NIH Clinical Center Releases CT Image Dataset

The NIH radiologists measured and marked clinically meaningful findings with the aid of a complex electronic bookmark tool that provides arrows, lines, diameters, and text that can tell the exact location and size of a lesion so experts can identify growth or new disease. The bookmarks, including a range of retrospective medical data, were used to develop the DeepLesion dataset. Unlike most lesion medical image datasets currently available, which can only detect one type of lesion, the database contains all critical radiology findings, such as lung nodules, liver tumors, enlarged lymph nodes, and so on.

The dataset released is large enough to train a deep neural network, which could enable the scientific community to create a large-scale universal lesion detector with one unified framework that could eventually serve as an initial screening tool for other specialist systems trained on certain types of lesions. In addition, DeepLesion marks multiple findings in one CT exam image, allowing researchers to analyze their relationship to make new discoveries, enabling whole body assessment of cancer burden. DeepLesion was introduced in a study published on July 20, 2018, in the Journal of Medical Imaging.

“Vast amounts of clinical annotations have been collected and stored in hospitals’ picture archiving and communication systems. These types of annotations, also known as bookmarks, are usually marked by radiologists during their daily workflow to highlight significant image findings that may serve as reference for later studies,” said senior author Ronald Summers, MD, PhD, and colleagues. “We propose to mine and harvest these abundant retrospective medical data to build a large-scale lesion image dataset.”

“In the future, the NIH Clinical Center hopes to keep improving the DeepLesion dataset by collecting more data, thus improving its detection accuracy,” stated the NIH in a press release. “The universal lesion detecting capability will become more reliable once researchers are able to leverage 3D and lesion type information. It may be possible to further extend DeepLesion to other image modalities such as MRI and combine data from multiple hospitals, as well.”

Related Links:
U.S. National Institutes of Health