qurAI

Information

Location

Netherlands

Website

https://qurai.amsterdam/

Editors

ivanai

csangut

Activity Overview

Contrast-Enhanced Cardiac CT Images

Review automatic whole-heart segmentations in contrast-enhanced CT [demo]

Reader Study

A deep learning algorithm has inferred cardiac segmentations from TAVI treatment planning CT. The reader's purpose is to review the quality of these segmentations.

AIROGS

Challenge

Artificial Intelligence for RObust Glaucoma Screening Challenge

Fluid Segmentation in Retinal Optical Coherence Tomography (OCT)

Algorithm

Segments intraretinal fluid, subretinal fluid, and pigment epithelial detachments in Optical Coherence Tomography scans. Optimized for Spectralis, Cirrus and Topcon scanners.

Calcium scoring in non-contrast CT showing the heart

Algorithm

Automatic quantification of calcifications in the three main coronary arteries (LAD, LCX, RCA) and the thoracic aorta in non-contrast CT scans.

Whole-heart segmentation in non-contrast-enhanced CT

Algorithm

Algorithm for the automatic segmentation of cardiac structures in non-contrast-enhanced CT images. The structures to be segmented are the left ventricular myocardium, left ventricular cavity, right ventricle, left atrium, right atrium, ascending aorta, and the pulmonary artery trunk until the first bifurcation. The underlying publication can be found under https://doi.org/10.1002/mp.14451

AIROGS Baseline

Algorithm

A baseline algorithm for the AIROGS challenge

AIROGS Classifier

Algorithm

A end to end deep learning based system to find referable glaucoma and non-referable glaucoma.

Age-related macular degeneration (AMD) Staging in Optical Coherence Tomography (OCT) with UBIX for Increased Reliability

Algorithm

Deep learning models for optical coherence tomography (OCT) classification often perform well on data from scanners that were also used during training. However, when these models are applied to data from different vendors, their performance tends to drop substantially. Artifacts that only occur within scans from specific scanners are major causes of this poor generalizability. We aimed to improve this generalizability of deep learning models for OCT classifi- cation. To reduce the effect of vendor-specific artifacts, we propose Uncertainty-Based Instance eXclusion (UBIX), of which we define a hard and a soft variant. UBIX aims to suppress the contributions of B-scans with unseen artifacts to the final OCT-level outputs. Suppression is based on out-of-distribution detection of B-scans, which are instances in our multiple instance learning approach.

Whole-heart segmentation in 3D contrast-enhanced cardiac CT

Algorithm

Algorithm that segments the left ventricular cavity and myocardium, left atrium, right atrium , and right ventricle.