Faculty Bibliography

PURPOSE/OBJECTIVE:To introduce a method for the estimation of the ideal current patterns (ICP) that yield optimal signal-to-noise ratio (SNR) for realistic heterogeneous tissue models in MRI. THEORY AND METHODS/METHODS:The ICP were calculated for different surfaces that resembled typical radiofrequency (RF) coil formers. We constructed numerical electromagnetic (EM) bases to accurately represent EM fields generated by RF current sources located on the current-bearing surfaces. Using these fields as excitations, we solved the volume integral equation and computed the EM fields in the sample. The fields were appropriately weighted to calculate the optimal SNR and the corresponding ICP. We demonstrated how to qualitatively use ICP to guide the design of a coil array to maximize SNR inside a head model. RESULTS: CONCLUSION/CONCLUSIONS:ICP can be calculated for human tissue models, potentially guiding the design of application-specific RF coil arrays.

Different pathologies of the hip are characterized by the abnormal shape of the bony structures of the joint, namely the femur and the acetabulum. Three-dimensional (3D) models of the hip can be used for diagnosis, biomechanical simulation, and planning of surgical treatments. These models can be generated by building 3D surfaces of the joint's structures segmented on magnetic resonance (MR) images. Deep learning can avoid time-consuming manual segmentations, but its performance depends on the amount and quality of the available training data. Data augmentation and transfer learning are two approaches used when there is only a limited number of datasets. In particular, data augmentation can be used to artificially increase the size and diversity of the training datasets, whereas transfer learning can be used to build the desired model on top of a model previously trained with similar data. This study investigates the effect of data augmentation and transfer learning on the performance of deep learning for the automatic segmentation of the femur and acetabulum on 3D MR images of patients diagnosed with femoroacetabular impingement. Transfer learning was applied starting from a model trained for the segmentation of the bony structures of the shoulder joint, which bears some resemblance to the hip joint. Our results suggest that data augmentation is more effective than transfer learning, yielding a Dice similarity coefficient compared to ground-truth manual segmentations of 0.84 and 0.89 for the acetabulum and femur, respectively, whereas the Dice coefficient was 0.78 and 0.88 for the model based on transfer learning. The Accuracy for the two anatomical regions was 0.95 and 0.97 when using data augmentation, and 0.87 and 0.96 when using transfer learning. Data augmentation can improve the performance of deep learning models by increasing the diversity of the training dataset and making the models more robust to noise and variations in image quality. The proposed segmentation model could be combined with radiomic analysis for the automatic evaluation of hip pathologies.

The increasing availability of high-performance gradient systems in human MRI scanners has generated great interest in diffusion microstructural imaging applications such as axonal diameter mapping. Practically, sensitivity to axon diameter in diffusion MRI is attained at strong diffusion weightings (Formula presented.), where the deviation from the expected (Formula presented.) scaling in white matter yields a finite transverse diffusivity, which is then translated into an axon diameter estimate. While axons are usually modeled as perfectly straight, impermeable cylinders, local variations in diameter (caliber variation or beading) and direction (undulation) are known to influence axonal diameter estimates and have been observed in microscopy data of human axons. In this study, we performed Monte Carlo simulations of diffusion in axons reconstructed from three-dimensional electron microscopy of a human temporal lobe specimen using simulated sequence parameters matched to the maximal gradient strength of the next-generation Connectome 2.0 human MRI scanner ((Formula presented.) 500 mT/m). We show that axon diameter estimation is accurate for nonbeaded, nonundulating fibers; however, in fibers with caliber variations and undulations, the axon diameter is heavily underestimated due to caliber variations, and this effect overshadows the known overestimation of the axon diameter due to undulations. This unexpected underestimation may originate from variations in the coarse-grained axial diffusivity due to caliber variations. Given that increased axonal beading and undulations have been observed in pathological tissues, such as traumatic brain injury and ischemia, the interpretation of axon diameter alterations in pathology may be significantly confounded.

Proper animal conditioning is a key factor in the quality and success of preclinical neuroimaging applications. Here, we introduce an open-source easy-to-modify multimodal 3D printable design for rodent conditioning for magnetic resonance imaging (MRI) or other imaging modalities. Our design can be used for brain imaging in anesthetized or awake mice, and in anesthetized rats. We show ease of use and reproducibility of subject conditioning with anatomical T2-weighted imaging for both mice and rats. We also demonstrate the application of our design for awake functional MRI in mice using both visual evoked potential and olfactory stimulation paradigms. In addition, using a combined MRI, positron emission tomography and X-ray computed tomography experiment, we demonstrate that our proposed cradle design can be utilized for multiple imaging modalities.

