Convolutional Neural Network Ensemble Segmentation With Ratio-Based Sampling for the Arteries and Veins in Abdominal CT Scans.

OBJECTIVE: Three-dimensional (3D) blood vessel structure information is important for diagnosis and treatment in various clinical scenarios. We present a fully automatic method for the extraction and differentiation of the arterial and venous vessel trees from abdominal contrast enhanced computed tomography (CE-CT) volumes using convolutional neural networks (CNNs). METHODS: We used a novel ratio-based sampling method to train 2D and 3D versions of the U-Net, the V-Net and the DeepVesselNet. Networks were trained with a combination of the Dice and cross entropy loss. Performance was evaluated on 20 IRCAD subjects. Best performing networks were combined into an ensemble. We investigated seven different weighting schemes. Trained networks were additionally applied to 26 BTCV cases to validate the generalizability. RESULTS: Based on our experiments, the optimal configuration is an equally weighted ensemble of 2D and 3D U- and V-Nets. Our method achieved Dice similarity coefficients of 0.758 ± 0.050 (veins) and 0.838 ± 0.074 (arteries) on the IRCAD data set. Application to the BTCV data set showed a high transfer ability. CONCLUSION: Abdominal vascular structures can be segmented more accurately using ensembles than individual CNNs. 2D and 3D networks have complementary strengths and weaknesses. Our ensemble of 2D and 3D U-Nets and V-Nets in combination with ratio-based sampling achieves a high agreement with manual annotations for both artery and vein segmentation. Our results surpass other state-of-the-art methods. SIGNIFICANCE: Our segmentation pipeline can provide valuable information for the planning of living donor organ transplantations.

Deterministic Arterial Input Function selection in DCE-MRI for automation of quantitative perfusion calculation of colorectal cancer.

Development of a deterministic algorithm for automated detection of the Arterial Input Function (AIF) in DCE-MRI of colorectal cancer. Using a filter pipeline to determine the AIF region of interest. Comparison to algorithms from literature with mean squared error and quantitative perfusion parameter Ktrans. The AIF found by our algorithm has a lower mean squared error (0.0022 ±â€¯0.0021) in reference to the manual annotation than comparable algorithms. The error of Ktrans (21.52 ±â€¯17.2%) is lower than that of other algorithms. Our algorithm generates reproducible results and thus supports a robust and comparable perfusion analysis.

Simulation-based deep artifact correction with Convolutional Neural Networks for limited angle artifacts.

Non-conventional scan trajectories for interventional three-dimensional imaging promise low-dose interventions and a better radiation protection to the personnel. Circular tomosynthesis (cTS) scan trajectories yield an anisotropical image quality distribution. In contrast to conventional Computed Tomographies (CT), the reconstructions have a preferred focus plane. In the other two perpendicular planes, limited angle artifacts are introduced. A reduction of these artifacts leads to enhanced image quality while maintaining the low dose. We apply Deep Artifact Correction (DAC) to this task. cTS simulations of a digital phantom are used to generate training data. Three U-Net-based networks and a 3D-ResNet are trained to estimate the correction map between the cTS and the phantom. We show that limited angle artifacts can be mitigated using simulation-based DAC. The U-Net-corrected cTS achieved a Root Mean Squared Error (RMSE) of 124.24 Hounsfield Units (HU) on 60 simulated test scans in comparison to the digital phantoms. This equals an error reduction of 59.35% from the cTS. The achieved image quality is similar to a simulated cone beam CT (CBCT). Our network was also able to mitigate artifacts in scans of objects which strongly differ from the training data. Application to real cTS test scans showed an error reduction of 45.18% and 26.4% with the 3D-ResNet in reference to a high-dose CBCT.

Synthesis of CT images from digital body phantoms using CycleGAN.

PURPOSE: The potential of medical image analysis with neural networks is limited by the restricted availability of extensive data sets. The incorporation of synthetic training data is one approach to bypass this shortcoming, as synthetic data offer accurate annotations and unlimited data size. METHODS: We evaluated eleven CycleGAN for the synthesis of computed tomography (CT) images based on XCAT body phantoms. The image quality was assessed in terms of anatomical accuracy and realistic noise properties. We performed two studies exploring various network and training configurations as well as a task-based adaption of the corresponding loss function. RESULTS: The CycleGAN using the Res-Net architecture and three XCAT input slices achieved the best overall performance in the configuration study. In the task-based study, the anatomical accuracy of the generated synthetic CTs remained high ([Formula: see text] and [Formula: see text]). At the same time, the generated noise texture was close to real data with a noise power spectrum correlation coefficient of [Formula: see text]. Simultaneously, we observed an improvement in annotation accuracy of 65% when using the dedicated loss function. The feasibility of a combined training on both real and synthetic data was demonstrated in a blood vessel segmentation task (dice similarity coefficient [Formula: see text]). CONCLUSION: CT synthesis using CycleGAN is a feasible approach to generate realistic images from simulated XCAT phantoms. Synthetic CTs generated with a task-based loss function can be used in addition to real data to improve the performance of segmentation networks.

Tomosynthesis implementation with adaptive online calibration on clinical C-arm systems.

PURPOSE: Cone beam computed tomography (CBCT) systems offer physicians crucial 3D and 2D imaging capabilities during interventions. However, certain medical applications only require very specific information from the CBCTs (e.g., determination of the position of high-contrast objects). In diagnostics, tomosynthesis techniques can be used in these cases to minimize dose exposure. Therefore, integrating such techniques on CBCT systems could also be beneficial for interventions. In this paper, we investigate the performance of our implementation of circular tomosynthesis on a CBCT device. METHODS: The tomosynthesis scan trajectory is realized with step-and-shoot on a clinical C-arm device. The online calibration algorithm uses conventionally acquired 3D CBCT of the scanned object as prior knowledge to correct the imaging geometries. The online calibration algorithm was compared to an offline calibration to test its performance. A ball bearing phantom was used to evaluate the reconstructions with respect to geometric distortions. The evaluation was done for three different scenarios to test the robustness of our tomosynthesis implementation against object deviations (e.g., pen) and different object positioning. RESULTS: The circular tomosynthesis was tested on a ball bearing and an anthropomorphic phantom. The results show that the calibration is robust against isocenter shifts and object deviations in the CBCT. All reconstructions used 100 projections and displayed limited angle artifacts. The accuracy of the positions and shapes of high-contrast objects were, however, determined precisely. (The maximal center position deviation is 0.31 mm.) CONCLUSION: For medical procedures that primarily determine the precise position of high-contrast objects, circular tomosynthesis could offer an approach to reduce dose exposure.

Correction to: Tomosynthesis implementation with adaptive online calibration on clinical C-arm systems.

The original version of this article unfortunately contained a mistake in table 2.