FibroidX: Vision Transformer-Powered Prognosis and Recurrence Prediction for Uterine Fibroids Using Ultrasound...

Area of Science:

• Gynecological imaging analysis
• Artificial intelligence in healthcare
• Machine learning for disease prediction

Background:

• Uterine fibroids (UFs) significantly impact women's reproductive health and quality of life.
• Accurate prognosis and recurrence prediction of UFs are crucial for personalized treatment and reducing long-term consequences.
• Conventional prediction methods using imaging and statistical models often lack accuracy and objectivity.

Purpose of the Study:

• To introduce FibroidX, an AI-driven system for enhanced uterine fibroid prognosis and recurrence prediction.
• To overcome limitations of conventional methods by automating feature extraction and providing customized risk evaluations.
• To improve the accuracy and reliability of predicting UF disease progression, symptom severity, treatment response, and post-treatment regrowth.

Main Methods:

• Utilized vision transformers and self-attention mechanisms within the FibroidX model.
• Trained the model on a dataset of 1990 ultrasound images, split into 80% training and 20% testing sets.
• Evaluated model performance using metrics including accuracy, precision, recall, F1-score, and AUC-ROC.

Main Results:

• FibroidX achieved a high accuracy of 98.4%, outperforming baseline models (92.3% and 94.1%).
• Demonstrated strong performance with precision (97.8%), recall (96.9%), and an F1-score of 97.3%.
• Achieved an AUC-ROC score of 0.99, indicating excellent class distinction.

Conclusions:

• FibroidX is effective and reliable for UF prediction tasks, suitable for real-time applications with an average inference time of 0.02s.
• The AI model demonstrated a 15% increase in accuracy and a 12% reduction in false positive rates compared to traditional machine learning techniques.
• FibroidX offers a significant advancement in personalized risk assessment for uterine fibroids.