Automated detection of critical findings in multi-parametric brain MRI using a system of 3D neural networks.

With the rapid growth and increasing use of brain MRI, there is an interest in automated image classification to aid human interpretation and improve workflow. We aimed to train a deep convolutional neural network and assess its performance in identifying abnormal brain MRIs and critical intracranial findings including acute infarction, acute hemorrhage and mass effect. A total of 13,215 clinical brain MRI studies were categorized to training (74%), validation (9%), internal testing (8%) and external testing (8%) datasets. Up to eight contrasts were included from each brain MRI and each image volume was reformatted to common resolution to accommodate for differences between scanners. Following reviewing the radiology reports, three neuroradiologists assigned each study to abnormal vs normal, and identified three critical findings including acute infarction, acute hemorrhage, and mass effect. A deep convolutional neural network was constructed by a combination of localization feature extraction (LFE) modules and global classifiers to identify the presence of 4 variables in brain MRIs including abnormal, acute infarction, acute hemorrhage and mass effect. Training, validation and testing sets were randomly defined on a patient basis. Training was performed on 9845 studies using balanced sampling to address class imbalance. Receiver operating characteristic (ROC) analysis was performed. The ROC analysis of our models for 1050 studies within our internal test data showed AUC/sensitivity/specificity of 0.91/83%/86% for normal versus abnormal brain MRI, 0.95/92%/88% for acute infarction, 0.90/89%/81% for acute hemorrhage, and 0.93/93%/85% for mass effect. For 1072 studies within our external test data, it showed AUC/sensitivity/specificity of 0.88/80%/80% for normal versus abnormal brain MRI, 0.97/90%/97% for acute infarction, 0.83/72%/88% for acute hemorrhage, and 0.87/79%/81% for mass effect. Our proposed deep convolutional network can accurately identify abnormal and critical intracranial findings on individual brain MRIs, while addressing the fact that some MR contrasts might not be available in individual studies.

Impact of COVID-19 social distancing regulations on outpatient diagnostic imaging volumes and no-show rates.

BACKGROUND: The coronavirus disease 2019 (COVID-19) pandemic has significantly impacted outpatient radiology practices, necessitating change in practice infrastructure and workflow. OBJECTIVE: The purpose of this study was to assess the consequences of social distancing regulations on 1) outpatient imaging volume and 2) no-show rates per imaging modality. METHODS: Volume and no-show rates of a large, multicenter metropolitan healthcare system outpatient practice were retrospectively stratified by modality including radiography, CT, MRI, ultrasonography, PET, DEXA, and mammography from January 2 to July 21, 2020. Trends were assessed relative to timepoints of significant state and local social distancing regulatory changes. RESULTS: The decline in imaging volume and rise in no-show rates was first noted on March 10, 2020 following the declaration of a state of emergency in New York State (NYS). Total outpatient imaging volume declined 85% from baseline over the following 5 days. Decreases varied by modality: 88% for radiography, 75% for CT, 73% for MR, 61% for PET, 80% for ultrasonography, 90% for DEXA, and 85% for mammography. Imaging volume and no-show rate recovery preceded the mask mandate of April 15, 2020, and further trended along with New York City's reopening phases. No-show rates recovered within 2 months of the height of the pandemic, however, outpatient imaging volume has yet to recover to baseline after 3 months. CONCLUSION: The total outpatient imaging volume declined alongside an increase in the no-show rate following the declaration of a state of emergency in NYS. No-show rates recovered within 2 months of the height of the pandemic with imaging volume yet to recover after 3 months. CLINICAL IMPACT: Understanding the impact of social distancing regulations on outpatient imaging volume and no-show rates can potentially aid other outpatient radiology practices and healthcare systems in anticipating upcoming changes as the COVID-19 pandemic evolves.

Sources of Revenue Loss and Recovery in Radiology Practices During the Coronavirus Disease 2019 (COVID-19) Pandemic.

RATIONALE AND OBJECTIVES: This study seeks to quantify the financial impact of COVID-19 on radiology departments, and to describe the structure of both volume and revenue recovery. MATERIALS AND METHODS: Radiology studies from a large academic health system were retrospectively studied from the first 33 weeks of 2020. Volume and work relative value unit (wRVU) data were aggregated on a weekly basis for three periods: Presurge (weeks 1-9), surge (10-19), and recovery (20-33), and analyzed compared to the pre-COVID baseline stratified by modality, specialty, patient service location, and facility type. Mean and median wRVU per study were used as a surrogate for case complexity. RESULTS: During the pandemic surge, case volumes fell 57%, while wRVUs fell by 69% relative to the pre-COVID-19 baseline. Mean wRVU per study was 1.13 in the presurge period, 1.03 during the surge, and 1.19 in the recovery. Categories with the greatest mean complexity declines were radiography (-14.7%), cardiothoracic imaging (-16.2%), and community hospitals overall (-15.9%). Breast imaging (+6.5%), interventional (+5.5%), and outpatient (+12.1%) complexity increased. During the recovery, significant increases in complexity were seen in cardiothoracic (0.46 to 0.49), abdominal (1.80 to 1.91), and neuroradiology (2.46 to 2.56) at stand-alone outpatient centers with similar changes at community hospitals. At academic hospitals, only breast imaging complexity remained elevated (1.32 from 1.17) during the recovery. CONCLUSION: Reliance on volume alone underestimates the financial impact of the COVID-19 pandemic as there was a disproportionate loss in high-RVU studies. However, increased complexity of outpatient cases has stabilized overall losses during the recovery.

Cytoarchitecture of the human cerebral cortex: MR microscopy of excised specimens at 9.4 Tesla.

BACKGROUND AND PURPOSE: The laminar patterns displayed by MR microscopy (MRM) form one basis for the classification of the cytoarchitectonic areas (Brodmann areas). It is plausible that in the future MRM may depict Brodmann areas directly, and not only by inference from gross anatomic location. Our purpose was to depict the laminar cytoarchitecture of excised, formalin-fixed specimens of human cerebral cortex by use of 9.4-T MR and to correlate MR images with histologic stains of the same sections. METHODS: Formalin-fixed samples of human sensory isocortex (calcarine, Heschl's, and somatosensory cortices), motor isocortex (hand motor area of M1), polar isocortex (frontal pole), allocortex (hippocampal formation), and transitional periallocortex (retrosplenial cortex) were studied by MRM at 9.4 T with intermediate-weighted pulse sequences for a total overnight acquisition time of 14 hours 17 minutes for each specimen. The same samples were then histologically analyzed to confirm the MR identification of the cortical layers. Curves representing the change in MR signal intensity across the cortex were generated to display the signal intensity profiles for each type of cortex. RESULTS: High-field-strength MR imaging at a spatial resolution of 78 x 78 x 500 micro m resolves the horizontal lamination of isocortex, allocortex, and periallocortex and displays specific intracortical structures such as the external band of Baillarger. The signal intensity profiles demonstrate the greatest hypointensity at the sites of maximum myelin concentration and maximum cell density and show gradations of signal intensity inversely proportional to varying cell density. CONCLUSION: MRM at 9.4 T depicts important aspects of the cytoarchitecture of normal formalin-fixed human cortex.