aisc-team-b1/PMC-CaseReport
收藏资源简介：
PMC-CaseReport数据集是一个用于视觉问答（VQA）任务的医学案例报告数据集，包含训练集和测试集，分别有316,838和120,836个样本。每个样本包含PMC_id、context、question、answer、inline和img_ref等特征。数据集中的文本部分和图像部分可以分别下载，图像部分以PMCxxxxxxx_figid.jpg的格式命名。数据集中的context字段包含案例报告中的上下文信息，question字段是生成的问题，answer字段是正确答案，img_ref字段是与图像相关的ID列表。需要注意的是，部分案例中的答案可能会在context中泄露，且由于数据收集时间窗口的原因，部分图像可能缺失。
The PMC-CaseReport dataset is a medical case report dataset designed for visual question answering (VQA) tasks. It comprises a training set and a test set, containing 316,838 and 120,836 samples respectively. Each sample includes features such as PMC_id, context, question, answer, inline, and img_ref. The textual and image components of the dataset can be downloaded separately, with image files named in the format PMCxxxxxxx_figid.jpg. Specifically, the context field holds the contextual information from the case report, the question field contains a generated question, the answer field holds the correct answer, and the img_ref field is a list of IDs associated with relevant images. It is worth noting that the answers to some cases may be inadvertently leaked in the context, and some images may be missing due to the data collection time window.
PMC-CaseReport Dataset 概述
数据集结构
-
特征字段：
- PMC_id: 字符串类型，对应的PMC论文ID。
- context: 字符串类型，病例报告讨论图像前的上下文。
- question: 字符串类型，生成的问题。
- answer: 字符串类型，正确答案。
- inline: 字符串类型，原文中的内联句子，不应输入网络。
- img_ref: 字符串类型，相关图像ID的列表。
-
数据集划分：
- 训练集：316838个样本，占用634222398字节。
- 测试集：120836个样本，占用253538916字节。
-
数据集大小：总下载大小为139781550字节，总数据集大小为887761314字节。
样本示例
| PMC_id | PMC9052276 |
|---|---|
| context | We report the case of a 73-year-old female who presented to the ER with left-sided body weakness of unclear duration.She had an ischemic stroke four years prior with no residual neurologic deficits, a myocardial infarction requiring coronary artery bypass grafting (CABG) two years prior, hypertension, and dementia. Her vital signs were blood pressure (BP) 117/78 mmHg, pulse 121 beats per minute, temperature 98.9 F, respiratory rate (RR) 18 cycles/minute, and oxygen saturation (SpO2) of 97% on ambient air.She was disoriented to place and time with a Glasgow Coma Score (GCS) of 14 (E4V4M6).Her speech was slurred, cranial nerves (CN) 2-12 were grossly intact, motor strength on the left upper and lower extremities was 0/5 and on the right upper and lower extremities was 4/5, and the sensation was preserved in all extremities.The patient had a National Institutes of Health Stroke Scale (NIHSS) score of 16 and a Modified Rankin Score (mRS) of 5 points.A non-contrast head CT scan revealed evidence of old lacuna infarcts in the basal ganglia and thalamus.No intracranial hemorrhage or acute infarct was found.CT perfusion was not done as our center lacks the resources needed to perform that. |
| inline | A brain MRI scan showed an acute pontine stroke (Figures and old infarcts |
| question | What did the brain MRI scan reveal? |
| answer | The brain MRI scan showed an acute pontine stroke and old infarcts. |
| img_ref | "[FIG1, FIG3, FIG4]" |
注意事项
- 上下文未严格过滤，少数情况下答案可能泄露在上下文中。
- 由于论文更新，部分图像可能不在图像库中。
