Towards automatic bone age estimation from MRI: localization of 3D anatomical landmarks

Area of Science:

• Medical Imaging
• Forensic Science
• Computer Vision

Background:

• Bone age estimation (BAE) is critical in forensic medicine.
• Recent advancements shift BAE from X-ray to MRI-based imaging.
• Automating BAE from MRI requires precise hand bone joint localization, which is challenging due to anatomical variations and complex hand structures.

Purpose of the Study:

• To develop an automated landmark localization algorithm for MRI-based BAE.
• To improve the accuracy and efficiency of joint detection in hand MR images.

Main Methods:

• A novel algorithm employing multiple random regression forests for landmark localization.
• Initial analysis of global hand shape to model landmark configuration.
• Refinement using local image information to enhance prediction accuracy.

Main Results:

• The algorithm achieved a mean landmark localization error of 1.4 ± 1.5mm on a dataset of 60 T1-weighted MR images.
• Demonstrated superior performance compared to related approaches.
• Reported a low outlier rate of only 0.25% for errors greater than 10mm.

Conclusions:

• The proposed random regression forest algorithm effectively automates landmark localization for MRI-based BAE.
• This method offers high accuracy and robustness, outperforming existing techniques.
• The findings support the potential of automated MRI analysis for forensic and clinical applications.