A deep-learning method using computed tomography scout images for estimating patient body weight.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate patient body weight is crucial for determining contrast medium dose, drug dosing, and radiation dose management.
• Estimating patient body weight accurately and efficiently before computed tomography (CT) scans, especially in emergency settings, remains a challenge.
• Current methods for time-efficient body weight estimation before diagnostic CT scans lack high accuracy.

Purpose of the Study:

• To develop and evaluate deep-learning models for automatic body weight prediction from CT scout images.
• To assess the accuracy and clinical utility of AI-based body weight estimation in adult patients.
• To provide a solution for managing contrast medium and radiation doses in cases of unknown patient body weight.

Main Methods:

• Development of deep-learning models using a dataset of 1831 chest and 519 abdominal CT scout images with corresponding body weights.
• Evaluation of model performance through correlation analysis and mean absolute error calculations.
• Validation of the models' ability to predict body weight from CT scout images.

Main Results:

• Strong correlations were observed between actual and predicted body weights for both chest (ρ = 0.947, p < 0.001) and abdominal datasets (ρ = 0.869, p < 0.001).
• Mean absolute errors in body weight prediction were 2.75 kg for the chest dataset and 4.77 kg for the abdominal dataset.
• The deep learning models demonstrated clinically acceptable accuracy in estimating body weights.

Conclusions:

• Deep learning models can accurately estimate patient body weight from CT scout images.
• This AI-driven method offers a clinically useful tool for managing contrast medium and radiation doses in adult patients with unknown body weights.
• The proposed approach has the potential to improve patient safety and optimize medical imaging protocols.