We propose Physics-Informed Fourier Networks for Electrical Properties (EP) Tomography (PIFON-EPT), a novel deep learning-based method for EP reconstruction using noisy and/or incomplete magnetic resonance (MR) measurements. Our approach leverages the Helmholtz equation to constrain two networks, responsible for the denoising and completion of the transmit fields, and the estimation of the object's EP, respectively. We embed a random Fourier features mapping into our networks to enable efficient learning of high-frequency details encoded in the transmit fields. We demonstrated the efficacy of PIFON-EPT through several simulated experiments at 3 and 7 T (T) MR imaging, and showed that our method can reconstruct physically consistent EP and transmit fields. Specifically, when only 20% of the noisy measured fields were used as inputs, PIFON-EPT reconstructed the EP of a phantom with ≤ 5% error, and denoised and completed the measurements with ≤ 1% error. Additionally, we adapted PIFON-EPT to solve the generalized Helmholtz equation that accounts for gradients of EP between inhomogeneities. This yielded improved results at interfaces between different materials without explicit knowledge of boundary conditions. PIFON-EPT is the first method that can simultaneously reconstruct EP and transmit fields from incomplete noisy MR measurements, providing new opportunities for EPT research.

3D imaging enables accurate diagnosis by providing spatial information about organ anatomy. However, using 3D images to train AI models is computationally challenging because they consist of 10x or 100x more pixels than their 2D counterparts. To be trained with high-resolution 3D images, convolutional neural networks resort to downsampling them or projecting them to 2D. We propose an effective alternative, a neural network that enables efficient classification of full-resolution 3D medical images. Compared to off-the-shelf convolutional neural networks, our network, 3D Globally-Aware Multiple Instance Classifier (3D-GMIC), uses 77.98%-90.05% less GPU memory and 91.23%-96.02% less computation. While it is trained only with image-level labels, without segmentation labels, it explains its predictions by providing pixel-level saliency maps. On a dataset collected at NYU Langone Health, including 85,526 patients with full-field 2D mammography (FFDM), synthetic 2D mammography, and 3D mammography, 3D-GMIC achieves an AUC of 0.831 (95% CI: 0.769-0.887) in classifying breasts with malignant findings using 3D mammography. This is comparable to the performance of GMIC on FFDM (0.816, 95% CI: 0.737-0.878) and synthetic 2D (0.826, 95% CI: 0.754-0.884), which demonstrates that 3D-GMIC successfully classified large 3D images despite focusing computation on a smaller percentage of its input compared to GMIC. Therefore, 3D-GMIC identifies and utilizes extremely small regions of interest from 3D images consisting of hundreds of millions of pixels, dramatically reducing associated computational challenges. 3D-GMIC generalizes well to BCS-DBT, an external dataset from Duke University Hospital, achieving an AUC of 0.848 (95% CI: 0.798-0.896).

While musculoskeletal imaging volumes are increasing, there is a relative shortage of subspecialized musculoskeletal radiologists to interpret the studies. Will artificial intelligence (AI) be the solution? For AI to be the solution, the wide implementation of AI-supported data acquisition methods in clinical practice requires establishing trusted and reliable results. This implementation will demand close collaboration between core AI researchers and clinical radiologists. Upon successful clinical implementation, a wide variety of AI-based tools can improve the musculoskeletal radiologist's workflow by triaging imaging examinations, helping with image interpretation, and decreasing the reporting time. Additional AI applications may also be helpful for business, education, and research purposes if successfully integrated into the daily practice of musculoskeletal radiology. The question is not whether AI will replace radiologists, but rather how musculoskeletal radiologists can take advantage of AI to enhance their expert capabilities.

