Advanced AI-Powered System for Comprehensive Thyroid Cancer Detection and Malignancy Risk Assessment.

Area of Science:

• Oncology
• Medical Imaging
• Bioinformatics

Background:

• Thyroid cancer incidence is increasing globally.
• Accurate diagnosis is crucial to avoid unnecessary surgical interventions.
• Computer-aided diagnosis (CAD) systems offer potential for improved accuracy.

Purpose of the Study:

• To introduce a novel hybrid CAD system for comprehensive thyroid cancer diagnostics.
• To combine convolutional neural networks (CNNs) for image analysis and molecular data for risk prediction.
• To enhance diagnostic reliability and support clinical decision-making.

Main Methods:

• A two-module hybrid system was developed: a CNN model for ultrasound image analysis and a molecular data analysis module for malignancy risk prediction.
• Transfer learning and various image augmentation techniques were employed for the CNN module.
• The system was evaluated for its diagnostic performance and accuracy in classifying thyroid nodules and predicting malignancy.

Main Results:

• The CNN module achieved 93.65% accuracy, 100% sensitivity, and 69.23% specificity in classifying thyroid nodules from ultrasound images.
• The molecular data analysis module demonstrated strong performance with low Mean Squared Error (MSE) values (training: 4.24 × 10^-5, testing: 6.31 × 10^-3).
• The hybrid system offers decoupled or combined usage for in-depth thyroid cancer diagnosis.

Conclusions:

• The proposed hybrid CAD system shows competitive performance compared to existing systems.
• This AI-driven approach provides reliable diagnostic insights for thyroid cancer management.
• The system supports clinicians in making informed decisions, potentially improving patient outcomes by avoiding unnecessary surgeries.