An overview of the National COVID-19 Chest Imaging Database: data quality and cohort analysis

Abstract

Background: The National COVID-19 Chest Imaging Database (NCCID) is a centralized database containing mainly chest X-rays and computed tomography scans from patients across the UK. The objective of the initiative is to support a better understanding of the coronavirus SARS-CoV-2 disease (COVID-19) and the development of machine learning technologies that will improve care for patients hospitalized with a severe COVID-19 infection. This article introduces the training dataset, including a snapshot analysis covering the completeness of clinical data, and availability of image data for the various use-cases (diagnosis, prognosis, longitudinal risk). An additional cohort analysis measures how well the NCCID represents the wider COVID-19-affected UK population in terms of geographic, demographic, and temporal coverage.

Findings: The NCCID offers high-quality DICOM images acquired across a variety of imaging machinery; multiple time points including historical images are available for a subset of patients. This volume and variety make the database well suited to development of diagnostic/prognostic models for COVID-associated respiratory conditions. Historical images and clinical data may aid long-term risk stratification, particularly as availability of comorbidity data increases through linkage to other resources. The cohort analysis revealed good alignment to general UK COVID-19 statistics for some categories, e.g., sex, whilst identifying areas for improvements to data collection methods, particularly geographic coverage.

Conclusion: The NCCID is a growing resource that provides researchers with a large, high-quality database that can be leveraged both to support the response to the COVID-19 pandemic and as a test bed for building clinically viable medical imaging models.

Keywords: COVID-19; SARS-CoV2; machine learning; medical imaging; thoracic imaging.

Figure 1:

Diagram of the data collection pipeline for the NCCID warehouse.

Figure 2:

Completeness of clinical data fields related to (A) dates, (B) patient medical history, (C) symptoms on admissions and (D) COVID-related information. bp: blood presure; crp: C-reactive protein; itu: intensive therapy unit; o2: oxygen; pmh: past medical history; wcc: white blood cell count.

Figure 3:

Number of historical/acute/total image studies per NCCID COVID-positive patient (n = 2,826) for (A) X-rays and (B) CTs. In both sets of box plots, outliers are indicated by dots outside the limit of the plot whiskers, and whiskers correspond to Q1 or Q3 ±1.5*iqr (interquartile range).

Figure 4:

(A) Number of days between the patient’s RT-PCR swab test and the image acquisition (n_XRAY = 2,410, n_CT = 507) and (B) Number of days between patient symptom onset and image acquisition (n_XRAY = 803, n_CT = 133). In both sets of box plots, outliers are indicated by dots outside the limit of the plot whiskers, and whiskers correspond to Q1 or Q3 ±1.5*iqr (interquartile range).

Figure 5:

Number of COVID-positive and negative (A) X-ray studies by manufacturer and (B) CT studies by manufacturer. In both cases the manufacturers are ordered by highest to lowest total (positive+negative) number of studies.

Figure 6:

NCCID positive and negative patients submitted by region, sorted by total contribution.

Figure 7:

Comparison of national COVID-19 admissions at a regional level with NCCID positive cases.

Figure 8:

Comparison of sex split within (A) the NCCID COVID-19 patients, the general UK population (as reported in the 2011 census), and COVID-19 hospital admissions (reported by ISARIC); (B) NCCID recorded deaths and NHS England COVID-19 hospital mortality data.

Figure 9:

Comparison of ethnicity proportions within (A) the NCCID COVID-19 patients, the UK population (as reported in the 2011 national census), and COVID-19 hospital admissions (reported by ISARIC); (B) the NCCID recorded deaths and NHS England COVID-19 hospital mortality data.

Figure 10:

Comparison of age proportions between COVID-19 hospital admissions (reported by PHE) and NCCID positive patients for (A) England, (B) East of England, (C) London, (D) Midlands, (E) Northeast and Yorkshire, (F) Northwest (G), Southeast, and (H) Southwest.

Figure 11:

Comparison of age distributions between recorded COVID-19 deaths (as reported by NHSE) and the NCCID (England only).

Figure 12:

Comparison of COVID-19 admissions to NCCID positive cases by week.