Analysis of Convolutional Neural Network Segmentation Algorithm of Adenomyoma

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Gynecology

Background:

• Adenomyoma is a prevalent uterine condition significantly impacting women's health.
• Ultrasound is the primary, cost-effective imaging modality for gynecological disease diagnosis and screening.
• Interpreting ultrasound images for adenomyoma requires significant clinical expertise, posing challenges for diagnosis and increasing physician workload.

Purpose of the Study:

• To investigate deep learning-based segmentation methods for uterine adenomyoma in ultrasound images.
• To evaluate the effectiveness of the Deeplab network for this specific application.
• To address the need for automated tools in gynecological ultrasound image analysis.

Main Methods:

• Utilized convolutional neural networks (CNNs), specifically the Deeplab network.
• Compared the performance of Deeplab against the Fully Convolutional Network (FCN) for image segmentation.
• Focused on segmenting uterine adenomyoma from ultrasound images.

Main Results:

• The Deeplab network demonstrated superior performance as an image segmentation model for uterine adenomyoma compared to the FCN network.
• The study identified Deeplab as a suitable deep learning model for this task.

Conclusions:

• Deep learning, particularly the Deeplab network, offers a viable solution for uterine adenomyoma ultrasound image segmentation.
• Implementing this technology can potentially reduce the diagnostic burden on clinicians.
• This research aims to fill a gap in automated analysis for uterine adenomyoma ultrasound imaging.