Automated Uterine Fibroids Detection in Ultrasound Images Using Deep Convolutional Neural Networks.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Deep Learning for Diagnostics

Background:

• Uterine fibroids (UF) are common benign tumors in women of childbearing age.
• Early diagnosis of UF is crucial for effective treatment.
• Automated diagnosis using deep learning (DL) shows potential for improving UF detection accuracy.

Purpose of the Study:

• To evaluate state-of-the-art DL architectures (VGG16, ResNet50, InceptionV3) and a proposed DPCNN for UF detection.
• To assess the impact of fine-tuning strategies on pre-trained DL models for UF diagnosis.
• To establish a foundation for advanced computer-aided diagnosis systems for UF.

Main Methods:

• Utilized an ultrasound image dataset from Kaggle, applying scaling, normalization, and data augmentation for preprocessing.
• Trained and validated VGG16, ResNet50, InceptionV3, and a novel DPCNN architecture.
• Evaluated model performance using various metrics, including accuracy.

Main Results:

• The proposed DPCNN architecture achieved the highest accuracy at 99.8%.
• Fine-tuned InceptionV3 and ResNet50 models demonstrated strong performance with 90% and 89% accuracy, respectively.
• The VGG16 model achieved 85% accuracy, indicating varying performance among DL architectures.

Conclusions:

• Deep learning-based methods are effective for automated UF detection from medical images.
• The DPCNN architecture offers superior performance for UF detection compared to existing DL models.
• Further research in DL for medical imaging analysis can lead to enhanced computer-aided diagnosis systems for UF.