Effectiveness of temporal subtraction computed tomography images using deep learning in detecting vertebral bone...

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Artificial Intelligence in Medicine

Background:

• Accurate detection of vertebral bone metastases is crucial for cancer patient management.
• Traditional CT imaging can be limited in identifying subtle bone lesions.
• Deep learning and advanced image processing techniques offer potential for improved diagnostic performance.

Purpose of the Study:

• To evaluate the clinical effectiveness of temporal subtraction computed tomography (TS CT) enhanced by deep learning for detecting vertebral bone metastases.
• To assess whether TS CT improves diagnostic accuracy compared to standard CT interpretation.

Main Methods:

• A retrospective study utilizing TS CT, involving deep learning for registration error estimation and false positive suppression.
• Paired CT images from patients with and without vertebral metastases were analyzed.
• Six board-certified radiologists and five radiology residents interpreted images with and without TS CT.

Main Results:

• TS CT significantly improved the overall figure of merit (FOM) for both board-certified radiologists (0.848 to 0.876) and residents (0.752 to 0.799).
• Sub-analysis showed improved FOM for osteoblastic lesions in both radiologist groups using TS CT.
• TS CT demonstrated significant improvement in detecting lesions in the vertebral arch for the resident group.

Conclusions:

• Temporal subtraction computed tomography (TS CT) is an effective tool for enhancing the detection of vertebral bone metastases.
• The deep learning-enhanced TS CT improves diagnostic accuracy for both experienced radiologists and residents.
• This technique holds promise for improving the clinical management of patients with bone metastases.