Deep learning to detect left ventricular structural abnormalities in chest X-rays.

BACKGROUND AND AIMS: Early identification of cardiac structural abnormalities indicative of heart failure is crucial to improving patient outcomes. Chest X-rays (CXRs) are routinely conducted on a broad population of patients, presenting an opportunity to build scalable screening tools for structural abnormalities indicative of Stage B or worse heart failure with deep learning methods. In this study, a model was developed to identify severe left ventricular hypertrophy (SLVH) and dilated left ventricle (DLV) using CXRs. METHODS: A total of 71 589 unique CXRs from 24 689 different patients completed within 1 year of echocardiograms were identified. Labels for SLVH, DLV, and a composite label indicating the presence of either were extracted from echocardiograms. A deep learning model was developed and evaluated using area under the receiver operating characteristic curve (AUROC). Performance was additionally validated on 8003 CXRs from an external site and compared against visual assessment by 15 board-certified radiologists. RESULTS: The model yielded an AUROC of 0.79 (0.76-0.81) for SLVH, 0.80 (0.77-0.84) for DLV, and 0.80 (0.78-0.83) for the composite label, with similar performance on an external data set. The model outperformed all 15 individual radiologists for predicting the composite label and achieved a sensitivity of 71% vs. 66% against the consensus vote across all radiologists at a fixed specificity of 73%. CONCLUSIONS: Deep learning analysis of CXRs can accurately detect the presence of certain structural abnormalities and may be useful in early identification of patients with LV hypertrophy and dilation. As a resource to promote further innovation, 71 589 CXRs with adjoining echocardiographic labels have been made publicly available.

A comparison of pulmonary embolism in pediatric and adult patients with acute COVID-19.

BACKGROUND: COVID-19 is associated with pulmonary embolism (PE) in adults. However, the rate of PE in pediatric patients with acute COVID-19 evaluated by CT pulmonary angiography (CTPA) has not been evaluated. OBJECTIVE: Determine PE rate in pediatric patients with acute COVID-19 and compare to adults. MATERIALS AND METHODS: A retrospective review of CTPA studies, performed between March 2020 and January 2021 on pediatric patients with acute COVID-19, but not MIS-C, was performed. CTPAs performed on an adult cohort of acute COVID-19 patients during April 2020 were reviewed for comparison. Pediatric and chest radiologists independently reviewed CTPAs of pediatric and adult patients, respectively. RESULTS: Of the 355 acute COVID-19 pediatric patients treated during the study period, 14 (16.6 ± 4.8y, median-18.5y, 64% female) underwent CTPA. Of the 1868 acute COVID-19 adults treated during two weeks in April 2020, 50 (57.2 ± 17.0y, median-57.0y, 42% female) underwent CTPA. The PE rate was 14% in the pediatric group (2 patients) and 18% in the adult group (9 patients) (p = 1.0). Both pediatric patients with PE were obese, over 18y, and had asthma, diabetes mellitus, or hypertension. No child<18y with acute COVID-19 had PE. In the adult cohort, higher alanine-aminotransferase and D-dimer levels were associated with PE (p = 0.04 and p = 0.004, respectively). CONCLUSION: Despite similar PE rates in pediatric and adult patients, PE occurred in acute COVID-19 pediatric patients who were >18y, obese, and had at least 1 comorbidity. Children <18y with COVID-19 did not have PE.

Bladder prolapse through a patent urachus presenting as an umbilical mass in the newborn: characteristic prenatal sonographic findings and the diagnostic benefit of postnatal cystography.

Bladder prolapse through a patent urachus is rare. We present a newborn with an unusual exophytic, erythematous umbilical mass. Voiding cystography readily demonstrated continuity of the bladder dome with the umbilical mass, confirming bladder prolapse through a patent urachus. The diagnosis of bladder prolapse was rapidly made in a second newborn with similar imaging and clinical findings and confirmed by cystography. We discuss the embryology of this condition including the association with a vesico-allantoic cyst in utero. Pre- and postnatal images are presented. The use of cystography in diagnosis is emphasized.

The challenging task of enumerating blasts in the bone marrow.

Enumeration of blasts in the bone marrow is critical for diagnostic, prognostic, and therapeutic response evaluation in myelodysplastic syndromes, myeloproliferative neoplasms and acute leukemias. However, few studies have examined the accuracy and precision of marrow blast counting using standard microscopic procedures. In our study, 4 experienced hematopathologists evaluated blast percentages in marrow using either differential counts on aspirate smears or visual estimates on CD34-stained trephine biopsies. Results of an independent observer's manual counts of individual labeled and unlabeled cells performed on high resolution digital images of CD34-stained trephine biopsies were designated as the "Digital Reference." Hematopathologists' blast counts showed excellent interobserver reproducibility, but the counts in smears and trephine biopsies correlated poorly with each other. Compared to the Digital Reference, both smear and trephine evaluations showed positive bias and high variability. The biopsy showed less variability but higher positive bias relative to the smears, indicating that counts were overestimated more in the hematopathologists' biopsy evaluation. Flow cytometric counts correlated well with the Digital Reference, and cases with high blast count generally showed worse cytogenetic findings. Our results demonstrate the need for better counting methods if significant decisions are made based on microscopic enumeration of blasts. Further efforts should be made to develop markers to better define blast cells and perhaps incorporate automated digital imaging technologies to enumerate them. Also, consideration should be given to quantifying blasts per marrow area in biopsies instead of per nucleated cells.

