Cervical spine fracture detection in computed tomography using convolutional neural networks.

Objective.In the context of primary in-hospital trauma management timely reading of computed tomography (CT) images is critical. However, assessment of the spine is time consuming, fractures can be very subtle, and the potential for under-diagnosis or delayed diagnosis is relevant. Artificial intelligence is increasingly employed to assist radiologists with the detection of spinal fractures and prioritization of cases. Currently, algorithms focusing on the cervical spine are commercially available. A common approach is the vertebra-wise classification. Instead of a classification task, we formulate fracture detection as a segmentation task aiming to find and display all individual fracture locations presented in the image.Approach.Based on 195 CT examinations, 454 cervical spine fractures were identified and annotated by radiologists at a tertiary trauma center. We trained for the detection a U-Net via four-fold-cross validation to segment spine fractures and the spine via a multi-task loss. We further compared advantages of two image reformation approaches-straightened curved planar reformatted (CPR) around the spine and spinal canal aligned volumes of interest (VOI)-to achieve a unified vertebral alignment in comparison to processing the Cartesian data directly.Main results.Of the three data versions (Cartesian, reformatted, VOI) the VOI approach showed the best detection rate and a reduced computation time. The proposed algorithm was able to detect 87.2% of cervical spine fractures at an average number of false positives of 3.5 per case. Evaluation of the method on a public spine dataset resulted in 0.9 false positive detections per cervical spine case.Significance.The display of individual fracture locations as provided with high sensitivity by the proposed voxel classification based fracture detection has the potential to support the trauma CT reading workflow by reducing missed findings.

Fast CBCT Reconstruction using Convolutional Neural Networks for Arbitrary Robotic C-arm Orbits.

Cone-beam CT (CBCT) with non-circular acquisition orbits has the potential to improve image quality, increase the field-of view, and facilitate minimal interference within an interventional imaging setting. Because time is of the essence in interventional imaging scenarios, rapid reconstruction methods are advantageous. Model-Based Iterative Reconstruction (MBIR) techniques implicitly handle arbitrary geometries; however, the computational burden for these approaches is particularly high. The aim of this work is to extend a previously proposed framework for fast reconstruction of non-circular CBCT trajectories. The pipeline combines a deconvolution operation on the backprojected measurements using an approximate, shift-invariant system response prior to processing with a Convolutional Neural Network (CNN). We trained and evaluated the CNN for this approach using 1800 randomized arbitrary orbits. Noisy projection data were formed from 1000 procedurally generated tetrahedral phantoms as well as anthropomorphic data in the form of 800 CT and CBCT images from the Lung Image Database Consortium Image Collection (LIDC). Using this proposed reconstruction pipeline, computation time was reduced by 90% as compared to MBIR with only minor differences in performance. Quantitative comparisons of nRMSE, FSIM and SSIM are reported. Performance was consistent for projection data simulated with acquisition orbits the network has not previously been trained on. These results suggest the potential for fast processing of arbitrary CBCT trajectory data with reconstruction times that are clinically relevant and applicable - facilitating the application of non-circular orbits in CT image-guided interventions and intraoperative imaging.

Development of an abdominal phantom for the validation of an oligometastatic disease diagnosis workflow.

