Ground Glass Lesions on Chest Imaging: Evaluation of Reported Incidence in Cancer Patients Using Natural Language Processing

Abstract

Background. Ground glass opacities (GGOs) on computed tomography (CT) have gained significant recent attention, with unclear incidence and epidemiologic patterns. Natural language processing (NLP) is a powerful computing tool that collects variables from unstructured data fields. Our objective was to characterize trends of GGO detection using NLP.Methods. Patients were identified at a large quaternary referral center who underwent chest CT from 2000 to 2016 via query of institutional databases. NLP was used to identify imaging reports with GGOs and to obtain additional demographic data. Incidence of reported GGOs was tracked over time. Multivariate regression was used to identify predictors of GGOs identified on chest CT.Results. A total of 244,391 chest CTs were included, with 35,386 (14.5%) revealing GGOs. There was a significant relationship between advancing year of chest CT and likelihood of reported GGOs (p < 0.001). GGOs were more likely to occur in older subjects (60.5 vs 58.5 years, p

Image‐guided video‐assisted thoracoscopic resection (iVATS): Translation to clinical practice—real‐world experience

“…The ground glass and partially solid nodules are particularly challenging, as imaging characterization, biopsy, and even surgical resection by standard of care procedures can be difficult. This often results in prolonged follow‐up till the lesions are deemed resectable 5,30‐32 . In addition, it is difficult to characterize small partially solid nodules based on imaging alone and there continues to be significant interobserver variability in reporting and characterizing ground glass nodules 31 .…”

“…This poses a significant diagnostic burden on radiologists, pulmonologists, and surgeons. In addition, ground glass nodules have gained significant attention with the new pathological classification of lung cancer 4,5 . Management of pulmonary nodules detected by CT depends on their size, radiographic appearance, rate of growth, and pretest probability of malignancy.…”

A systematic review of natural language processing applied to radiology reports

“…Out of these papers, the majority (n = 8) created cohorts for specific medical conditions including fatty liver disease [92,93] hepatocellular cancer [94], ureteric stones [95], vertebral fracture [96], traumatic brain injury [97,98], and leptomeningeal disease secondary to metastatic breast cancer [99]. Five papers identified cohorts focused on particular radiology findings including ground glass opacities (GGO) [100], cerebral microbleeds (CMB) [101], pulmonary nodules [102,103], changes in the spine correlated to back pain [1] and identifying radiological evidence of people having suffered a fall. One paper focused on identifying abnormalities of specific anatomical regions of the ear within an audiology imaging database [104] and another paper aimed to create a cohort of people with any rare disease (within existing ontologies -Orphanet Rare Disease Ontology ORDO and Radiology Gamuts Ontology RGO).…”

Advanced Data Analytics for Clinical Research Part II: Application to Cardiothoracic Surgery

“…If a patient population is identified by a specific term in written text within the medical record, then by using all variations of that term, a specific patient cohort can be efficiently identified, as demonstrated in Examples 2 and 3 using all variations of GGOs. 6,11 Finally, for a patient population requiring manual review, NLP can still be used to augment the search results to exclude patients and lessen the burden of manual chart review. As seen in Example 4, 13 2,988 patients were narrowed to 849, substantially lowering the burden of manual chart review.…”