Metatarsalgia.

Although causes of metatarsalgia are multifactorial, in practice these most commonly include osseous stress reaction or fracture (including subchondral injury) and interdigital neuroma or plantar plate tear with adjacent pseudoneuroma. The various roles of radiography, ultrasonography, and MR imaging are discussed, and relevant technical issues and imaging findings are reviewed in order to facilitate accurate diagnosis and guide proper treatment of metatarsalgia.

Mini-review of hair cortisol concentration for evaluation of Cushing syndrome.

INTRODUCTION: The diagnosis of endogenous Cushing syndrome is often challenging and requires multiple repeated blood, urine, and saliva tests to detect elevated cortisol levels. Hair cortisol concentration has been described as a marker of long-term exposure to systemic cortisol in patients with Cushing syndrome. Like hemoglobin A1c is used to detect serum glucose exposure over months, segmental hair cortisol can help identify patients with milder forms of and/or periodic or cyclical Cushing syndrome, which may reduce time and costs associated with collection of urine, salivary, and serum cortisol. AREAS COVERED: Success of hair cortisol in detection of Cushing syndrome will be discussed in context of current literature, including differences between total or segmental hair cortisol in accurately determining timeline of cortisol exposure. Optimal methods of hair collection, storage, processing, and analysis and efforts toward standardization will be a major focus. EXPERT COMMENTARY: Recent evidence suggests increased sensitivity and specificity of hair cortisol in detecting Cushing syndrome. Future guidelines should consider this test as a routine part of the repertoire of screening tests for Cushing syndrome. Possible confounders to explain discrepant results in the literature will be discussed.

Hair cortisol in the evaluation of Cushing syndrome.

PURPOSE: Hair cortisol evaluation has been used to help detect patients with suspected Cushing syndrome. Our goal was to correlate segmental hair cortisol with biochemical testing in patients with Cushing syndrome and controls. This study was a prospective analysis of hair cortisol in confirmed Cushing syndrome cases over 16 months. METHODS: Thirty-six subjects (26.5 ± 18.9 years, 75% female, and 75% Caucasian) were analyzed by diurnal serum cortisol, 24 h urinary free cortisol corrected for body surface area (UFC/BSA), and 24 h urinary 17-hydroxysteroids corrected for creatinine (17OHS/Cr). Thirty patients were diagnosed with Cushing syndrome, and six were defined as controls. 3-cm hair samples nearest to the scalp, cut into 1-cm segments (proximal, medial, and distal), were analyzed for cortisol by enzyme immunoassay and measured as pmol cortisol/g dry hair. Hair cortisol levels were compared with laboratory testing done within previous 2 months of the evaluation. RESULTS: Proximal hair cortisol was higher in Cushing syndrome patients (266.6 ± 738.4 pmol/g) than control patients (38.9 ± 25.3 pmol/g) (p = 0.003). Proximal hair cortisol was highest of all segments in 25/36 (69%) patients. Among all subjects, proximal hair cortisol was strongly correlated with UFC/BSA (r = 0.5, p = 0.005), midnight serum cortisol (r = 0.4, p = 0.03), and 17OHS/Cr, which trended towards significance (r = 0.3, p = 0.06). CONCLUSIONS: Among the three examined hair segments, proximal hair contained the highest cortisol levels and correlated the most with the initial biochemical tests for Cushing syndrome in our study. Further studies are needed to validate proximal hair cortisol in the diagnostic workup for Cushing syndrome.

Primary Aldosteronism and ARMC5 Variants.

CONTEXT: Primary aldosteronism is one of the leading causes of secondary hypertension, causing significant morbidity and mortality. A number of genetic defects have recently been identified in primary aldosteronism, whereas we identified mutations in ARMC5, a tumor-suppressor gene, in cortisol-producing primary macronodular adrenal hyperplasia. OBJECTIVE: We investigated a cohort of 56 patients who were referred to the National Institutes of Health for evaluation of primary aldosteronism for ARMC5 defects. METHODS: Patients underwent step-wise diagnosis, with measurement of serum aldosterone and plasma renin activity followed by imaging, saline suppression and/or oral salt loading tests, plus adrenal venous sampling. Cortisol secretion was also evaluated; unilateral or bilateral adrenalectomy was performed, if indicated. DNA, protein, and transfection studies in H295R cells were conducted by standard methods. RESULTS: We identified 12 germline ARMC5 genetic alterations in 20 unrelated and two related individuals in our cohort (39.3%). ARMC5 sequence changes in 6 patients (10.7%) were predicted to be damaging by in silico analysis. All affected patients carrying a variant predicted to be damaging were African Americans (P = .0023). CONCLUSIONS: Germline ARMC5 variants may be associated with primary aldosteronism. Additional cohorts of patients with primary aldosteronism and metabolic syndrome, particularly African Americans, should be screened for ARMC5 sequence variants because these may underlie part of the known increased predisposition of African Americans to low renin hypertension.