PURPOSE: The liver is a common site for metastatic disease, which is a challenging and life-threatening condition with a grim prognosis and outcome. We propose a standardized workflow for the diagnosis of oligometastatic disease (OMD), as a gold standard workflow has not been established yet. The envisioned workflow comprises the acquisition of a multimodal image data set, novel image processing techniques, and cone beam computed tomography (CBCT)-guided biopsy for subsequent molecular subtyping. By combining morphological, molecular, and functional information about the tumor, a patient-specific treatment planning is possible. We designed and manufactured an abdominal liver phantom that we used to demonstrate multimodal image acquisition, image processing, and biopsy of the OMD diagnosis workflow. METHODS: The anthropomorphic abdominal phantom contains a rib cage, a portal vein, lungs, a liver with six lesions, and a hepatic vessel tree. This phantom incorporates three different lesion types with varying visibility under computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography CT (PET-CT), which reflects clinical reality. The phantom is puncturable and the size of the corpus and the organs is comparable to those of a real human abdomen. By using several modern additive manufacturing techniques, the manufacturing process is reproducible and allows to incorporate patient-specific anatomies. As a first step of the OMD diagnosis workflow, a preinterventional CT, MRI, and PET-CT data set of the phantom was acquired. The image information was fused using image registration and organ information was extracted via image segmentation. A CBCT-guided needle puncture experiment was performed, where all six liver lesions were punctured with coaxial biopsy needles. RESULTS: Qualitative observation of the image data and quantitative evaluation using contrast-to-noise ratio (CNR) confirms that one lesion type is visible only in MRI and not CT. The other two lesion types are visible in CT and MRI. The CBCT-guided needle placement was performed for all six lesions, including those visible only in MRI and not CBCT. This was possible by successfully merging multimodal preinterventional image data. Lungs, bones, and liver vessels serve as realistic inhibitions during needle path planning. CONCLUSIONS: We have developed a reusable abdominal phantom that has been used to validate a standardized OMD diagnosis workflow. Utilizing the phantom, we have been able to show that a multimodal imaging pipeline is advantageous for a comprehensive detection of liver lesions. In a CBCT-guided needle placement experiment we have punctured lesions that are invisible in CBCT using registered preinterventional MRI scans for needle path planning.

Deep Learning-Based Total Kidney Volume Segmentation in Autosomal Dominant Polycystic Kidney Disease Using Attention, Cosine Loss, and Sharpness Aware Minimization.

Early detection of the autosomal dominant polycystic kidney disease (ADPKD) is crucial as it is one of the most common causes of end-stage renal disease (ESRD) and kidney failure. The total kidney volume (TKV) can be used as a biomarker to quantify disease progression. The TKV calculation requires accurate delineation of kidney volumes, which is usually performed manually by an expert physician. However, this is time-consuming and automated segmentation is warranted. Furthermore, the scarcity of large annotated datasets hinders the development of deep learning solutions. In this work, we address this problem by implementing three attention mechanisms into the U-Net to improve TKV estimation. Additionally, we implement a cosine loss function that works well on image classification tasks with small datasets. Lastly, we apply a technique called sharpness aware minimization (SAM) that helps improve the generalizability of networks. Our results show significant improvements (p-value < 0.05) over the reference kidney segmentation U-Net. We show that the attention mechanisms and/or the cosine loss with SAM can achieve a dice score (DSC) of 0.918, a mean symmetric surface distance (MSSD) of 1.20 mm with the mean TKV difference of −1.72%, and R2 of 0.96 while using only 100 MRI datasets for training and testing. Furthermore, we tested four ensembles and obtained improvements over the best individual network, achieving a DSC and MSSD of 0.922 and 1.09 mm, respectively.

An anthropomorphic pelvis phantom for MR-guided prostate interventions.

PURPOSE: To design and manufacture a pelvis phantom for magnetic resonance (MR)-guided prostate interventions, such as MRGB (MR-guided biopsy) or brachytherapy seed placement. METHODS: The phantom was designed to mimic the human pelvis incorporating bones, bladder, prostate with four lesions, urethra, arteries, veins, and six lymph nodes embedded in ballistic gelatin. A hollow rectum enables transrectal access to the prostate. To demonstrate the feasibility of the phantom for minimal invasive MRI-guided interventions, a targeted inbore MRGB was performed. The needle probe was rectally inserted and guided using an MRI-compatible remote controlled manipulator (RCM). RESULTS: The presented pelvis phantom has realistic imaging properties for MR imaging (MRI), computed tomography (CT) and ultrasound (US). In the targeted inbore MRGB, a prostate lesion was successfully hit with an accuracy of 3.5 mm. The experiment demonstrates that the limited size of the rectum represents a realistic impairment for needle placements. CONCLUSION: The phantom provides a valuable platform for evaluating the performance of MRGB systems. Interventionalists can use the phantom to learn how to deal with challenging situations, without risking harm to patients.

Generation of annotated multimodal ground truth datasets for abdominal medical image registration.

