Thyroid segmentation and volume estimation in ultrasound images

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Thyroidology

Background:

• Thyroid gland volume measurement is crucial for diagnosing pathologies.
• Current methods rely on computed tomography (CT) scans, which are costly, time-consuming, and involve radiation exposure.
• Ultrasound (US) imaging is a safer, more affordable alternative but lacks precise volumetric measurement capabilities.

Purpose of the Study:

• To develop and validate a comprehensive solution for estimating thyroid gland volume directly from ultrasound images.
• To provide a non-invasive, cost-effective, and accurate method for thyroid volume assessment.

Main Methods:

• Utilized a radial basis function neural network for thyroid gland block classification.
• Employed a region-growing method to identify the integral region of interest.
• Applied a particle swarm optimization algorithm for accurate parameter estimation of thyroid volume.

Main Results:

• The proposed approach demonstrated successful segmentation of the thyroid region in US images.
• Accurate estimation of thyroid gland volume was achieved directly from US imaging data.
• The method shows significant promise for clinical application.

Conclusions:

• The developed technique offers a reliable and efficient method for thyroid volume estimation using ultrasound.
• This approach can potentially replace or supplement CT scans, reducing patient radiation exposure and healthcare costs.
• The findings support the integration of advanced imaging analysis techniques in routine thyroid diagnostics.