The diagnostic performance of Magnetic resonance imaging-derived apparent diffusion coefficient histogram analysis in distinguishing between benign and malignant parotid gland tumors

Abstract Background Parotid gland tumors constitute the majority of salivary gland neoplasms, with a predominance of benign lesions. Malignant counterparts, although less common, necessitate more aggressive treatment strategies. Differentiating between benign and malignant tumors based on clinical findings alone is challenging. Magnetic Resonance Imaging (MRI), particularly when combined with diffusion-weighted imaging and apparent diffusion coefficient (ADC) analysis, provides valuable insights into lesion characteristics. Histogram analysis of ADC maps may enhance diagnostic precision by capturing the distribution of diffusion values within tumors. We aimed to assess the diagnostic accuracy of the MRI and ADC histogram analysis and the inter-observer agreement in differentiating benign and malignant parotid gland masses. Results This cross-sectional retrospective study involved 100 patients who underwent MRI examination and had histopathologically proven parotid tumors. Conventional and diffusion-weighted MRI sequences were analyzed by two radiologists. ADC values and histogram metrics were compared between benign and malignant lesions, and inter-observer agreement was evaluated. Among 100 patients (57% males), 59% had benign and 41% had malignant parotid tumors. Most lesions were unilateral (80%) and located in the superior lobe (57%). Malignancy was associated with solid composition ( P = 0.027), hypointense T2 ( P = 0.011), ill-defined margins ( P = 0.010), high diffusion ( P < 0.001), lymphadenopathy ( P = 0.001), skin infiltration ( P = 0.023), and vascular invasion ( P = 0.006). MRI feature assessment showed excellent agreement between observers, with κ values ranging from 0.820 to 1.000 ( P < 0.001), confirming high reliability in interpreting composition, signals, margins, enhancement, and diffusion. Malignant lesions had significantly lower ADC values (median = 0.9 vs 1.4–1.5; P < 0.001) and higher skewness ( P = 0.008). Lesions’ ADC values showed near-perfect inter-observer agreement (ICC = 0.959, P < 0.001). Using ADC < 1.15 × 10⁻ 3 mm 2 /s, observer 1 achieved 80.6% accuracy (AUC = 0.847, P < 0.001), and observer 2 reached 78.6% accuracy (AUC = 0.857, P < 0.001), with high sensitivity and specificity in both. Mean ADC had the best performance (AUC = 0.793, accuracy = 74.3%, P < 0.001). Minimum ADC showed 64.3% accuracy (AUC = 0.720), and Maximum ADC had the highest sensitivity (80.5%) but lower accuracy (67.2%, AUC = 0.712, P < 0.001). Conclusions MRI combined with ADC histogram analysis demonstrates high diagnostic performance and excellent inter-observer agreement in differentiating benign and malignant parotid gland tumors. This approach can enhance preoperative assessment and guide clinical decision-making.