Smart forecasting of artifacts in contrast-enhanced breast MRI before contrast agent administration.

OBJECTIVES: To evaluate whether artifacts on contrast-enhanced (CE) breast MRI maximum intensity projections (MIPs) might already be forecast before gadolinium-based contrast agent (GBCA) administration during an ongoing examination by analyzing the unenhanced T1-weighted images acquired before the GBCA injection. MATERIALS AND METHODS: This IRB-approved retrospective analysis consisted of n = 2884 breast CE MRI examinations after intravenous administration of GBCA, acquired with n = 4 different MRI devices at different field strengths (1.5 T/3 T) during clinical routine. CE-derived subtraction MIPs were used to conduct a multi-class multi-reader evaluation of the presence and severity of artifacts with three independent readers. An ensemble classifier (EC) of five DenseNet models was used to predict artifacts for the post-contrast subtraction MIPs, giving as the input source only the pre-contrast T1-weighted sequence. Thus, the acquisition directly preceded the GBCA injection. The area under ROC (AuROC) and diagnostics accuracy scores were used to assess the performance of the neural network in an independent holdout test set (n = 285). RESULTS: After majority voting, potentially significant artifacts were detected in 53.6% (n = 1521) of all breast MRI examinations (age 49.6 ± 12.6 years). In the holdout test set (mean age 49.7 ± 11.8 years), at a specificity level of 89%, the EC could forecast around one-third of artifacts (sensitivity 31%) before GBCA administration, with an AuROC = 0.66. CONCLUSION: This study demonstrates the capability of a neural network to forecast the occurrence of artifacts on CE subtraction data before the GBCA administration. If confirmed in larger studies, this might enable a workflow-blended approach to prevent breast MRI artifacts by implementing in-scan personalized predictive algorithms. CLINICAL RELEVANCE STATEMENT: Some artifacts in contrast-enhanced breast MRI maximum intensity projections might be predictable before gadolinium-based contrast agent injection using a neural network. KEY POINTS: • Potentially significant artifacts can be observed in a relevant proportion of breast MRI subtraction sequences after gadolinium-based contrast agent administration (GBCA). • Forecasting the occurrence of such artifacts in subtraction maximum intensity projections before GBCA administration for individual patients was feasible at 89% specificity, which allowed correctly predicting one in three future artifacts. • Further research is necessary to investigate the clinical value of such smart personalized imaging approaches.

Advanced neural networks for classification of MRI in psoriatic arthritis, seronegative, and seropositive rheumatoid arthritis.

OBJECTIVES: To evaluate whether neural networks can distinguish between seropositive RA, seronegative RA, and PsA based on inflammatory patterns from hand MRIs and to test how psoriasis patients with subclinical inflammation fit into such patterns. METHODS: ResNet neural networks were utilized to compare seropositive RA vs PsA, seronegative RA vs PsA, and seropositive vs seronegative RA with respect to hand MRI data. Results from T1 coronal, T2 coronal, T1 coronal and axial fat-suppressed contrast-enhanced (CE), and T2 fat-suppressed axial sequences were used. The performance of such trained networks was analysed by the area under the receiver operating characteristics curve (AUROC) with and without presentation of demographic and clinical parameters. Additionally, the trained networks were applied to psoriasis patients without clinical arthritis. RESULTS: MRI scans from 649 patients (135 seronegative RA, 190 seropositive RA, 177 PsA, 147 psoriasis) were fed into ResNet neural networks. The AUROC was 75% for seropositive RA vs PsA, 74% for seronegative RA vs PsA, and 67% for seropositive vs seronegative RA. All MRI sequences were relevant for classification, however, when deleting contrast agent-based sequences the loss of performance was only marginal. The addition of demographic and clinical data to the networks did not provide significant improvements for classification. Psoriasis patients were mostly assigned to PsA by the neural networks, suggesting that a PsA-like MRI pattern may be present early in the course of psoriatic disease. CONCLUSION: Neural networks can be successfully trained to distinguish MRI inflammation related to seropositive RA, seronegative RA, and PsA.

Automated artifact detection in abbreviated dynamic contrast-enhanced (DCE) MRI-derived maximum intensity projections (MIPs) of the breast.

