-CMRI datasets by validating the proposed metric on an external dataset and establishing a human-in-the-loop framework to improve the segmentation results. In the proposed approach, we referred the top 10% most uncertain segmentations as detected by our dQC tool to the human expert for refinement. This approach resulted in a significant increase in the Dice score (p < 0.001) and a notable decrease in the number of images with failed segmentation (16.2% to 11.3%) whereas the alternative approach of randomly selecting the same number of segmentations for human referral did not achieve any significant improvement. Our results suggest that the proposed dQC framework has the potential to accurately identify poor-quality segmentations and may enable efficient DNN-based analysis of DCE-CMRI in a human-in-the-loop pipeline for clinical interpretation and reporting of dynamic CMRI datasets.

The association of obesity and obesity-related conditions with carotid extra-medial thickness in children and adolescents.

BACKGROUND: Carotid extra-medial thickness (EMT) and carotid intima-media thickness (cIMT) provide information concerning vascular changes. OBJECTIVES: In this study, we aimed to evaluate the association between carotid EMT and obesity and its metabolic complications in children. METHODS: The study included 38 obese subjects and 30 age-matched and sex-matched healthy controls aged between 7 and 17 years. For all subjects, complete blood count, fasting blood glucose, serum insulin, aspartate aminotransferase, alanine aminotransferase, HDL cholesterol, total cholesterol and triglyceride levels were measured. The carotid EMT and cIMT were measured by an expert radiologist in all patients. RESULTS: Body mass index (BMI) (28.8 ± 3 vs. 18.1 ± 2.2, p < 0.001), total cholesterol (167.9 ± 34.8 mg dL-1 vs. 150.5 ± 28.1 mg dL-1 , p = 0.029), homeostatic model assessment of insulin resistance (HOMA-IR) (4.3 vs. 1.7, p < 0.001), cIMT (0.51 ± 0.08 mm vs. 0.45 ± 0.06 mm, p < 0.001) and carotid EMT (0.74 ± 0.11 mm vs. 0.64 ± 0.1 mm, p < 0.001) were significantly higher in obese subjects than in controls, while HDL cholesterol (41.6 ± 6.5 mg dL-1 vs. 49.5 ± 7.5 mg dL-1 , p < 0.001) was lower in obesity group. Among the obese subjects, the HOMA-IR values (4.7 vs. 3.6, p = 0.027), cIMT (0.54 ± 0.07 mm vs. 0.49 ± 0.07 mm, p = 0.039) and carotid EMT (0.79 ± 0.1 mm vs. 0.7 ± 0.1 mm, p = 0.013) were significantly higher in post-pubertal children compared with prepubertal children. BMI, cut-off values of HOMA-IR and cIMT were significantly associated with increased carotid EMT (p < 0.001, p = 0.023 and p < 0.001, respectively). The only independent risk factor affecting carotid EMT was BMI (p < 0.001). CONCLUSION: We have found that carotid EMT is associated with cIMT, obesity and insulin resistance and the assessment of carotid EMT may provide additional information concerning early vascular disease.

Is localized prostate cancer an obstacle for an immediate consideration for renal transplantation? A case report.

Currently, renal failure patients with a history of prostate cancer are recommended to have a 2 to 5-year disease-free interval prior to being allowed to receive a kidney transplant. This disease is now amenable to curative therapy if diagnosed at an early stage when the tumor is organ-confined. We report a patient undergoing immediate renal transplantation following a laparoscopic radical prostatectomy for the treatment of prostate cancer. Candidates for renal transplantation who are diagnosed with early stage, organ-confined prostate cancer may be immediately considered for transplantation following radical prostatectomy in view of the high likelihood of cure of their prostate cancer.

First A2-to-O renal transplantation in the 21st Century: a step in the transplantation history of Turkey.

Renal transplantation is the best treatment modality for patients with end-stage renal disease. Turkey is a country with limited cadaveric donor organ programs. Herein we have reported the first A2-to-O living donor kidney transplantation in Turkey. A 20-year-old female patient was admitted for a living related renal transplantation from her only potential donor her mother. She was blood group O and her mother was blood group A2. Three plasmapheresis sessions followed by intravenous immunoglobulin (IVIG) were performed every other day in the week prior to transplantation. Daclizumab was administered at the time of transplantation with an additional four doses every 2 weeks after the procedure. The immunsuppressive regimen included tacrolimus, mycophenolate mofetil, and prednisolone. Eight plasmapheresis sessions followed by IVIG were performed in the first 2 weeks posttransplant. Six months after transplantation, the serum creatinine was 1 mg/dL. Our experience showed that A2-to-O renal transplantation can be safely performed and may expand the pool of living kidney donors in Turkey.

An unusual presentation of sirolimus associated cough in a renal transplant recipient.

Sirolimus-associated pulmonary problems are rare but life threatening. Pulmonary problems due to sirolimus treatment are interstitial pneumonitis, bronchiolitis obliterans organizing pneumonia (BOOP), and alveolar hemorrhage. We present a case of sirolimus-related cough in the absence of any pulmonary radiological findings. A 55-year-old man with a history of 4 years of hemodialysis therapy because of end-stage renal disease of unknown etiology underwent cadaveric renal transplantation in June 2006. Three days following the initiation of sirolimus therapy he complained of dry cough and fever. There were no clinical or laboratory findings compatible with specific pulmonary disease. After switching sirolimus to tacrolimus, the cough improved within 1-2 days and resolved in 5 days. Sirolimus should be considered in the differential diagnosis of pulmonary problems in the early posttransplantation period even in the absence of radiological findings.

Seroprevalence of hepatitis C virus infection and evaluation of serum aminotransferase levels among haemodialysis patients in Izmir, Turkey.

The seroprevalence of hepatitis C virus (HCV) infection was investigated among haemodialysis (HD) patients. Mean serum aminotransferase levels were also compared over 3 months in HCV-seropositive patients with and without viraemia, as well as in HCV-seronegative HD patients and HCV-seropositive, non-uraemic, viraemic patients. Seroprevalence of HCV infection was 19% among the 437 HD patients tested. Of the 61 HD HCV-seropositive, hepatotoxic medication- and alcohol-free patients, 38 (62%) were found to be viraemic, using quantitative HCV-RNA, on at least one occasion. Mean serum aminotransferase levels were significantly higher in viraemic HD patients (compared with non-viraemic patients), suggesting that HCV-RNA positivity is an important predictor of increased enzyme activity in these patients. As expected, aminotransferase levels in HCV-seropositive HD patients tended to be lower than levels in HCV-seropositive non-uraemic patients.

Behçet's disease and diverticulosis.

Behçet's disease (BD) is a multisystem disorder characterized by vasculitis. The aim of this report is to present a patient with BD and diverticular disease of the colon and discuss the possible association between BD and diverticulosis. To our knowledge, diverticular disease of the colon has not been previously reported in a patient with BD. We conclude that the significance of this association between BD and diverticulosis needs to be clarified.