Deep neural network to locate and segment brain tumors outperformed the expert technicians who created the training data.

Purpose: Deep learning (DL) algorithms have shown promising results for brain tumor segmentation in MRI. However, validation is required prior to routine clinical use. We report the first randomized and blinded comparison of DL and trained technician segmentations. Approach: We compiled a multi-institutional database of 741 pretreatment MRI exams. Each contained a postcontrast T1-weighted exam, a T2-weighted fluid-attenuated inversion recovery exam, and at least one technician-derived tumor segmentation. The database included 729 unique patients (470 males and 259 females). Of these exams, 641 were used for training the DL system, and 100 were reserved for testing. We developed a platform to enable qualitative, blinded, controlled assessment of lesion segmentations made by technicians and the DL method. On this platform, 20 neuroradiologists performed 400 side-by-side comparisons of segmentations on 100 test cases. They scored each segmentation between 0 (poor) and 10 (perfect). Agreement between segmentations from technicians and the DL method was also evaluated quantitatively using the Dice coefficient, which produces values between 0 (no overlap) and 1 (perfect overlap). Results: The neuroradiologists gave technician and DL segmentations mean scores of 6.97 and 7.31, respectively ( p < 0.00007 ). The DL method achieved a mean Dice coefficient of 0.87 on the test cases. Conclusions: This was the first objective comparison of automated and human segmentation using a blinded controlled assessment study. Our DL system learned to outperform its "human teachers" and produced output that was better, on average, than its training data.

MR Imaging and Cochlear Implants with Retained Internal Magnets: Reducing Artifacts near Highly Inhomogeneous Magnetic Fields.

The number of patients receiving cochlear implants and auditory brainstem implants for severe to profound sensorineural hearing loss has rapidly increased. These implants consist of an internal component implanted between the skull and the temporal scalp and an external removable speech processor unit. A small magnet within the internal component is commonly used to hold the external speech processor unit in place. Several cochlear implant models have recently received U.S. Food and Drug Administration and European Economic Area regulatory approval to allow magnetic resonance (MR) imaging examinations to be performed under certain specified conditions. The small internal magnet presents a challenge for imaging of the head and neck near the implant, creating a nonlinear magnetic field inhomogeneity and significant MR imaging artifacts. Fat-saturation failures and susceptibility artifacts severely degrade image quality. Typical artifacts at diffusion-weighted imaging and accelerated imaging are exacerbated. Each examination may require impromptu adjustments to allow visualization of the tissue or contrast of interest. Patients may also be quite uncomfortable during the examination, as a result of either imposed magnetic forces or a tight head wrap that is often applied to minimize internal magnet movement. Translational forces and torque sometimes displace the implanted magnet even when a head wrap is used. Diseases such as neurofibromatosis type 2 that are associated with bilateral vestibular schwannomas and hearing loss often require lifelong tumor surveillance with MR imaging. A collaborative team of radiologists, technologists, and/or medical physicists or MR imaging scientists, armed with strategies to mitigate artifacts near implanted magnets, can customize the examination for better visualization of tissue and consistent comparison examinations over time. ©RSNA, 2018.

Double Inversion Recovery Magnetic Resonance Imaging in Identifying Focal Cortical Dysplasia.

BACKGROUND: Focal cortical dysplasia is commonly recognized in pediatric epilepsy surgery. Despite characteristic radiographic features, focal cortical dysplasia can be subtle on magnetic resonance imaging. Double inversion recovery acquisition suppresses the white matter signal, which may enhance visualization of abnormal features at the gray-white matter interface. We assessed the ability of double inversion recovery to distinguish focal cortical dysplasia from periventricular nodular heterotopia and normal brain. METHODS: Patients with focal cortical dysplasia were identified from our patient database, as was a control group comprising patients with periventricular nodular heterotopia and healthy persons. A senior neuroradiologist reviewed all clinical images and classified them as patients with focal cortical dysplasia (n = 16) or control subjects (periventricular nodular heterotopia, n = 13; normal, n = 20). Four neuroradiologists reviewed the de-identified and randomized double inversion recovery and magnetization prepared rapid acquired gradient echoes (MPRAGE) sequences for each person and scored them as normal, focal cortical dysplasia, or periventricular nodular heterotopia. RESULTS: Among individual reviewers, double inversion recovery showed sensitivity from 50% to 88% and specificity from 67% to 91% in detecting focal cortical dysplasia. The sensitivity was notably higher in reviewers with more clinical experience with the technique. Consensus agreement among the three most experienced reviewers gave a sensitivity of 88% (95% confidence interval [CI], 72% to 97%) and specificity of 88% (95% CI, 62% to 98%) for double inversion recovery and sensitivity of 44% (95% CI, 20% to 70%) and specificity of 100% (95% CI, 89% to 100%) for MPRAGE. CONCLUSIONS: Double inversion recovery is sensitive for detection of focal cortical dysplasia with experienced users, particularly when there is consensus agreement. The use of two clinically available magnetic resonance imaging acquisitions-double inversion recovery and another sequence with high specificity such as MPRAGE-would be complementary in the evaluation of lesional epilepsy.