OBJECTIVES: To automatically detect MRI artifacts on dynamic contrast-enhanced (DCE) maximum intensity projections (MIPs) of the breast using deep learning. METHODS: Women who underwent clinically indicated breast MRI between October 2015 and December 2019 were included in this IRB-approved retrospective study. We employed two convolutional neural network architectures (ResNet and DenseNet) to detect the presence of artifacts on DCE MIPs of the left and right breasts. Networks were trained on images acquired up to and including the year 2018 using a 5-fold cross-validation (CV). Ensemble classifiers were built with the resulting CV models and applied to an independent holdout test dataset, which was formed by images acquired in 2019. RESULTS: Our study sample contained 2265 examinations from 1794 patients (median age at first acquisition: 50 years [IQR: 17 years]), corresponding to 1827 examinations of 1378 individuals in the training dataset and 438 examinations of 416 individuals in the holdout test dataset with a prevalence of image-level artifacts of 53% (1951/3654 images) and 43% (381/876 images), respectively. On the holdout test dataset, the ResNet and DenseNet ensembles demonstrated an area under the ROC curve of 0.92 and 0.94, respectively. CONCLUSION: Neural networks are able to reliably detect artifacts that may impede the diagnostic assessment of MIPs derived from DCE subtraction series in breast MRI. Future studies need to further explore the potential of such neural networks to complement quality assurance and improve the application of DCE MIPs in a clinical setting, such as abbreviated protocols. KEY POINTS: • Deep learning classifiers are able to reliably detect MRI artifacts in dynamic contrast-enhanced protocol-derived maximum intensity projections of the breast. • Automated quality assurance of maximum intensity projections of the breast may be of special relevance for abbreviated breast MRI, e.g., in high-throughput settings, such as cancer screening programs.

Detection of femoropopliteal arterial steno-occlusion at MR angiography: initial experience with artificial intelligence.

BACKGROUND: This study evaluated a deep learning (DL) algorithm for detecting vessel steno-occlusions in patients with peripheral arterial disease (PAD). It utilised a private dataset, which was acquired and annotated by the authors through their institution and subsequently validated by two blinded readers. METHODS: A single-centre retrospective study analysed 105 magnetic resonance angiography (MRA) images using an EfficientNet B0 DL model. Initially, inter-reader variability was assessed using the complete dataset. For a subset of these images (29 from the left side and 35 from the right side) where digital subtraction angiography (DSA) data was available as the ground truth, the model's accuracy and the area under the curve at receiver operating characteristics analysis (ROC-AUC) were evaluated. RESULTS: A total of 105 patient examinations (mean age, 75 years ±12 [mean ± standard deviation], 61 men) were evaluated. Radiologist-DL model agreement had a quadratic weighted Cohen κ ≥ 0.72 (left side) and ≥ 0.66 (right side). Radiologist inter-reader agreement was ≥ 0.90 (left side) and ≥ 0.87 (right side). The DL model achieved a 0.897 accuracy and a 0.913 ROC-AUC (left side) and 0.743 and 0.830 (right side). Radiologists achieved 0.931 and 0.862 accuracies, with 0.930 and 0.861 ROC-AUCs (left side), and 0.800 and 0.799 accuracies, with 0.771 ROC-AUCs (right side). CONCLUSION: The DL model provided valid results in identifying arterial steno-occlusion in the superficial femoral and popliteal arteries on MRA among PAD patients. However, it did not reach the inter-reader agreement of two radiologists. RELEVANCE STATEMENT: The tested DL model is a promising tool for assisting in the detection of arterial steno-occlusion in patients with PAD, but further optimisation is necessary to provide radiologists with useful support in their daily routine diagnostics. KEY POINTS: • This study focused on the application of DL for arterial steno-occlusion detection in lower extremities on MRA. • A previously developed DL model was tested for accuracy and inter-reader agreement. • While the model showed promising results, it does not yet replace human expertise in detecting arterial steno-occlusion on MRA.

Lesion-conditioning of synthetic MRI-derived subtraction-MIPs of the breast using a latent diffusion model.

