In recent years, with the development of society and the westernization of dietary patterns in China, the incidence and mortality rates of PCa have been increasing year by year[1]. The clinical characteristics of prostate cancer (PCa) vary from person to person. Some patients have indolent growth of PCa, which does not significantly impact their life and health, while for others it can be fatal. Moreover, different patients show varying sensitivities to treatment options. The main reason for this phenomenon is the heterogeneity of the tumor, which refers to the genetic and molecular changes that occur during tumor growth and affect its growth rate, invasiveness, drug sensitivity, and prognosis[15]. DCE-MRI can help detect and diagnose tumor by highlighting the differences in distribution between tumor tissue and normal tissue using a contrast agent, which has been widely used in the clinical diagnosis and treatment of various cancers, including gliomas, breast tumors and renal tumors[16–18].
In this study, the quantitative parameters Ktrans and Kep of PCa (average, 50th, 25th, 75th) were found to be higher than those of the non-tumor group, which is consistent with some previous research findings[19, 20]. This is mainly due to the reliance of cancer tissue on the generation of neovascularization, which leads to more robust neovascular tissue with larger interstitial gaps and increased permeability[21]. Although there was a significant difference in Ve 25th, the AUC and specificity of Ve 25th were relatively low, indicating its limited diagnostic value for PCa. Therefore, we believe that the quantitative analysis of Ktrans and Kep has certain diagnostic value for detecting PCa, which is consistent with the findings of other studies[20, 22, 23]. Furthermore, the results of this study showed that Ktrans had higher AUC, sensitivity and specificity compared to Kep parameters. Among them, Ktrans 25th had the highest diagnostic value, with an AUC of 0.839 and specificity and sensitivity of 81.2% and 80.2% respectively. This indicates that when the parameter value of 25% of the pixels within the entire tumor is below this threshold, it reflects tumor heterogeneity and demonstrates better diagnostic performance than the average value. There was a certain degree of overlap in the parameters between the two groups, which may be attributed to the older age of the patients included in this study, as most of them exhibited signs of benign prostatic hyperplasia. Both PCa and benign prostatic hyperplasia can lead to increased microvascular density, and the discontinuity of the vascular basement membrane in the lesion may contribute to increased vascular permeability.
The widespread application of the PI-RADS system requires good diagnostic consistency. PI-RADS v2.1, based on the 2.0 scoring system, further refines and modifies the system, reducing ambiguous descriptions and overlapping criteria, thus improving diagnostic consistency. According to PI-RADS v2.1, patients with scores of 4–5 should be considered for biopsy, while patients with scores of 1–2 are recommended for follow-up. In this study, the detection rates of PCa for scores 1, 2, 4 and 5 were 0%, 9.5%, 68.3% and 94.1% respectively. These findings are highly consistent with the guidelines and can effectively avoid overtreatment. The detection of cancer lesions by PI-RADS is dependent on their volume[24, 25]. When the lesion volume is between 0.5-1.0 cm3, the detection rates of PCa are increased by 1.4 times. When the lesion volume is ≥ 1.0 cm3, the detection rates are increased by 1.8 times. The management of lesions with a score of 3 is currently a subject of significant controversy. In this study, the detection rates of PCa for PI-RADS v2.1 score 3 were 36.0%. It is recommended to consider other clinical indicators and actively perform biopsy when necessary, which is consistent with the viewpoint of Tsai[26].
Quantitative histogram analysis and PI-RADS scoring have their respective advantages and limitations in the diagnosis of PCa. When used alone, they do not show a higher diagnostic value. However, when combined in a diagnostic model, the diagnostic performance significantly improved with AUC values greater than 0.9. The reason for this improvement may be that PI-RADS scoring provides a more comprehensive and accurate assessment of factors such as involvement of the prostatic capsule, involvement of adjacent tissues, and pelvic lymph node metastasis. On the other hand, quantitative histogram analysis is more sensitive in evaluating small lesions and tumor heterogeneity. Combining the strengths of both approaches can significantly enhance the detection of PCa and provide more reliable evidence for clinical physicians.
Different pathological grades of PCa exhibit differences in treatment approaches and prognosis. Therefore, it is particularly important to assess tumor invasiveness. The Gleason grading system is the most widely used system internationally and has undergone continuous updates and improvements since its inception. It provides crucial guidance for clinicians in selecting treatment plans and evaluating prognosis[27, 28]. The results of this study demonstrate that the establishment of a combined diagnostic model using quantitative histogram parameters from DCE-MRI and PI-RADS v2.1 has better diagnostic value in distinguishing between Gleason 6 group and Gleason ≥ 7 group, Gleason ≤ 3 + 4 group and Gleason ≥ 4 + 3 group, and Gleason ≤ 4 + 3 group and Gleason ≥ 8 group. The AUC values of the combined model are higher than those of any single parameter, and both sensitivity and specificity in the differential diagnosis are improved to some extent. Although the AUC value of the combined diagnosis between Gleason ≤ 8 group and Gleason ≥ 9 group does not show a significant difference compared to the use of individual parameters alone, the AUC of the combined diagnosis is still slightly higher than the diagnostic performance of individual parameters. The combined application of both approaches as a guide for biopsy can help differentiate between low-risk and intermediate-to-high-risk PCa, thereby reducing unnecessary biopsy, overdiagnosis and overtreatment. This provides clinicians with more reference criteria for the diagnosis and treatment of PCa.
This study has certain limitations. Some of the pathological results were obtained from biopsy, which may lead to missed diagnosis of small lesions. Manual delineation of regions of interest (ROI) may introduce some errors and biases. However, in this study, the selection of ROI was based on abnormal signals observed in MRI and combined with the lesion location determined by biopsy or gross specimens, ensuring a one-to-one correspondence between ROI placement and pathological regions. The PI-RADS score is influenced by various factors, such as differences in equipment, scanning parameters and individual variations[29]. Given the complexity of medical imaging, the integration of physician experience with artificial intelligence (AI) is a current research focus. In the future, it is expected that multicenter, large-sample experiments combining physician experience with AI will be conducted to overcome these limitations.