|
Description
|
The 2D ultrasound imaging database contains the kidney location and complete diagnosis of 1985 images (with 450 healthy and 1535 pathological kidneys), annotated by two experienced nephrologists from Hospital Universitario Ramón y Cajal. The annotations for each clinical cases include: a polygonal segmentation mask of the kidney, indicators of presence of global pathologies, and also indicators and bounding boxes of the local lesions (when present).
The images are B-mode renal ultrasound images retrospectively collected from patients aged 18 and above, with a balanced distribution across sex. Left and right kidneys are also balanced in the collection, and both transversal (93%) and longitudinal (7%) images are present. Images were anonymized to ensure that they do not contain any personal information that could lead to the identification of the patients, and were collected during the years 2009 and 2018 at the rate of one image per clinical case.
|
|
Notes
| See map
Description of the project:
Ultrasound is a widespread imaging modality, with special application in medical fields such as nephrology. URI-CADS (Computer-Aided Diagnosis System for Ultrasound Renal Imaging) proposes a fully-automated computer-aided diagnosis system for ultrasound renal imaging.Our system is based in a multi-task architecture, which is implemented by a three-branched convolutional neural network, is capable of segmenting the kidney and detecting global and local pathologies with no need of human interaction during diagnosis. The integration of different image perspectives at distinct granularities enhanced the proposed diagnosis. We rely on a large (1985 images) and demanding ultrasound renal imaging database, which has been annotated on the basis of an exhaustive taxonomy of two global and nine local pathologies(including cysts, lithiasis, hydronephrosis, angiomyolipoma, etc), establishing a benchmark for ultrasound renal interpretation. The dataset is described below.
Methodology:
Images were acquired through a Toshiba Xario-660a ecographer with 3MHz and 3.5MHz convex multi-frequency probes and different capture parameters: field of view, zoom, etc. The varied collection of images were written in JPG format, with variable sizes in the range of [375 − 600, 382 − 810] height-width pixels.
An ultrasound renal image depicts a kidney (either in transversal or longitudinal position) which may exhibit pathologies at two different levels: global and/or local. This hierarchical point of view, with two different levels of granularity, is inherent to the interpretation of the ultrasound kidney image and provides valuable information. Two experienced nephrologists from Hospital Universitario Ramón y Cajal, have independently annotated each clinical case, and consensus was reached by discussion in the event of disagreement. Both facultatives were blinded to the patient record. The annotation process has been performed manually through an ad-hoc annotation application. The annotation of each image includes the following fields: an associated segmentation mask of the kidney (a polygonal segmentation delineated over an ellipse drawn by the nephrologists); an indicator of whether it contains global pathologies and, if so, which ones; and, if present, the bounding box coordinates of the local lesions and their indicators.
Based on this dataset, we have designed and implemented a a fully automated Computer-Aided Diagnosis System for Ultrasound Renal Imaging (URI-CADS). To the best of our knowledge, this is the first attempt to simultaneously segment and perform a comprehensive characterization of a complete kidney pathology taxonomy in a real scenario, and to establish a benchmark in this field. The software implementation of this system is publicly available and and further detailed in the following paragraphs.
Software:
Our architecture, based on Mask R-CNN and Faster R-CNN, can be divided into two main blocks: the first one is called SCD-CNN (Segmentation, Classification and Detection CNN), responsible for obtaining the kidney segmentation mask and pathology predictions, from both image- and region-based approaches; and the second one is the Diagnosis Generation Module, which, from the description provided by the SCD-CNN, combines the predictions coming from image- and region-based perspectives to provide a tentative diagnosis for each clinical case.
The code is developed in Pytorch and includes our trained models. It provides the kidney segmentation masks, global pathologies and local pathology detections, and evaluates the corresponding AUC (Area Under the ROC curve) for binary (healthy vs. pathological) and multi-pathological (healthy, poor corticomedullary distinction, hyper-echoic cortex, cyst, pyramid, hydronephrosis and others) diagnosis.
The software corresponding to the article is available in https://github.com/miguel55/URI-CADS. |
