Automated Patient Registration in Magnetic Resonance Imaging Using Deep Learning-Based Height and Weight Estimation with 3D Camera: A Feasibility Study.

RATIONALE AND OBJECTIVES: Accurate and efficient estimation of patient height and weight is crucial to ensure patient safety and optimize the quality of magnetic resonance imaging (MRI) procedures. Several height and weight estimation methods have been proposed for use in adult patient management, but none is widely established. Estimation by the medical technologists for radiology (MTR) based on personal experience remains to be the most common method. This study aimed to compare a novel deep learning (DL)-based 3-dimensional (3D) camera estimation method to MTR staff in terms of estimation accuracy. METHODS: A retrospective study was conducted to compare the accuracy of height and weight estimation with a DL-based 3D camera algorithm to the accuracy of height and weight estimation by the MTR. Depth images of the patients were captured during the regular imaging workflow on a low field 0.55 T MRI scanner (MAGNETOM Free.Max, Siemens Healthineers, Erlangen, Germany) and then processed retrospectively. Depth images of a total of 161 patients were used to validate the accuracy of the height and weight estimation algorithm. The accuracy of each estimation method was evaluated by computing the proportions of the estimates within 5% and 15% of actual height (PH05, PH15) and within 10% and 20% of actual weight (PW10, PW20). An acceptable accuracy for height estimation was predetermined to be PH05 = 95% and PH15 = 99% and an acceptable accuracy for weight estimation was predetermined to be PW10 = 70% and PW20 = 95%. The bias in height and weight estimation was measured by the mean absolute percentage error (MAPE). RESULTS: The retrospective study included 161 adult patients. For 148/161 patients complying with inclusion criteria, DL-based 3D camera algorithm outperformed the MTR in estimating the patient's height and weight in term of accuracy (3D camera: PH05 =98.6%, PH15 =100%, PW10 =85.1%, PW20 =95.9%; MTR: PH05 =92.5%, PH15 =100%, PW10 =75.0%, PW20 =93.2%). MTR had a slightly higher bias in their estimates compared to the DL-based 3D camera algorithm (3D camera: MAPE height=1.8%, MAPE weight=5.6%, MTR: MAPE height=2.2%, MAPE weight=7.5%) CONCLUSION: This study has demonstrated that the estimation of the patient's height and weight by a DL-based 3D camera algorithm is accurate and robust. It has the potential to complement the regular MRI workflows, by providing further automation during patient registration.

Diagnostic Image Quality of a Low-Field (0.55T) Knee MRI Protocol Using Deep Learning Image Reconstruction Compared with a Standard (1.5T) Knee MRI Protocol.

OBJECTIVES: Low-field MRI at 0.55 Tesla (T) with deep learning image reconstruction has recently become commercially available. The objective of this study was to evaluate the image quality and diagnostic reliability of knee MRI performed at 0.55T compared with 1.5T. METHODS: A total of 20 volunteers (9 female, 11 male; mean age = 42 years) underwent knee MRI on a 0.55T system (MAGNETOM Free.Max, Siemens Healthcare, Erlangen, Germany; 12-channel Contour M Coil) and a 1.5T scanner (MAGNETOM Sola, Siemens Healthcare, Erlangen, Germany; 18-channel transmit/receive knee coil). Standard two-dimensional (2D) turbo spin echo (TSE), fat-suppressed (fs) proton density-weighted (PDw), T1w TSE, and T2w TSE sequences were acquired in approximately 15 min. In total, 2 radiologists blinded to the field strength subjectively assessed all MRI sequences (overall image quality, image noise, and diagnostic quality) using a 5-point Likert scale (1-5; 5 = best). Additionally, both radiologists evaluated the possible pathologies of menisci, ligaments, and cartilage. Contrast ratios (CRs) of different tissues (bone, cartilage, and menisci) were determined on coronal PDw fs TSE images. The statistical analysis included Cohen's kappa and the Wilcoxon rank sum test. RESULTS: The overall image quality of the 0.55T T2w, T1w, and PDw fs TSE sequences was diagnostic and rated similar for T1w (p > 0.05), but lower for PDw fs TSE and T2w TSE compared with 1.5T (p < 0.05). The diagnostic accordance of meniscal and cartilage pathologies at 0.55T was similar to 1.5T. The CRs of the tissues were not significantly different between 1.5T and 0.55T (p > 0.05). The inter-observer agreement of the subjective image quality was generally fair between both readers and almost perfect for the pathologies. CONCLUSIONS: Deep learning-reconstructed TSE imaging at 0.55T yielded diagnostic image quality for knee MRI compared with standard 1.5T MRI. The diagnostic performance of meniscal and cartilage pathologies was equal for 0.55T and 1.5T without a significant loss of diagnostic information.

Voxel-Based Analysis of the Relation of 3'-Deoxy-3'-[18F]fluorothymidine ([18F]FLT) PET and Diffusion-Weighted (DW) MR Signals in Subcutaneous Tumor Xenografts Does Not Reveal a Direct Spatial Relation of These Two Parameters.