PURPOSE: Sparsity of annotated data is a major limitation in medical image processing tasks such as registration. Registered multimodal image data are essential for the diagnosis of medical conditions and the success of interventional medical procedures. To overcome the shortage of data, we present a method that allows the generation of annotated multimodal 4D datasets. METHODS: We use a CycleGAN network architecture to generate multimodal synthetic data from the 4D extended cardiac-torso (XCAT) phantom and real patient data. Organ masks are provided by the XCAT phantom; therefore, the generated dataset can serve as ground truth for image segmentation and registration. Realistic simulation of respiration and heartbeat is possible within the XCAT framework. To underline the usability as a registration ground truth, a proof of principle registration is performed. RESULTS: Compared to real patient data, the synthetic data showed good agreement regarding the image voxel intensity distribution and the noise characteristics. The generated T1-weighted magnetic resonance imaging, computed tomography (CT), and cone beam CT images are inherently co-registered. Thus, the synthetic dataset allowed us to optimize registration parameters of a multimodal non-rigid registration, utilizing liver organ masks for evaluation. CONCLUSION: Our proposed framework provides not only annotated but also multimodal synthetic data which can serve as a ground truth for various tasks in medical imaging processing. We demonstrated the applicability of synthetic data for the development of multimodal medical image registration algorithms.

Accelerated white matter lesion analysis based on simultaneous T1 and T2∗ quantification using magnetic resonance fingerprinting and deep learning.

PURPOSE: To develop an accelerated postprocessing pipeline for reproducible and efficient assessment of white matter lesions using quantitative magnetic resonance fingerprinting (MRF) and deep learning. METHODS: MRF using echo-planar imaging (EPI) scans with varying repetition and echo times were acquired for whole brain quantification of T1 and T2∗ in 50 subjects with multiple sclerosis (MS) and 10 healthy volunteers along 2 centers. MRF T1 and T2∗ parametric maps were distortion corrected and denoised. A CNN was trained to reconstruct the T1 and T2∗ parametric maps, and the WM and GM probability maps. RESULTS: Deep learning-based postprocessing reduced reconstruction and image processing times from hours to a few seconds while maintaining high accuracy, reliability, and precision. Mean absolute error performed the best for T1 (deviations 5.6%) and the logarithmic hyperbolic cosinus loss the best for T2∗ (deviations 6.0%). CONCLUSIONS: MRF is a fast and robust tool for quantitative T1 and T2∗ mapping. Its long reconstruction and several postprocessing steps can be facilitated and accelerated using deep learning.

23 Na MRI in ischemic stroke: Acquisition time reduction using postprocessing with convolutional neural networks.

Quantitative 23 Na magnetic resonance imaging (MRI) provides tissue sodium concentration (TSC), which is connected to cell viability and vitality. Long acquisition times are one of the most challenging aspects for its clinical establishment. K-space undersampling is an approach for acquisition time reduction, but generates noise and artifacts. The use of convolutional neural networks (CNNs) is increasing in medical imaging and they are a useful tool for MRI postprocessing. The aim of this study is 23 Na MRI acquisition time reduction by k-space undersampling. CNNs were applied to reduce the resulting noise and artifacts. A retrospective analysis from a prospective study was conducted including image datasets from 46 patients (aged 72 ± 13 years; 25 women, 21 men) with ischemic stroke; the 23 Na MRI acquisition time was 10 min. The reconstructions were performed with full dataset (FI) and with a simulated dataset an image that was acquired in 2.5 min (RI). Eight different CNNs with either U-Net-based or ResNet-based architectures were implemented with RI as input and FI as label, using batch normalization and the number of filters as varying parameters. Training was performed with 9500 samples and testing included 400 samples. CNN outputs were evaluated based on signal-to-noise ratio (SNR) and structural similarity (SSIM). After quantification, TSC error was calculated. The image quality was subjectively rated by three neuroradiologists. Statistical significance was evaluated by Student's t-test. The average SNR was 21.72 ± 2.75 (FI) and 10.16 ± 0.96 (RI). U-Nets increased the SNR of RI to 43.99 and therefore performed better than ResNet. SSIM of RI to FI was improved by three CNNs to 0.91 ± 0.03. CNNs reduced TSC error by up to 15%. The subjective rating of CNN-generated images showed significantly better results than the subjective image rating of RI. The acquisition time of 23 Na MRI can be reduced by 75% due to postprocessing with a CNN on highly undersampled data.

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.

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.