Automatic abdominal fat assessment in obese mice using a segmental shape model

Area of Science:

• Medical imaging
• Computational biology
• Obesity research

Background:

• Obesity is a complex metabolic disorder.
• Accurate assessment of abdominal fat distribution is crucial for understanding obesity-related health risks.
• Current methods for fat quantification can be labor-intensive.

Purpose of the Study:

• To develop an automated image analysis technique for quantifying abdominal fat in obese (ob/ob) mice.
• To adapt the method for high-resolution 7 Tesla (7T) magnetic resonance imaging (MRI).

Main Methods:

• A novel segmental shape model was developed to differentiate visceral adipose tissue (VAT) from subcutaneous adipose tissue (SAT).
• The model utilizes shape and distance constraints to segment fat tissues.
• Adaptive fuzzy C-means clustering was employed to handle image intensity variations.

Main Results:

• The automated method demonstrated high accuracy, with correlation coefficients for total adipose tissue (TAT), SAT, and VAT sizes ranging from 0.907 to 0.950.
• Dice coefficients for the positional accuracy of TAT, SAT, and VAT were between 0.920 and 0.941.
• Validation was performed against manual segmentations on 109 MRI slices from 7 obese mice.

Conclusions:

• The developed computerized image analysis method provides reliable quantification and distribution assessment of abdominal fat tissues.
• The automated approach shows strong agreement with manual segmentations.
• This method has the potential to enhance laboratory automation in obesity research.