# Using non-Gaussian diffusion models to distinguish benign from malignant head and neck lesions

## TL;DR

This study shows that non-Gaussian diffusion models can help distinguish between benign and malignant head and neck tumors, with one parameter performing particularly well.

## Contribution

The study introduces the use of FROC and CTRW diffusion models for differentiating head and neck lesions with improved diagnostic accuracy.

## Key findings

- Parameters like αCTRW showed the highest diagnostic performance with significant AUC values.
- Several diffusion parameters correlated with Ki-67 expression, indicating their potential to reflect tumor heterogeneity.
- FROC and CTRW models outperformed conventional DWI in distinguishing benign and malignant lesions.

## Abstract

This study aims to investigate the application value of fractional-order calculus (FROC) and continuous-time random-walk (CTRW) derived multiple parameters in distinguishing benign and malignant head and neck lesions and compare their performance with conventional diffusion-weighted imaging (DWI).

A retrospective analysis was conducted on 70 pathologically confirmed cases, including 23 benign lesions (BL) and 47 malignant lesions (ML). ML was further classified into lymphoma subgroups (LS, 11 cases, 15 lesions) and malignant lesions subgroups excluding lymphoma (MLS, 36 cases). DWI scans with 12 b-values were performed before treatment, and seven diffusion parameters—ADC, DFROC, βFROC, μFROC, DCTRW, αCTRW, and βCTRW—were extracted from conventional DWI, FROC, and CTRW diffusion models. Independent t-tests or U-tests were used to compare parameter differences among BL, ML, LS, and MLS. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves, with area under the curve (AUC) compared via DeLong analysis. Pearson correlation analysis was conducted to explore relationships between diffusion parameters and Ki-67 expression in the MLS group.

ADC, DFROC, μFROC, DCTRW, and αCTRW showed significant differences between all groups, αCTRW demonstrated the highest diagnostic performance (AUC). Significant correlations were found between Ki-67 expression and DFROC (r = -0.367, p = 0.028), DCTRW (r = -0.376, p = 0.024), αCTRW (r = -0.418, p = 0.011), and βCTRW (r = 0.525, p = 0.001).

Multiple diffusion parameters derived from FROC and CTRW models effectively differentiate between benign and malignant head and neck lesions, reflecting tumor heterogeneity. Among them, αCTRW showed the best diagnostic performance, making it a promising non-invasive imaging biomarker for quantitative assessment and differential diagnosis of head and neck tumors, thereby improving diagnostic accuracy.