Congenital mydriasis and prune belly syndrome in a child with an ACTA2 mutation.

We report the association of congenital mydriasis with prune belly syndrome and cerebrovascular anomalies in a 9-year-old boy who was found to have an ACTA2 mutation. This case illustrates the spectrum of systemic malformations that are attributable to mutations in ACTA2 and expands the spectrum of cerebrovascular anomalies that are now known to accompany congenital mydriasis.

"Ocular moyamoya" syndrome in a patient with features of microcephalic osteodysplastic primordial dwarfism type II.

Primordial dwarfism refers to severely impaired growth beginning early in fetal life. There are many genetic causes of primordial dwarfism, including disorders classified as microcephalic osteodysplastic primordial dwarfism. Microcephalic osteodysplastic primordial dwarfism type II is an autosomal-recessive disease characterized by small stature, bone and dental anomalies, and characteristic facies. Affected patients have a high risk of stroke secondary to progressive cerebral vascular anomalies, which often are classified as moyamoya disease. We present the case of a boy with features suggestive of MOPD II with unilateral moyamoya cerebrovascular changes and correlative moyamoya collaterals involving the iris of the ipsilateral eye.

Assessment of normal myelination with magnetic resonance imaging.

White matter myelination is essential to postnatal neurologic maturation and can be accurately evaluated by magnetic resonance imaging (MRI). Accordingly, MRI pulse sequences should be optimized for detection of myelin in young children. T1-weighted images are most useful during the first year of life. These demonstrate myelin-related white matter hyperintensity consequent to increasing cholesterol and galactocerebroside within myelin membranes. T2-weighted images are most useful in later stages of myelination, during which time elaboration of myelin leads to reduction in brain water content with associated T2 hypointensity. Additional information regarding the status of myelination can be obtained from T2-weighted fluid attenuation inversion recovery (FLAIR) and diffusion tensor imaging (DTI) pulse sequences. Clinically useful milestones for assessment of myelination across all these MRI pulse sequences are available as guidelines to image interpretation. Evaluation of myelination status using a combination of T1- and T2-weighted images should be routine in the interpretation of all pediatric brain MRI exams.

3 Tesla magnetic resonance imaging with and without corticotropin releasing hormone stimulation for the detection of microadenomas in Cushing's syndrome.

OBJECTIVE: We sought to determine if higher resolution 3 Tesla (T) magnetic resonance imaging (MRI) with or without ovine corticotropin releasing hormone (o-CRH) stimulation would increase the sensitivity for detection of pituitary microadenomas in ACTH-dependent Cushing's syndrome (CS). DESIGN AND PATIENTS: We prospectively identified 23 patients over a 2-year period with clinical and biochemical evidence of ACTH-dependent CS with no lesion (n = 11) or equivocal lesion (n = 10) on 1.5T MRI. Subsequently, two additional MRIs were performed in random order: 3T nonstimulated MRI or 3T MRI with o-CRH in all patients. Three neuroradiologists reviewed all examinations in a randomized blinded fashion. Patients were divided into four groups, depending on the outcome of their evaluation and treatment for CS. Two patients had to be excluded, and so we report on 21 subjects. MEASUREMENTS AND RESULTS: Both 3T MRI without (P < 0.016) and with o-CRH stimulation (P < 0.013) was significantly more sensitive for detection of pituitary microadenomas than 1.5T MRI for Group 1 (definitive proof of Cushing's disease, n = 10). Group 2 (those in group 1, plus three patients where dynamic/invasive testing suggested pituitary source) also showed a significant (P < 0.012) advantage for 3T. There was no difference between the 3T and the 3T o-CRH examinations for any of the pulse sequences. We did not observe a statistically significant difference in other patient groups [patients with recurrent CD (n = 6) and patients with ectopic CS (n = 2)]. CONCLUSIONS: The results of our prospective blinded studies suggest that 3T MRI of pituitary gland should be considered in evaluation of patients with ACTH-dependent CD when 1.5T imaging is negative or equivocal.

