Automatic volumetry on MR brain images can support diagnostic decision making

Area of Science:

• Medical Imaging
• Neuroscience
• Radiology

Background:

• Current diagnostic decisions in clinical imaging rely heavily on visual interpretation, which can lead to diagnostic uncertainty, particularly in differentiating dementia from normal aging.
• Magnetic Resonance (MR) brain imaging offers potential for quantitative analysis to improve diagnostic accuracy.

Purpose of the Study:

• To investigate if extracting volumetric data from MR brain images and presenting it as a color overlay can assist diagnostic readers in classifying dementia versus normal aging.
• To evaluate the impact of this visual aid on diagnostic confidence.

Main Methods:

• A proof-of-concept reader study using MR brain images from 36 subjects (7 with Alzheimer's Disease (AD), 29 controls).
• Automatic segmentation of MR images into 43 regions using atlas registration and label propagation.
• Readers reviewed grayscale MR images, then had the option to view a false-color overlay representing volumetric size rank compared to a reference group.

Main Results:

• The color overlay was found to be useful in 18 out of 28 diagnoses, impacting reader confidence.
• The impact of the additional volumetric information on diagnostic confidence was statistically significant (p < 0.02).

Conclusions:

• Volumetric anatomical information extracted from brain images using automatic segmentation and presented as color overlays can support diagnostic decision-making.
• This approach shows promise for improving the accuracy and confidence in diagnosing conditions like dementia.