Histopathological studies suggest that cerebral small vessel tortuosity is crucial in age-related blood flow reduction and cellular degeneration. However, in vivo evidence is lacking. Here, we used Ferumoxytol-enhanced 7T MRI to directly visualize cerebral small vessels (>300 µm), enabling the identification of vascular tortuosity and exploration of its links to age, tissue atrophy, and vascular risk factors. High-resolution 2D/3D gradient echo MRI at 7T enhanced with Ferumoxytol, an ultrasmall superparamagnetic iron oxide (USPIO), was obtained and analyzed for cerebral small medullary artery tortuosity from 37 healthy participants (21-70 years; mean/SD: 38±14 years; 19 females). Tortuous artery count and tortuosity indices were compared between young and old groups. Age effects on vascular tortuosity were examined through partial correlations and multiple linear regression, adjusting for sex, body mass index (BMI), blood pressure (BP), and other vascular risk factors. Associations between tortuous medullary arteries and tissue atrophy, perivascular spaces (PVS), and white matter (WM) hyperintensities were explored. Age and BMI, rather than BP, showed positive correlations with both tortuous artery count and tortuosity indices. A significant correlation existed between the number of tortuous arteries and WM atrophy. WM lesions were found in proximity to or at the distal ends of tortuous medullary arteries, especially within the deep WM. Moreover, the elderly population displayed a higher prevalence of PVS, including those containing enclosed tortuous arteries. Leveraging the blooming effect of Ferumoxytol, 7T MRI excels in directly detecting cerebral small arterial tortuosity in vivo, unveiling its associations with age, BMI, tissue atrophy, WMH and PVS.

Creating a patient-centered experience is becoming increasingly important for radiology departments around the world. The goal of patient-centered radiology is to ensure that radiology services are sensitive to patients' needs and desires. This article provides a framework for addressing the patient's experience by dividing their imaging journey into three distinct time periods: pre-exam, day of exam, and post-exam. Each time period has aspects that can contribute to patient anxiety. Although there are components of the patient journey that are common in all regions of the world, there are also unique features that vary by location. This paper highlights innovative solutions from different parts of the world that have been introduced in each of these time periods to create a more patient-centered experience. CLINICAL RELEVANCE STATEMENT: Adopting innovative solutions that help patients understand their imaging journey and decrease their anxiety about undergoing an imaging examination are important steps in creating a patient centered imaging experience. KEY POINTS: • Patients often experience anxiety during their imaging journey and decreasing this anxiety is an important component of patient centered imaging. • The patient imaging journey can be divided into three distinct time periods: pre-exam, day of exam, and post-exam. • Although components of the imaging journey are common, there are local differences in different regions of the world that need to be considered when constructing a patient centered experience.

PURPOSE/OBJECTIVE:To assess the diagnostic performance of FDG-PET/MRI for the preoperative diagnosis and staging of peritoneal carcinomatosis (PC) using surgical Sugarbaker's PC index (PCI) as the reference in a multireader pilot study. METHODS:Fourteen adult patients (M/F: 3/11, mean age: 57 ± 12 year) with PC were prospectively included in this single-center study. Patients underwent FDG-PET/MRI prior to surgery (mean delay: 14 d, range: 1-63 d). Images were reviewed independently by 2 abdominal radiologists and 2 nuclear medicine physicians. The radiologists assessed contrast-enhanced abdominal MR images, while the nuclear medicine physicians assessed PET images fused with T2-weighted images. The abdomen was divided in 13 regions, scored from 0 to 3. A hybrid FDG-PET/MRI radiological PCI was created by combining the study data. Radiological PCI was compared to the surgical PCI on a per-patient and per-region basis. Inter-reader agreement was evaluated. RESULTS:Mean surgical PCI was 10 ± 8 (range: 0-24). Inter-reader agreement was almost perfect for all sets for radiologic PCI (Kappa: 0.81-0.98). PCI scores for all reading sets significantly correlated with the surgical PCI score (r range: 0.57-0.74, p range: < 0.001-0.003). Pooled per-patient sensitivity, specificity, and accuracy were 75%/50%/71.4% for MRI, 66.7%/50%/64.3% for FDG-PET, and 91.7%/50%/85.7% for FDG-PET/MRI, without significant difference (p value range 0.13-1). FDG-PET/MRI achieved 100% sensitivity and 100% specificity for a cutoff PCI of 20. Per-region sensitivity and accuracy were lower: 37%/61.8% for MRI, 17.8%/64.3% for FDG-PET, and 52.7%/60.4% for FDG-PET/MRI, with significantly higher sensitivity for FDG-PET/MRI. Per-region specificity was higher for FDG-PET (95%) compared to MRI (78.4%) and FDG-PET/MRI (66.5%). CONCLUSION/CONCLUSIONS:FDG-PET/MRI achieved an excellent diagnostic accuracy per-patient and weaker performance per-region for detection of PC. The added value of PET/MRI compared to MRI and FDG-PET remains to be determined.