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.

Heroin Inhalation Leukoencephalopathy: An Overlooked Entity in the Opioid Epidemic

Heroin inhalation leukoencephalopathy (HIL) is a rare complication of vaporized heroin inhalation leading to white matter degeneration resulting in a range of neurologic disturbances including softened speech, cerebellar ataxia, behavioral changes, cerebellar gait abnormalities, and even respiratory failure resulting in death in the most severe cases. Magnetic resonance imaging (MRI) most commonly demonstrates bilateral hyperintensities affecting the basal ganglia, periventricular white matter, and cerebellum. In this case report, we present a relatively mild case of HIL in a young female patient to describe the characteristic challenges associated with the condition's presentation, diagnosis, and treatment. While healthcare workers everywhere are addressing a complex and ever-changing opioid epidemic, we strive to raise awareness about HIL as only one of a variety of complications resulting from opioid use disorder.

Call-Fleming syndrome: headache in a 16-year-old girl

BACKGROUND

Call-Fleming syndrome, also known as reversible cerebral vasoconstriction syndrome, is an important cause of severe headache characterized by segmental constriction of cerebral arteries in multiple vascular distributions. It is commonly described in adults, with a female predominance.

PATIENT

We report a case of a 16-year-old girl with history of anxiety, attention deficit hyperactivity disorder, and migraines on several medications presenting with 2 weeks of worsening headaches.

RESULTS

Cranial computed tomography was normal, but magnetic resonance imaging revealed cortical subarachnoid hemorrhage. Follow-up imaging demonstrated extensive vasoconstriction of small- to medium-sized cerebral arteries. Sertraline and methylphenidate were discontinued, and nifedipine was started. Symptoms rapidly improved, and repeat angiography at 2 months showed no vasoconstriction.

CONCLUSIONS

Call-Fleming syndrome is an important cause of thunderclap headache and should be considered in the pediatric population, especially in the setting of certain medication usage and other known risk factors.

Intracranial lesions mimicking neoplasms

CONTEXT

A broad spectrum of nonneoplastic conditions can mimic a brain tumor, both clinically and radiologically. In this review we consider these, taking into consideration the following etiologic categories: infection, demyelination, vascular diseases, noninfectious inflammatory disorders, and iatrogenic conditions. We give an overview of such diseases, which represent a potential pitfall for pathologists and other clinicians involved in patient care, and present selected cases from each category.

OBJECTIVE

To illustrate the radiologic and pathologic features of nontumoral intracranial lesions that can clinically and radiologically mimic neoplasia.

DATA SOURCES

Case-derived material and literature review.

CONCLUSIONS

A variety of nonneoplastic lesions can present clinically and radiologically as primary or metastatic central nervous system tumors and result in surgical biopsy or resection of the lesion. In such situations, the pathologist has an important role to play in correctly determining the nature of these lesions. Awareness of the entities that can present in this way will assist the pathologist in the correct diagnosis of these lesions.

High-grade glioma before and after treatment with radiation and Avastin: initial observations

We evaluate the effects of adjuvant treatment with the angiogenesis inhibitor Avastin (bevacizumab) on pathological tissue specimens of high-grade glioma. Tissue from five patients before and after treatment with Avastin was subjected to histological evaluation and compared to four control cases of glioma before and after similar treatment protocols not including bevacizumab. Clinical and radiographic data were reviewed. Histological analysis focused on microvessel density and vascular morphology, and expression patterns of vascular endothelial growth factor-A (VEGF-A) and the hematopoietic stem cell, mesenchymal, and cell motility markers CD34, smooth muscle actin, D2-40, and fascin. All patients with a decrease in microvessel density had a radiographic response, whereas no response was seen in the patients with increased microvessel density. Vascular morphology showed apparent "normalization" after Avastin treatment in two cases, with thin-walled and evenly distributed vessels. VEGF-A expression in tumor cells was increased in two cases and decreased in three and did not correlate with treatment response. There was a trend toward a relative increase of CD34, smooth muscle actin, D2-40, and fascin immunostaining following treatment with Avastin. Specimens from four patients with recurrent malignant gliomas before and after adjuvant treatment (not including bevacizumab) had features dissimilar from our study cases. We conclude that a change in vascular morphology can be observed following antiangiogenic treatment. There seems to be no correlation between VEGF-A expression and clinical parameters. While the phenomena we describe may not be specific to Avastin, they demonstrate the potential of tissue-based analysis for the discovery of clinically relevant treatment response biomarkers.