A multiparameter diagnostic model based on MRI volumetric ADC histogram and clinical variables accurately differentiates thymic epithelial tumors from mediastinal lymphomas.

[BACKGROUND] The management and prognosis of each type of anterior mediastinal mass differ substantially. Radical thymectomy is regarded as the preferred surgical approach for resectable thymic epithelial tumors (TETs), whereas chemotherapy is the recommended treatment for mediastinal lymphoma after confirming the histological diagnosis through needle biopsy, and surgical procedures should be avoided. Consequently, an accurate diagnosis of mediastinal lymphoma and TETs holds paramount importance in clinical treatment and prognosis for patients with thymic neoplasms.

[METHODS] Patients of TETs and mediastinal lymphomas with histopathological proof were included in the present study. The ADC histogram parameters were extracted from ADC maps. Clinical characteristics, radiological features and ADC histogram metrics (including ADCmin, ADCmax, and ADCmean; 5th, 10th, 25th, 50th, 75th, 90th and 95th percentiles of ADC values; skewness and kurtosis) were evaluated between two groups. Multivariate logistic regression was used to build a comprehensive diagnostic model. Receiver operator characteristics (ROC) curve analysis was subsequently carried out to evaluate diagnostic performance. A nomogram was developed to differentiate TETs and mediastinal lymphomas.

[RESULTS] A cohort of 130 consecutive patients, comprising 93 individuals with TETs and 37 with mediastinal lymphomas, was enrolled in the study. TETs comprised 57 low-risk thymomas (61.3%), 20 high-risk thymomas (21.5%), and 16 thymic carcinomas (17.2%); mediastinal lymphomas comprised 13 Hodgkin lymphoma (HL) (35.1%) and 24 non-Hodgkin lymphoma (NHL) (64.9%). It was found that patients with mediastinal lymphomas were significantly younger compared to those with TETs (38.11 ± 13.51 years vs. 53.66 ± 12.99 years,  < 0.001). The rate of serum lactate dehydrogenase (LDH) elevation was markedly higher in the lymphoma group (54.1% vs. 2.2%,  < 0.001). The maximal diameter of lesions and skewness were significantly larger in patients with mediastinal lymphoma, whereas the 25th -95th percentile of ADC values, ADCmax and ADCmean were significantly lower compared to patients with TETs (all  < 0.05). ADC histogram parameters did not differ among TET subtypes (all  > 0.05), whereas NHL had lower 10th -95th percentile of ADC values and ADCmean than HL (all  < 0.05). The comprehensive diagnostic model was established based on forward stepwise regression, including age, serum LDH level and skewness, with higher AUC than skewness alone (0.914, 95%CI: 0.850–0.977 vs. 0.785, 95%CI: 0.701–0.869,  < 0.01). The predictive C-index nomogram performance was 0.917 (95%CI: 0.915–0.918).

[CONCLUSION] The comprehensive diagnostic model, integrating ADC histogram parameters and clinical characteristics, demonstrated significant potential in distinguishing between TETs and mediastinal lymphomas.

[SUPPLEMENTARY INFORMATION] The online version contains supplementary material available at 10.1186/s12880-025-02147-5.