Accurate, automated extraction of clinical stroke information from unstructured text has several important applications. ICD-9/10 codes can misclassify ischemic stroke events and do not distinguish acuity or location. Expeditious, accurate data extraction could provide considerable improvement in identifying stroke in large datasets, triaging critical clinical reports, and quality improvement efforts. In this study, we developed and report a comprehensive framework studying the performance of simple and complex stroke-specific Natural Language Processing (NLP) and Machine Learning (ML) methods to determine presence, location, and acuity of ischemic stroke from radiographic text. We collected 60,564 Computed Tomography and Magnetic Resonance Imaging Radiology reports from 17,864 patients from two large academic medical centers. We used standard techniques to featurize unstructured text and developed neurovascular specific word GloVe embeddings. We trained various binary classification algorithms to identify stroke presence, location, and acuity using 75% of 1,359 expert-labeled reports. We validated our methods internally on the remaining 25% of reports and externally on 500 radiology reports from an entirely separate academic institution. In our internal population, GloVe word embeddings paired with deep learning (Recurrent Neural Networks) had the best discrimination of all methods for our three tasks (AUCs of 0.96, 0.98, 0.93 respectively). Simpler NLP approaches (Bag of Words) performed best with interpretable algorithms (Logistic Regression) for identifying ischemic stroke (AUC of 0.95), MCA location (AUC 0.96), and acuity (AUC of 0.90). Similarly, GloVe and Recurrent Neural Networks (AUC 0.92, 0.89, 0.93) generalized better in our external test set than BOW and Logistic Regression for stroke presence, location and acuity, respectively (AUC 0.89, 0.86, 0.80). Our study demonstrates a comprehensive assessment of NLP techniques for unstructured radiographic text. Our findings are
Objectives: This study examines the unique factors that influence loss to follow-up after newborn hearing screening for patients at a Massachusetts urban safety-net hospital. We seek to characterize our patient population, investigate correlations between patient factors and rates of follow-up, and understand gaps in care. Design: A retrospective chart review was conducted of patients born at an urban safety-net hospital from January 2015 through May 2018 who did not pass the newborn hearing screening in one or both ears. A total of 197 infants were included in our study. Outcomes of interest included rates and latency of follow-up appointments, infant demographics (sex, race, birth weight, risk factors for hearing loss), and maternal factors (age, marital status, smoking status, number of children). Results: From January 2015 through May 2018, 17% (n = 34) of infants were lost to follow-up. Of those who attended an initial audiology evaluation, the median time between screening and appointment was 29 days. Newborns were 3.5 times at risk of being lost to follow-up if their mothers smoked during pregnancy compared to those whose mothers did not smoke. Further, newborns with multiple siblings in the home were less likely to utilize any audiological services. High-risk infants, such as those with an extended stay in the neonatal intensive care unit, were found to have higher rates of loss to follow-up. Conclusions: Our results indicate that patients at urban safety-net hospitals require increased support to decrease rates of loss to follow-up. In particular, strategies to aid mothers who smoke, have multiple children, or have high-risk infants can address gaps in care for newborns after hearing screening.
Introduction: Chronic Helicobacter pylori (HP) infection has been shown to be strongly associated with development of gastric malignancies, mostly gastric adenocarcinomas and lymphomas. We present a case of primary gastric leiomyosarcoma in a patient with persistent epigastric pain and HP infection. Case Description/Methods: A 59-year-old female presented to clinic with persistent, postprandial epigastric pain. Cardiopulmonary workup was negative. She had a history of esophageal leiomyoma resected endoscopically with negative margins 10 years ago. She was treated in the past for HP infection but never had repeat testing to confirm eradication. Abdominal computed tomography showed focal hepatic infiltration. She was advised to have a repeat upper endoscopy for surveillance but was lost to follow-up due to psychiatric comorbidities. She presented 6 years later with epigastric pain. Repeat CT abdomen showed a new 2.8 cm hypoattenuating lesion in the lesser curvature of the stomach. Patient did not follow up due to psychosocial issues. Two years later, she returned to clinic for epigastric pain and underwent repeat upper endoscopy and endoscopic ultrasound (EUS) with fine needle aspiration (FNA). EGD showed sub-gastric nodule attached gastric wall, about 30 mm in maximal dimension and gastric biopsy confirmed active HP infection. The EUS showed a 31.7x23.5 mm hypoechoic homogeneous gastric mass along the lesser gastric curvature. Cytology showed cellular spindle cell neoplasia with myogenic differentiation concerning for possible leiomyoma or leiomyosarcoma. Abdominal magnetic resonance imaging and chest computed tomography revealed no evidence of abdominal or thoracic metastasis. Subsequently, partial underwent sleeve gastrectomy. Pathology revealed Grade 1 myxoid leiomyosarcoma with negative margins. (Figure ) Discussion: Prior studies demonstrated associations between HP infection and gastric malignancies -typically gastric adenocarcinoma or lymphomas. Few studies investigated the relationship between HP infection and gastric leiomyosarcomas.4,5 Importantly, this patient never had clearance of HP infection and may have allowed a leiomyosarcoma to develop. It is imperative that clinicians confirm eradication of HP infection given risk of leiomyosarcoma development in addition to other malignancies HP infections are known to cause.[3613] Figure 1. Endoscopic and histologic examination of gastric leiomyosarcoma. (A) Endoscopic ultrasound visualization demonstrating tumor within lesser curvature of stomach. (B) Upper endoscopy revealing tumor approximately 4 cm from gastroesophageal junction. (C) Hematoxylin and eosin (H&E) stain of gastric tumor biopsy demonstrating spindle cell proliferation in a myxoid background with significant nuclear atypia and pleomorphism consistent with myxoid leiomyosarcoma. Magnification x400. (D) Immunohistochemical staining of gastric lesion biopsy demonstrating strong immunoreactivity to desmin. Not shown are positive stains for smooth muscle actin (SMA) as well as negative stains for C...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.