Machine learning approach to needle insertion site identification for spinal anesthesia in obese patients

Area of Science:

• Anesthesiology
• Medical Imaging
• Machine Learning

Background:

• Neuraxial anesthesia increasingly utilizes ultrasonography for spinal landmark identification.
• Obesity presents challenges in identifying correct needle insertion points for spinal anesthesia.
• An automated ultrasound-guided program was developed to assist anesthetists.

Purpose of the Study:

• To evaluate an automated ultrasound-guided program for spinal landmark identification.
• To assess the program's efficacy in assisting needle insertion point identification in obese patients.
• To determine the correlation between program-measured and clinician-measured depths.

Main Methods:

• A prospective cohort study of 48 obese patients undergoing spinal anesthesia for Cesarean section.
• Utilized a machine learning algorithm with automated spinal landmark ultrasound imaging.
• Identified the L3/4 interspinous space and posterior complex of the dura mater.

Main Results:

• Achieved a 79.1% first-attempt success rate for spinal anesthesia.
• Scanning durations for key landmarks were 21.0 and 11.0 seconds.
• Demonstrated good correlation (r=0.915) between program and clinician depth measurements.

Conclusions:

• The automated spinal landmark identification program aids needle insertion point identification in obese patients.
• The program shows good correlation with clinician measurements for depth to the dura mater.
• Future research should focus on improving imaging algorithms for needle guidance.