Relationship of T1, T2* durations and ADC values with tumor type and histopathological tumor degree in solid kidney tumors

Background This study aims to evaluate whether MRI parameters (pre-contrast T1, T2*, R2*, post-contrast T1, and ADC values) can accurately differentiate tumor type and grade in solid kidney tumors. Additionally, it seeks to develop a prediction tool using the Random Forest machine learning algorithm for tumor detection based on these imaging results. Methods Contrast-enhanced abdominal MRIs of 82 patients who underwent surgery for solid renal masses at our hospital (July 2019–January 2024) were analyzed retrospectively. Measurements were taken from tumor regions and the contralateral healthy kidney cortex using three ROIs (~ 15 mm²). Parameters included T2*, native and post-contrast T1, R2* values (1/T2*), and ADC values. Interobserver agreement was assessed by a second independent researcher. Results The study included 82 patients (32 women, 50 men), with a mean age of 59.3 years. The average tumor size was 60.7 mm. The tumors were categorized into eight subtypes, with clear cell carcinoma being the most common (n = 46), followed by papillary carcinoma (n = 8) and chromophobe carcinoma (n = 8). WHO grading classified tumors as severe (grades 3–4) or non-severe (grades 1–2). Severe tumors had significantly lower ADC values (p = 0.029) and larger sizes (p = 0.0017). Significant differences in T2 mDixon Quant, R2*, and ADC values were found across subtypes (p < 0.05). The Random Forest model effectively recognized tumor tissue. Conclusion MRI parameters, particularly ADC and R2*, effectively differentiate tumor type, grade, and tissue origin. The AI model showed high accuracy, warranting further validation with larger datasets.