Large, labeled datasets have driven deep learning methods to achieve expert-level performance on a variety of medical imaging tasks. We present CheXpert, a large dataset that contains 224,316 chest radiographs of 65,240 patients. We design a labeler to automatically detect the presence of 14 observations in radiology reports, capturing uncertainties inherent in radiograph interpretation. We investigate different approaches to using the uncertainty labels for training convolutional neural networks that output the probability of these observations given the available frontal and lateral radiographs. On a validation set of 200 chest radiographic studies which were manually annotated by 3 board-certified radiologists, we find that different uncertainty approaches are useful for different pathologies. We then evaluate our best model on a test set composed of 500 chest radiographic studies annotated by a consensus of 5 board-certified radiologists, and compare the performance of our model to that of 3 additional radiologists in the detection of 5 selected pathologies. On Cardiomegaly, Edema, and Pleural Effusion, the model ROC and PR curves lie above all 3 radiologist operating points. We release the dataset to the public as a standard benchmark to evaluate performance of chest radiograph interpretation models. 1

No abstract

Purpose:To investigate the potential of the normalized probabilistic atlases and computer-aided medical image analysis to automatically segment and quantify livers and spleens for extracting imaging biomarkers ͑volume and height͒. Methods: A clinical tool was developed to segment livers and spleen from 257 abdominal contrastenhanced CT studies. There were 51 normal livers, 44 normal spleens, 128 splenomegaly, 59 hepatomegaly, and 23 partial hepatectomy cases. 20 more contrast-enhanced CT scans from a public site with manual segmentations of mainly pathological livers were used to test the method. Data were acquired on a variety of scanners from different manufacturers and at varying resolution. Probabilistic atlases of livers and spleens were created using manually segmented data from ten noncontrast CT scans ͑five male and five female͒. The organ locations were modeled in the physical space and normalized to the position of an anatomical landmark, the xiphoid. The construction and exploitation of liver and spleen atlases enabled the automated quantifications of liver/spleen volumes and heights ͑midhepatic liver height and cephalocaudal spleen height͒ from abdominal CT data. The quantification was improved incrementally by a geodesic active contour, patient specific contrast-enhancement characteristics passed to an adaptive convolution, and correction for shape and location errors. Results: The livers and spleens were robustly segmented from normal and pathological cases. For the liver, the Dice/Tanimoto volume overlaps were 96.2%/92.7%, the volume/height errors were 2.2%/2.8%, the root-mean-squared error ͑RMSE͒ was 2.3 mm, and the average surface distance ͑ASD͒ was 1.2 mm. The spleen quantification led to 95.2%/91% Dice/Tanimoto overlaps, 3.3%/ 1.7% volume/height errors, 1.1 mm RMSE, and 0.7 ASD. The correlations ͑R 2 ͒ with clinical/ manual height measurements were 0.97 and 0.93 for the spleen and liver, respectively ͑p Ͻ 0.0001͒. No significant difference ͑p Ͼ 0.2͒ was found comparing interobserver and automaticmanual volume/height errors for liver and spleen. Conclusions: The algorithm is robust to segmenting normal and enlarged spleens and livers, and in the presence of tumors and large morphological changes due to partial hepatectomy. Imaging biomarkers of the liver and spleen from automated computer-assisted tools have the potential to assist the diagnosis of abdominal disorders from routine analysis of clinical data and guide clinical management.

Abstract. The paper presents the automated segmentation of spleen and liver from contrast-enhanced CT images of normal and hepato/splenomegaly populations. The method used 4 steps: (i) a mean organ model was registered to the patient CT; (ii) the first estimates of the organs were improved by a geodesic active contour; (iii) the contrast enhancements of liver and spleen were estimated to adjust to patient image characteristics, and an adaptive convolution refined the segmentations; (iv) lastly, a normalized probabilistic atlas corrected for shape and location for the precise computation of each organ's volume and height (mid-hepatic liver height and cephalocaudal spleen height). Results from test data demonstrated the method's ability to accurately segment the spleen (RMS error = 1.09mm; DICE/Tanimoto overlaps = 95.2/91) and liver (RMS error = 2.3mm, and DICE/Tanimoto overlaps = 96.2/92.7). The correlations (R 2 ) with clinical/manual height measurements were 0.97 and 0.93 for the spleen and liver respectively.

Rationale and Objectives To define systematic volumetric thresholds to identify and grade splenomegaly, and retrospectively evaluate the performance of radiologists to assess splenomegaly in computed tomography image data. Materials and Methods A clinical tool was developed to segment spleens from 172 contrast-enhanced clinical CT studies. There were 45 normal and 127 splenomegaly cases confirmed by radiological reports. Spleen volumes were compared to manual measurements using overlap/error. Volumetric thresholds for mild/massive splenomegaly were defined at 1/2.5 standard deviations above the average splenic volume of the healthy population. The thresholds were validated against consensus reports. The performance of radiologists in assessing splenomegaly was retrospectively evaluated. Results The automated segmentation of spleens was robust with volume overlap/error of 95.2/3.3%. There were no significant differences (p>0.2) between manual and automated segmentations for either normal/splenomegaly subgroups. Comparable correlations between interobserver and manual-automated measurements were found (R=0.99 for all). The average volume of normal spleens was 236.89±77.58 ml. For splenomegaly, average volume was 1004.75±644.27 ml. Volumetric thresholds of 314.47/430.84 ml were used to define mild/massive splenomegaly (+/−18.86 ml 95% CI). Radiologists disagreed in 23.25% (n=40) of the diagnosed cases. The area under the ROC curve of the volumetric criterion for splenomegaly detection was 0.96. Using the volumetric thresholds as the reference standard, the sensitivity of radiologists in detecting all/mild/massive splenomegaly was 95.0/66.6/99.0% at 78.0% specificity, respectively. Conclusion Thresholds for the identification and grading of splenomegaly from automatic volumetric spleen assessment were introduced. The volumetric thresholds match well with clinical interpretations for splenomegaly and may improve splenomegaly detection compared with splenic cephalocaudal height measurements or visual inspection commonly used in current clinical practice.