Diabetic Macular Edema VQA Dataset

<p>Medical VQA dataset built from the&nbsp;<a href="https://ieee-dataport.org/open-access/indian-diabetic-retinopathy-image-dataset-idrid">IDRiD</a>&nbsp;and&nbsp;<a href="https://www.adcis.net/en/third-party/e-ophtha/">eOphta</a>&nbsp;datasets. The dataset contains both healthy and unhealthy fundus images. For each image, a set of pre-defined questions is generated, including questions about regions (e.g. are there hard exudates in this region?), for which an associated mask denotes the location of the region.</p> <p>The motivation for this dataset includes the lack of public medical VQA datasets with related questions. In our dataset, questions are related because there is a high-level question about the DME grade of the image, and associated low-level questions that can lead to the answer of the high-level question. This allows to study the consistency of a VQA model i.e. how often the model produces contradictory answers to questions about a given image. Questions about regions are also a novel feature of this dataset.</p> <p>The dataset can be used for general VQA purposes, and also for the more specific purpose of consistency improvement.</p> <p>Number of images : Train: 433 Val: 112 Test: 134</p> <p>Number of QA pairs: Train: 9779 Val: 2380 Test: 1311</p> <p>More details can be found&nbsp;<a href="https://github.com/sergiotasconmorales/consistency_vqa/blob/master/DATA.md">here</a>.</p> <p>If you use this dataset, please make sure you cite <a href="https://arxiv.org/abs/2206.13296">our paper</a>:</p> <p><em>@inproceedings{tascon2022consistency,<br> &nbsp; title={Consistency-Preserving Visual Question Answering in Medical Imaging},<br> &nbsp; author={Tascon-Morales, Sergio and M&aacute;rquez-Neila, Pablo and Sznitman, Raphael},<br> &nbsp; booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},<br> &nbsp; pages={386--395},<br> &nbsp; year={2022},<br> &nbsp; organization={Springer}<br> }</em></p> <p>Do you need annotations about logical relations? No problem; check out our <a href="https://zenodo.org/record/7777849">DME VQA dataset with logical relations</a>.</p>