VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations.

Most of the existing chest X-ray datasets include labels from a list of findings without specifying their locations on the radiographs. This limits the development of machine learning algorithms for the detection and localization of chest abnormalities. In this work, we describe a dataset of more than 100,000 chest X-ray scans that were retrospectively collected from two major hospitals in Vietnam. Out of this raw data, we release 18,000 images that were manually annotated by a total of 17 experienced radiologists with 22 local labels of rectangles surrounding abnormalities and 6 global labels of suspected diseases. The released dataset is divided into a training set of 15,000 and a test set of 3,000. Each scan in the training set was independently labeled by 3 radiologists, while each scan in the test set was labeled by the consensus of 5 radiologists. We designed and built a labeling platform for DICOM images to facilitate these annotation procedures. All images are made publicly available in DICOM format along with the labels of both the training set and the test set.

Choroidal nevus transformation into melanoma: analysis of 2514 consecutive cases.

OBJECTIVE: To determine features that are predictive of growth of choroidal nevi into melanoma. METHODS: This was a retrospective medical record review of 2514 consecutive eyes; Kaplan-Meier estimates and Cox regression analyses were used. RESULTS: The median tumor basal diameter was 5.0 mm and thickness was 1.5 mm. Nevus growth into melanoma occurred in 2%, 9%, and 13% of eyes at 1, 5, and 10 years, respectively. Factors predictive of growth into melanoma by multivariable analysis included tumor thickness greater than 2 mm (P < .001), subretinal fluid (P = .002), symptoms (P = .002), orange pigment (P < .001), tumor margin within 3 mm of the optic disc (P = .001), ultrasonographic hollowness (P < .001), and halo absence (P = .009). A mnemonic device to recall risk factors of ocular melanoma is "To find small ocular melanoma using helpful hints," representing thickness, fluid, symptoms, orange pigment, margin, ultrasonographic hollowness, and halo absence. The median hazard ratio for those with 1 to 2 risk factors was 3; for 3 or 4 factors, 5; for 5 to 6 factors, 9; and for all 7 factors, 21. CONCLUSIONS: In an analysis of 2514 choroidal nevi, factors predictive of growth into melanoma included greater thickness, subretinal fluid, symptoms, orange pigment, margin near disc, and 2 new features: ultrasonographic hollowness and absence of halo.

Clinical spectrum of choroidal nevi based on age at presentation in 3422 consecutive eyes.

PURPOSE: To evaluate the clinical features of choroidal nevi based on patient age at presentation and to investigate features of the nevi that are predictive of patient symptoms. DESIGN: Observational case series. PARTICIPANTS: Three thousand four hundred twenty-two consecutive eyes of 3187 patients. METHODS: Retrospective clinic-based study of clinical features at referral. Cox proportional hazards regressions were used for evaluation of factors predictive of patient symptoms. MAIN OUTCOME MEASURES: Nevus features as related to patient age group at diagnosis (young [< or =20 years], mid-adult [21-50 years], older adult [>50 years]) and factors predictive of patient symptoms secondary to the nevus. RESULTS: Of the 3422 eyes with choroidal nevus, 63 (2%) were in young patients, 795 (23%) in mid-adults, and 2564 (75%) in older adults. The following factors showed no substantial increase or decrease by age category (young, mid-adult, older adult) at presentation: symptoms (14%, 12%, 13%), mean nevus base (5.6, 4.7, 5.2 mm), intrinsic nevus pigmentation (89%, 74%, 77%), related subretinal fluid (SRF) (11%, 15%, 9%), overlying orange pigment (6%, 10%, 6%), retinal pigment epithelial hyperplasia (0%, 9%, 7%), and retinal pigment epithelial atrophy (2%, 13%, 10%). The following factors statistically increased with age category: multiple nevi per eye (2%, 8%, 10%) (P = 0.0001), mean nevus thickness (1.2, 1.5, 1.6 mm) (P<0.0001), and overlying drusen (11%, 40%, 58%) (P<0.0001). Using multivariate analysis of the entire group, factors predictive of any symptom included nonpigmented nevus (P<0.001), location < or = 3 mm to foveola (P = 0.001), subfoveolar fluid (P = 0.002), any SRF (P = 0.02), and subfoveolar nevus (P = 0.027). CONCLUSIONS: Choroidal nevi show similar clinical features regardless of age of presentation, with the exception of increasing number of nevi per eye, slightly increasing thickness, and increasing drusen in adults versus younger patients. Symptomatic nevi are more likely to be nonpigmented, beneath the foveola, and with subfoveolar fluid.

Visual acuity in 3422 consecutive eyes with choroidal nevus.

OBJECTIVE: To evaluate visual acuity in eyes with choroidal nevus. DESIGN: This was an observational case series. Of 3422 consecutive eyes with choroidal nevus, vision loss at 15 years occurred in 2% of eyes with extrafoveolar nevus and in 26% of eyes with subfoveolar nevus, particularly those with overlying retinal pigment epithelial detachment and foveal edema. A retrospective medical record review was conducted, with evaluation of visual acuity at presentation and at final examination. The main outcome measure was visual acuity. RESULTS: The median visual acuity at presentation was 20/20 for eyes with either extrafoveolar or subfoveolar choroidal nevus. Using Kaplan-Meier estimates, vision loss of 3 or more logarithm of the minimum angle of resolution (logMAR) lines at 5, 10, and 15 years occurred in less than 1%, 1%, and 2% of eyes with extrafoveolar nevus compared with 15%, 20%, and 26% of eyes with subfoveolar choroidal nevus, respectively. By multivariate analysis, factors predictive of visual loss of 3 or more logMAR lines included subfoveolar nevus location (relative risk [RR], 15.52), juxtapapillary nevus location (RR, 4.52), initial visual acuity of 20/50 or worse (RR, 15.40), overlying retinal pigment epithelial detachment (RR, 22.16), and foveal edema (RR, 9.02). Factors predictive of poor final visual acuity of 20/200 or worse included subfoveolar nevus location (RR, 11.32), overlying orange pigment (RR, 3.68), overlying retinal pigment epithelial detachment (RR, 12.80), and foveal edema (RR, 18.72). CONCLUSION: Mild vision loss over many years should be anticipated in patients with subfoveolar choroidal nevus, particularly those with overlying retinal pigment epithelial detachment, orange pigment, and foveal edema.