Faculty Bibliography

PURPOSE/OBJECTIVE:To develop and evaluate a neural network-based method for Gibbs artifact and noise removal. METHODS:A convolutional neural network (CNN) was designed for artifact removal in diffusion-weighted imaging data. Two implementations were considered: one for magnitude images and one for complex images. Both models were based on the same encoder-decoder structure and were trained by simulating MRI acquisitions on synthetic non-MRI images. RESULTS:Both machine learning methods were able to mitigate artifacts in diffusion-weighted images and diffusion parameter maps. The CNN for complex images was also able to reduce artifacts in partial Fourier acquisitions. CONCLUSIONS:The proposed CNNs extend the ability of artifact correction in diffusion MRI. The machine learning method described here can be applied on each imaging slice independently, allowing it to be used flexibly in clinical applications.

PURPOSE/OBJECTIVE:To advance research in the field of machine learning for MR image reconstruction with an open challenge. METHODS:We provided participants with a dataset of raw k-space data from 1,594 consecutive clinical exams of the knee. The goal of the challenge was to reconstruct images from these data. In order to strike a balance between realistic data and a shallow learning curve for those not already familiar with MR image reconstruction, we ran multiple tracks for multi-coil and single-coil data. We performed a two-stage evaluation based on quantitative image metrics followed by evaluation by a panel of radiologists. The challenge ran from June to December of 2019. RESULTS:We received a total of 33 challenge submissions. All participants chose to submit results from supervised machine learning approaches. CONCLUSIONS:The challenge led to new developments in machine learning for image reconstruction, provided insight into the current state of the art in the field, and highlighted remaining hurdles for clinical adoption.

PURPOSE/OBJECTIVE:This study reports on the development of a novel 3D procedure planning technique to provide pre-ablation treatment planning for partial gland prostate cryoablation (cPGA). METHODS:Twenty men scheduled for partial gland cryoablation (cPGA) underwent pre-operative image segmentation and 3D modeling of the prostatic capsule, index lesion, urethra, rectum, and neurovascular bundles based upon multi-parametric MRI data. Pre-treatment 3D planning models were designed including virtual 3D cryotherapy probes to predict and plan cryotherapy probe configuration needed to achieve confluent treatment volume. Treatment efficacy was measured with 6 month post-operative MRI, serum prostate specific antigen (PSA) at 3 and 6 months, and treatment zone biopsy results at 6 months. Outcomes from 3D planning were compared to outcomes from a series of 20 patients undergoing cPGA using traditional 2D planning techniques. RESULTS:Forty men underwent cPGA. The median age of the cohort undergoing 3D treatment planning was 64.8 years with a median pretreatment PSA of 6.97 ng/mL. The Gleason grade group (GGG) of treated index lesions in this cohort included 1 (5%) GGG1, 11 (55%) GGG2, 7 (35%) GGG3, and 1 (5%) GGG4. Two (10%) of these treatments were post-radiation salvage therapies. The 2D treatment cohort included 20 men with a median age of 68.5 yrs., median pretreatment PSA of 6.76 ng/mL. The Gleason grade group (GGG) of treated index lesions in this cohort included 3 (15%) GGG1, 8 (40%) GGG2, 8 (40%) GGG3, 1 (5%) GGG4. Two (10%) of these treatments were post-radiation salvage therapies. 3D planning predicted the same number of cryoprobes for each group, however a greater number of cryoprobes was used in the procedure for the prospective 3D group as compared to that with 2D planning (4.10 ± 1.37 and 3.25 ± 0.44 respectively, p = 0.01). At 6 months post cPGA, the median PSA was 1.68 ng/mL and 2.38 ng/mL in the 3D and 2D cohorts respectively, with a larger decrease noted in the 3D cohort (75.9% reduction noted in 3D cohort and 64.8% reduction 2D cohort, p 0.48). In-field disease detection was 1/14 (7.1%) on surveillance biopsy in the 3D cohort and 3/14 (21.4%) in the 2D cohort, p = 0.056) In the 3D cohort, 6 month biopsy was not performed in 4 patients (20%) due to undetectable PSA, negative MRI, and negative MRI Axumin PET. For the group with traditional 2D planning, treatment zone biopsy was positive in 3/14 (21.4%) of the patients, p = 0.056. CONCLUSIONS:3D prostate cancer models derived from mpMRI data provide novel guidance for planning confluent treatment volumes for cPGA and predicted a greater number of treatment probes than traditional 2D planning methods. This study prompts further investigation into the use of 3D treatment planning techniques as the increase of partial gland ablation treatment protocols develop.

OBJECTIVE:Deep Learning (DL) image reconstruction has the potential to disrupt the current state of MR imaging by significantly decreasing the time required for MR exams. Our goal was to use DL to accelerate MR imaging in order to allow a 5-minute comprehensive examination of the knee, without compromising image quality or diagnostic accuracy. METHODS:A DL model for image reconstruction using a variational network was optimized. The model was trained using dedicated multi-sequence training, in which a single reconstruction model was trained with data from multiple sequences with different contrast and orientations. Following training, data from 108 patients were retrospectively undersampled in a manner that would correspond with a net 3.49-fold acceleration of fully-sampled data acquisition and 1.88-fold acceleration compared to our standard two-fold accelerated parallel acquisition. An interchangeability study was performed, in which the ability of 6 readers to detect internal derangement of the knee was compared for the clinical and DL-accelerated images. RESULTS:The study demonstrated a high degree of interchangeability between standard and DL-accelerated images. In particular, results showed that interchanging the sequences would result in discordant clinical opinions no more than 4% of the time for any feature evaluated. Moreover, the accelerated sequence was judged by all six readers to have better quality than the clinical sequence. CONCLUSIONS:An optimized DL model allowed for acceleration of knee images which performed interchangeably with standard images for the detection of internal derangement of the knee. Importantly, readers preferred the quality of accelerated images to that of standard clinical images.

The current coronavirus disease 2019 (COVID-19) crisis continues to grow and has resulted in marked changes to clinical operations. In parallel with clinical preparedness, universities have shut down most scientific research activities. Radiology researchers are currently grappling with these challenges that will continue to affect current and future imaging research. The purpose of this article is to describe the collective experiences of a diverse international group of academic radiology research programs in managing their response to the COVID-19 pandemic. The acute response at six distinct institutions will be described first, exploring common themes, challenges, priorities, and practices. This will be followed by reflections about the future of radiology research in the wake of the COVID-19 pandemic.

PURPOSE/OBJECTIVE:To study the effects of magnetization transfer (MT, in which a semi-solid spin pool interacts with the free pool), in the context of magnetic resonance fingerprinting (MRF). METHODS: RESULTS:values (~47 ms vs. ~35 ms) can be observed in white matter if MT is accounted for. CONCLUSION/CONCLUSIONS:with MRF. A model that encompasses MT effects can improve the accuracy of estimated relaxation parameters and allows quantification of the fractional pool size.

PURPOSE/OBJECTIVE:, and proton density) and sodium density images in 1 single scan. We hope that the development of such capabilities will help to ease the implementation of sodium MRI in clinical trials and provide more opportunities for researchers to investigate metabolism through sodium MRI. METHODS: RESULTS: CONCLUSIONS:

A publicly available dataset containing k-space data as well as Digital Imaging and Communications in Medicine image data of knee images for accelerated MR image reconstruction using machine learning is presented.