European Radiolog: Deep Learning Improves Radiologists' Performance in CT Bone Metastasis Detection

Clinically, early diagnosis of bone metastases is a common and important task for radiologist.

Currently, several imaging modalities such as CT, MRI, bone scan, and PET are available for the diagnosis of bone metastases, each with distinct advantage.

However, detection of bone metastases on CT remains challenging for several reason.

Recently, published in European Radiology, a deep learning algorithm (DLA) was developed for the automatic detection of bone metastases on CT, and the clinical diagnosis effect of the algorithm was evaluated, which provided a basis for the early diagnosis and treatment of bone metastases in cancer patient.

This retrospective study included CT scans obtained at an institution between 2009 and 2019, collecting positive scans with bone metastases and negative scans without bone metastases to train DL.

A total of 269 positive scans were collected in the training dataset, including 1375 bone metastases and 463 negative scan.

This study successfully developed a deep learning-based algorithm for automatic detection of bone metastases on CT, providing technical support for the formulation of personalized treatment plans for cancer patient.

Original source:

Shunjiro Noguchi, Mizuho Nishio, Ryo Sakamoto, et a.