Classification of ovary abnormality using the probabilistic neural network (PNN)

Area of Science:

• Medical imaging
• Artificial intelligence in healthcare
• Reproductive endocrinology

Background:

• Lifestyle changes contribute to obesity and hyperinsulinemia, impacting female fertility.
• Accurate fertility status evaluation is crucial for women facing reproductive challenges.
• Ultrasonography offers a safe, noninvasive method for assessing fertility potential.

Purpose of the Study:

• To develop an automated ovarian classification system using ultrasonographic images.
• To improve upon manual follicle counting and area measurement, reducing errors.
• To correlate automated ovarian biomarkers with hormone levels like androgen, testosterone, and luteinizing hormone.

Main Methods:

• Image segmentation using active contour with split-Bregman optimization.
• Feature extraction via geometric and intensity methods.
• Feature selection using particle swarm optimization and principal component analysis, followed by probabilistic neural network classification.

Main Results:

• The proposed probabilistic neural network achieved 97% classification efficiency.
• This surpasses Support Vector Machine (92%) and Radial Basis Function (88%) methods.
• High classification rates were achieved using extensive feature selection.

Conclusions:

• Automated ovarian classification shows high accuracy in fertility assessment.
• Combining numerous features with selection methods enhances classification performance.
• This approach offers a more efficient and reliable tool for evaluating female fertility.