The purpose of this feasibility study is to investigate if latent diffusion models (LDMs) are capable to generate contrast enhanced (CE) MRI-derived subtraction maximum intensity projections (MIPs) of the breast, which are conditioned by lesions. We trained an LDM with n = 2832 CE-MIPs of breast MRI examinations of n = 1966 patients (median age: 50 years) acquired between the years 2015 and 2020. The LDM was subsequently conditioned with n = 756 segmented lesions from n = 407 examinations, indicating their location and BI-RADS scores. By applying the LDM, synthetic images were generated from the segmentations of an independent validation dataset. Lesions, anatomical correctness, and realistic impression of synthetic and real MIP images were further assessed in a multi-rater study with five independent raters, each evaluating n = 204 MIPs (50% real/50% synthetic images). The detection of synthetic MIPs by the raters was akin to random guessing with an AUC of 0.58. Interrater reliability of the lesion assessment was high both for real (Kendall's W = 0.77) and synthetic images (W = 0.85). A higher AUC was observed for the detection of suspicious lesions (BI-RADS ≥ 4) in synthetic MIPs (0.88 vs. 0.77; p = 0.051). Our results show that LDMs can generate lesion-conditioned MRI-derived CE subtraction MIPs of the breast, however, they also indicate that the LDM tended to generate rather typical or 'textbook representations' of lesions.

DeepNAPSI multi-reader nail psoriasis prediction using deep learning.

Nail psoriasis occurs in about every second psoriasis patient. Both, finger and toe nails can be affected and also severely destroyed. Furthermore, nail psoriasis is associated with a more severe course of the disease and the development of psoriatic arthritis. User independent quantification of nail psoriasis, however, is challenging due to the heterogeneous involvement of matrix and nail bed. For this purpose, the nail psoriasis severity index (NAPSI) has been developed. Experts grade pathological changes of each nail of the patient leading to a maximum score of 80 for all nails of the hands. Application in clinical practice, however, is not feasible due to the time-intensive manual grading process especially if more nails are involved. In this work we aimed to automatically quantify the modified NAPSI (mNAPSI) of patients using neuronal networks retrospectively. First, we performed photographs of the hands of patients with psoriasis, psoriatic arthritis, and rheumatoid arthritis. In a second step, we collected and annotated the mNAPSI scores of 1154 nail photos. Followingly, we extracted each nail automatically using an automatic key-point-detection system. The agreement among the three readers with a Cronbach's alpha of 94% was very high. With the nail images individually available, we trained a transformer-based neural network (BEiT) to predict the mNAPSI score. The network reached a good performance with an area-under-receiver-operator-curve of 88% and an area-under precision-recall-curve (PR-AUC) of 63%. We could compare the results with the human annotations and achieved a very high positive Pearson correlation of 90% by aggregating the predictions of the network on the test set to the patient-level. Lastly, we provided open access to the whole system enabling the use of the mNAPSI in clinical practice.

Image quality assessment using deep learning in high b-value diffusion-weighted breast MRI.

The objective of this IRB approved retrospective study was to apply deep learning to identify magnetic resonance imaging (MRI) artifacts on maximum intensity projections (MIP) of the breast, which were derived from diffusion weighted imaging (DWI) protocols. The dataset consisted of 1309 clinically indicated breast MRI examinations of 1158 individuals (median age [IQR]: 50 years [16.75 years]) acquired between March 2017 and June 2020, in which a DWI sequence with a high b-value equal to 1500 s/mm2 was acquired. From these, 2D MIP images were computed and the left and right breast were cropped out as regions of interest (ROI). The presence of MRI image artifacts on the ROIs was rated by three independent observers. Artifact prevalence in the dataset was 37% (961 out of 2618 images). A DenseNet was trained with a fivefold cross-validation to identify artifacts on these images. In an independent holdout test dataset (n = 350 images) artifacts were detected by the neural network with an area under the precision-recall curve of 0.921 and a positive predictive value of 0.981. Our results show that a deep learning algorithm is capable to identify MRI artifacts in breast DWI-derived MIPs, which could help to improve quality assurance approaches for DWI sequences of breast examinations in the future.

Imaging in inflammatory arthritis: progress towards precision medicine.

Imaging techniques such as ultrasonography and MRI have gained ground in the diagnosis and management of inflammatory arthritis, as these imaging modalities allow a sensitive assessment of musculoskeletal inflammation and damage. However, these techniques cannot discriminate between disease subsets and are currently unable to deliver an accurate prediction of disease progression and therapeutic response in individual patients. This major shortcoming of today's technology hinders a targeted and personalized patient management approach. Technological advances in the areas of high-resolution imaging (for example, high-resolution peripheral quantitative computed tomography and ultra-high field MRI), functional and molecular-based imaging (such as chemical exchange saturation transfer MRI, positron emission tomography, fluorescence optical imaging, optoacoustic imaging and contrast-enhanced ultrasonography) and artificial intelligence-based data analysis could help to tackle these challenges. These new imaging approaches offer detailed anatomical delineation and an in vivo and non-invasive evaluation of the immunometabolic status of inflammatory reactions, thereby facilitating an in-depth characterization of inflammation. By means of these developments, the aim of earlier diagnosis, enhanced monitoring and, ultimately, a personalized treatment strategy looms closer.

