Cataract Surgery Dataset for Eye Positioning and Alignment
Computer-assisted intraoperative intraocular lens (IOL) positioning and alignment is a valuable study. It is important to precisely position and align the axis of IOL during surgery to achieve optimal post-operative astigmatism correction. The cataract surgery dataset is proposed in the research paper “Computer-aided Intraoperative Toric Intraocular Lens Positioning and Alignment During Cataract Surgery”. Videos of 20 cataract patients under routine cataract surgery were collected from a tertiary eye hospital (Shanxi Eye hospital, Taiyuan, Shanxi, China) with a resolution of 1920 × 1080 pixels and a frame rate of 25 FPS. The videos of these patients were divided into three groups, 10 as the training set (Video 11-20), 3 as the evaluation set (Video 8-10) and 7 as the test set (Video 1-7). We manually labeled the limbus of the surgical images in the training set. The limbus and blood vessel bifurcation points near the limbus in the evaluation set and test set were manually labeled. This cataract surgery dataset can be used to develop and evaluate algorithms for eyeball rotation and positioning during cataract surgery.
The training set was constructed by taking one frame every 50 frames from 10 training videos. A total of 3105 images were labeled by an image analyst. The videos of the evaluation set and the test set were divided into several short video clips, and each clip is the first two seconds (50 frames) of each minute from each video. As a result, 53 video clips were obtained in the test set and 16 video clips were obtained in the evaluation set. After labeling the limbus, we selected four bifurcation points in each video clip and labeled them in each frame. The information of these points was recorded in the .txt files. The first column is the frame number, the second column is the point number, the third column is the point's x-axis value, and the fourth column is the point's y-axis value. Each video clip is separated by asterisks. If the coordinate of a point is (0, 0), it means that this point was not found in this frame, it may be too blurry or not in the image. The first four records of each video clip are the coordinates of these four points in the first frame. Video 1-3 in the test set were labeled by two image analysts to assess inter-observer variability. These three videos contain a total of 25 video clips.