Neonatal hypoxic-ischemic encephalopathy: multimodality imaging findings.

Diffuse hypoxic-ischemic brain injury in the neonate results in neonatal hypoxic-ischemic encephalopathy (HIE). Because of differences in brain maturity at time of insult, severity of hypotension, and duration of insult, there are four distinct patterns of brain injury. Cranial ultra-sonography and computed tomography reveal periventricular leukomalacia, germinal matrix hemorrhage, and hydrocephalus. Magnetic resonance imaging is the most sensitive modality for evaluating the patterns of brain injury. In preterm neonates, mild hypotension causes periventricular injury; severe hypotension results in infarction of the deep gray matter, brainstem, and cerebellum. In term neonates, mild hypotension causes parasagittal cortical and subcortical injury; severe hypotension causes characteristic injury of the lateral thalami, posterior putamina, hippocampi, corticospinal tracts, and sensorimotor cortex. Prompt recognition of these imaging findings can help exclude other causes of encephalopathy, affect prognosis, and facilitate earlier (although mostly supportive) treatment.

Different patterns of magnetic resonance imaging atrophy for frontotemporal lobar degeneration syndromes.

BACKGROUND: Frontotemporal lobar degeneration (FTLD) is an uncommon degenerative dementia that presents with focal cognitive and behavioral deficits. OBJECTIVE: To determine the correlation of the different presentations of FTLD with structural neuroimaging findings. DESIGN AND PATIENTS: In a blinded study, we retrospectively evaluated the clinical presentations and magnetic resonance imaging (MRI) patterns of atrophy in 59 patients with FTLD and 26 patients with probable Alzheimer disease at a memory disorders clinic. RESULTS: Analysis of variance revealed a significant difference in the patterns of atrophy in the FTLD and Alzheimer disease groups. Patients with FTLD presenting with altered personal conduct had significant bifrontal atrophy, whereas patients presenting with semantic dementia had significant left temporal and bifrontal atrophy compared with other groups. Disinhibited behavior and hyperphagia correlated with right frontal atrophy, and fluent, anomic aphasia correlated with left temporal atrophy. CONCLUSIONS: We found that the type of clinical presentation of FTLD correlates with specific areas of atrophy. Our method of analysis may be useful to elicit further anatomic-behavioral relationships in degenerative brain disorders.

Frequency of abnormal hand and wrist radiographs at time of diagnosis of polyarticular juvenile rheumatoid arthritis.

OBJECTIVE: To determine the frequency of radiographic abnormalities in hand/wrist radiographs of children with newly diagnosed polyarticular juvenile rheumatoid arthritis (polyJRA) because radiographs of small joints are an important tool in assessing outcomes in RA and there are clinical similarities between RA and polyJRA. METHODS: A medical record review was performed to identify cases of polyJRA seen at Mayo Clinic from January 1, 1994, to December 31, 2001. Hand/wrist radiographs, obtained at the time of diagnosis, were reviewed by 3 radiologists with attention to periarticular osteopenia, joint space narrowing (JSN), or erosion. At least 2 radiologists had to independently identify abnormal findings on the same radiograph. The relative carpal length (RCL), judged by Poznanski's method, was also determined. RESULTS: From the review of 159 medical records, 60 cases of newly diagnosed polyJRA were identified. Twenty-five of these had hand/wrist radiographs at diagnosis; 18 sets were available for this study. Of those, 2/3 were female, 6% (1/18) had subcutaneous nodules, 7% (1/14) had elevated levels of serum rheumatoid factor, and 44% (7/16) had elevated serum levels of antinuclear antibodies. Median age at diagnosis was 10.2 years, median duration of hand/wrist symptoms at diagnosis was 10 months, and median number of joints with either swelling, pain on range of motion (ROM), or limited ROM was 14.5. Sixty-one percent of radiographs taken at the time of diagnosis of polyJRA were abnormal. While 44% had periarticular osteopenia, 28% had either erosions or JSN. Six (33%) had RCL > 2 SD below the mean for age. Five (83%) of those with RCL, > 2 SD below the mean for age, had periarticular osteopenia, JSN, or erosion. CONCLUSION: We conclude the frequency of abnormal hand/wrist radiographs is very high very early in the course of polyJRA. More studies are needed to determine to what extent these radiographic abnormalities correlate with clinical outcomes.