PURPOSE: Multimodal molecular imaging allows a direct coregistration of different images, facilitating analysis of the spatial relation of various imaging parameters. Here, we further explored the relation of proliferation, as measured by [18F]FLT PET, and water diffusion, as an indicator of cellular density and cell death, as measured by diffusion-weighted (DW) MRI, in preclinical tumor models. We expected these parameters to be negatively related, as highly proliferative tissue should have a higher density of cells, hampering free water diffusion. PROCEDURES: Nude mice subcutaneously inoculated with either lung cancer cells (n = 11 A549 tumors, n = 20 H1975 tumors) or colorectal cancer cells (n = 13 Colo205 tumors) were imaged with [18F]FLT PET and DW-MRI using a multimodal bed, which was transferred from one instrument to the other within the same imaging session. Fiducial markers allowed coregistration of the images. An automatic post-processing was developed in MATLAB handling the spatial registration of DW-MRI (measured as apparent diffusion coefficient, ADC) and [18F]FLT image data and subsequent voxel-wise analysis of regions of interest (ROIs) in the tumor. RESULTS: Analyses were conducted on a total of 76 datasets, comprising a median of 2890 data points (ranging from 81 to 13,597). Scatterplots showing [18F]FLT vs. ADC values displayed various grades of relations (Pearson correlation coefficient (PCC) varied from - 0.58 to 0.49, median: -0.07). When relating PCC to tumor volume (median: 46 mm3, range: 3 mm3 to 584 mm3), lung tumors tended to have a more pronounced negative spatial relation of [18F]FLT and ADC with increasing tumor size. However, due to the low number of large tumors (> ~ 200 mm3), this conclusion has to be treated with caution. CONCLUSIONS: A spatial relation of water diffusion, as measured by DW-MRI, and cellular proliferation, as measured by [18F]FLT PET, cannot be detected in the experimental datasets investigated in this study.

Spatio-Temporal Multiscale Analysis of Western Diet-Fed Mice Reveals a Translationally Relevant Sequence of Events during NAFLD Progression.

Mouse models of non-alcoholic fatty liver disease (NAFLD) are required to define therapeutic targets, but detailed time-resolved studies to establish a sequence of events are lacking. Here, we fed male C57Bl/6N mice a Western or standard diet over 48 weeks. Multiscale time-resolved characterization was performed using RNA-seq, histopathology, immunohistochemistry, intravital imaging, and blood chemistry; the results were compared to human disease. Acetaminophen toxicity and ammonia metabolism were additionally analyzed as functional readouts. We identified a sequence of eight key events: formation of lipid droplets; inflammatory foci; lipogranulomas; zonal reorganization; cell death and replacement proliferation; ductular reaction; fibrogenesis; and hepatocellular cancer. Functional changes included resistance to acetaminophen and altered nitrogen metabolism. The transcriptomic landscape was characterized by two large clusters of monotonously increasing or decreasing genes, and a smaller number of 'rest-and-jump genes' that initially remained unaltered but became differentially expressed only at week 12 or later. Approximately 30% of the genes altered in human NAFLD are also altered in the present mouse model and an increasing overlap with genes altered in human HCC occurred at weeks 30-48. In conclusion, the observed sequence of events recapitulates many features of human disease and offers a basis for the identification of therapeutic targets.

Data-driven gating in PET: Influence of respiratory signal noise on motion resolution.

PURPOSE: Data-driven gating (DDG) approaches for positron emission tomography (PET) are interesting alternatives to conventional hardware-based gating methods. In DDG, the measured PET data themselves are utilized to calculate a respiratory signal, that is, subsequently used for gating purposes. The success of gating is then highly dependent on the statistical quality of the PET data. In this study, we investigate how this quality determines signal noise and thus motion resolution in clinical PET scans using a center-of-mass-based (COM) DDG approach, specifically with regard to motion management of target structures in future radiotherapy planning applications. METHODS: PET list mode datasets acquired in one bed position of 19 different radiotherapy patients undergoing pretreatment [18 F]FDG PET/CT or [18 F]FDG PET/MRI were included into this retrospective study. All scans were performed over a region with organs (myocardium, kidneys) or tumor lesions of high tracer uptake and under free breathing. Aside from the original list mode data, datasets with progressively decreasing PET statistics were generated. From these, COM DDG signals were derived for subsequent amplitude-based gating of the original list mode file. The apparent respiratory shift d from end-expiration to end-inspiration was determined from the gated images and expressed as a function of signal-to-noise ratio SNR of the determined gating signals. This relation was tested against additional 25 [18 F]FDG PET/MRI list mode datasets where high-precision MR navigator-like respiratory signals were available as reference signal for respiratory gating of PET data, and data from a dedicated thorax phantom scan. RESULTS: All original 19 high-quality list mode datasets demonstrated the same behavior in terms of motion resolution when reducing the amount of list mode events for DDG signal generation. Ratios and directions of respiratory shifts between end-respiratory gates and the respective nongated image were constant over all statistic levels. Motion resolution d/dmax could be modeled as d/dmax=1-e-1.52(SNR-1)0.52, with dmax as the actual respiratory shift. Determining dmax from d and SNR in the 25 test datasets and the phantom scan demonstrated no significant differences to the MR navigator-derived shift values and the predefined shift, respectively. CONCLUSIONS: The SNR can serve as a general metric to assess the success of COM-based DDG, even in different scanners and patients. The derived formula for motion resolution can be used to estimate the actual motion extent reasonably well in cases of limited PET raw data statistics. This may be of interest for individualized radiotherapy treatment planning procedures of target structures subjected to respiratory motion.