Gradient-based geometry learning for fan-beam CT reconstruction.

Objective.Incorporating computed tomography (CT) reconstruction operators into differentiable pipelines has proven beneficial in many applications. Such approaches usually focus on the projection data and keep the acquisition geometry fixed. However, precise knowledge of the acquisition geometry is essential for high quality reconstruction results. In this paper, the differentiable formulation of fan-beam CT reconstruction is extended to the acquisition geometry.Approach.The CT fan-beam reconstruction is analytically derived with respect to the acquisition geometry. This allows to propagate gradient information from a loss function on the reconstructed image into the geometry parameters. As a proof-of-concept experiment, this idea is applied to rigid motion compensation. The cost function is parameterized by a trained neural network which regresses an image quality metric from the motion-affected reconstruction alone.Main results.The algorithm improves the structural similarity index measure (SSIM) from 0.848 for the initial motion-affected reconstruction to 0.946 after compensation. It also generalizes to real fan-beam sinograms which are rebinned from a helical trajectory where the SSIM increases from 0.639 to 0.742.Significance.Using the proposed method, we are the first to optimize an autofocus-inspired algorithm based on analytical gradients. Next to motion compensation, we see further use cases of our differentiable method for scanner calibration or hybrid techniques employing deep models.

Deep Learning-Based Classification of Inflammatory Arthritis by Identification of Joint Shape Patterns-How Neural Networks Can Tell Us Where to "Deep Dive" Clinically.

Objective: We investigated whether a neural network based on the shape of joints can differentiate between rheumatoid arthritis (RA), psoriatic arthritis (PsA), and healthy controls (HC), which class patients with undifferentiated arthritis (UA) are assigned to, and whether this neural network is able to identify disease-specific regions in joints. Methods: We trained a novel neural network on 3D articular bone shapes of hand joints of RA and PsA patients as well as HC. Bone shapes were created from high-resolution peripheral-computed-tomography (HR-pQCT) data of the second metacarpal bone head. Heat maps of critical spots were generated using GradCAM. After training, we fed shape patterns of UA into the neural network to classify them into RA, PsA, or HC. Results: Hand bone shapes from 932 HR-pQCT scans of 617 patients were available. The network could differentiate the classes with an area-under-receiver-operator-curve of 82% for HC, 75% for RA, and 68% for PsA. Heat maps identified anatomical regions such as bare area or ligament attachments prone to erosions and bony spurs. When feeding UA data into the neural network, 86% were classified as "RA," 11% as "PsA," and 3% as "HC" based on the joint shape. Conclusion: We investigated neural networks to differentiate the shape of joints of RA, PsA, and HC and extracted disease-specific characteristics as heat maps on 3D joint shapes that can be utilized in clinical routine examination using ultrasound. Finally, unspecific diseases such as UA could be grouped using the trained network based on joint shape.

Deep learning methods allow fully automated segmentation of metacarpal bones to quantify volumetric bone mineral density.

Arthritis patients develop hand bone loss, which leads to destruction and functional impairment of the affected joints. High resolution peripheral quantitative computed tomography (HR-pQCT) allows the quantification of volumetric bone mineral density (vBMD) and bone microstructure in vivo with an isotropic voxel size of 82 micrometres. However, image-processing to obtain bone characteristics is a time-consuming process as it requires semi-automatic segmentation of the bone. In this work, a fully automatic vBMD measurement pipeline for the metacarpal (MC) bone using deep learning methods is introduced. Based on a dataset of HR-pQCT volumes with MC measurements for 541 patients with arthritis, a segmentation network is trained. The best network achieves an intersection over union as high as 0.94 and a Dice similarity coefficient of 0.97 while taking only 33 s to process a whole patient yielding a speedup between 2.5 and 4.0 for the whole workflow. Strong correlation between the vBMD measurements of the expert and of the automatic pipeline are achieved for the average bone density with 0.999 (Pearson) and 0.996 (Spearman's rank) with [Formula: see text] for all correlations. A qualitative assessment of the network predictions and the manual annotations yields a 65.9% probability that the expert favors the network predictions. Further, the steps to integrate the pipeline into the clinical workflow are shown. In order to make these workflow improvements available to others, we openly share the code of this work.