Panoramic radiography Images with diagnosis of need for Apicoectomy surgery as label

This data set consists of panoramic radiography images of patient’s jaws. data was labeled by skilled endodontist. this dataset was collected for the purpose of training machine learning algorithm to automate diagnosis workflow.

There is a test directory in the dataset this portion of data is remained unseen in training of model and was used only for assessment and comparison.

Data is annotated using bounding boxes, each bounding box captures a single tooth that either was subjected to root canal treatment in the past (class 0) or is considered in need of surgery by endodontist (class 1). Bounding boxes of each image can be found in text file with same name of image each line indicates a bonding box location and class in format of (class_id x_center y_center w h), keep in mind that this numbers are normalized. Beside bonding boxes there is binary label for each original image that indicate final result of image, if there is found one or more teeth in need of surgery final label is 1 meaning doctor deduced patient needs surgery treatment; otherwise , label is 0 meaning patient doesn’t need this treatment.

in training set there is a directory named single_teeth, contents of this directory are cropped from original image in the other word they are teeth that labels suggest need further investigation.

Original images are resized to (2000 ,1000, 3) for convince of use and storage they; however, single_teeth images have various shapes due nature of their content. also, all images in dataset are saved in PNG or JPG format not DICOM.