Improving Diagnostic Performance for Head and Neck Tumors with Simple Diffusion Kurtosis Imaging and Machine Learning...

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Mean kurtosis (MK) from simple diffusion kurtosis imaging (SDI) aids head and neck malignancy diagnosis.
• Smoothing filters can enhance diagnostic accuracy in SDI.
• Multi-parameter analysis with diffusion kurtosis imaging (DKI) shows improved diagnostic performance.

Purpose of the Study:

• Evaluate machine learning (ML)-based multi-parameter analysis of MK and apparent diffusion coefficient (ADC) values from SDI.
• Determine the utility of these parameters for differentiating benign and malignant head and neck tumors.
• Assess the impact of filter pre-processing on diagnostic performance.

Main Methods:

• Collected SDI data from 32 pathologically diagnosed head and neck tumors.
• Applied Gaussian filter for image pre-processing.
• Extracted MK and ADC values from tumor regions for ML model input.
• Utilized five ML algorithms and ROC analysis for model evaluation.

Main Results:

• Bi-parameter analysis combining MK and ADC values demonstrated strong diagnostic performance.
• Gradient boosting algorithm achieved the highest diagnostic performance.
• Area Under the Curve (AUC) reached 0.81 in the best-performing model.

Conclusions:

• ML-based bi-parameter analysis using SDI data is effective for differential diagnosis of head and neck tumors.
• This approach aids in determining appropriate treatment strategies.
• Filter pre-processing potentially enhances diagnostic accuracy in SDI analysis.