Liu Guiqin, Lu Yang, Dai Yongming, Xue Ke, Yi Yangyang, Xu Jianrong, Wu Dongmei, Wu Guangyu
Department of Radiology, Renji Hospital, School of Medicine, Shanghai Jiaotong University, 160 Pujian Road, Shanghai, 200127, China.
United Imaging Healthcare, Shanghai, China.
Abdom Radiol (NY). 2021 Jun;46(6):2740-2750. doi: 10.1007/s00261-020-02903-x. Epub 2021 Jan 3.
To compare various models of diffusion-weighted imaging including mono-exponential, bi-exponential, diffusion kurtosis (DK) and fractional-order calculus (FROC) models in diagnosing prostate cancer (PCa) in transition zone (TZ) and distinguish the high-grade PCa [Gleason score (GS) ≥ 7] lesions from the total of low-grade PCa (GS ≤ 6) lesions and benign prostatic hyperplasia (BPH) in TZ.
80 Patients with 103 lesions were included in this study. Nine metrics [including apparent diffusion coefficient (ADC) derived from mono-exponential model, slow diffusion coefficient (D), fast diffusion coefficient (D),, and f (the fraction of fast diffusion) from bi-exponential model; mean diffusivity (MD) and mean kurtosis (MK) from DK model; diffusion coefficient (D), fractional-order derivative in space (β), and spatial metric (μ) from FROC model] were calculated. Comparisons between BPH and PCa lesions as well as between clinically significant PCa (CsPCa) (GS ≥ 7, n = 31) and clinically insignificant lesions (Cins) (GS ≤ 6 and BPH, n = 72) of these metrics were conducted. Mann-Whitney U-test and receiver operating characteristic (ROC) analysis were used for statistical evaluations.
The areas under the ROC curve (AUC) values of β derived from FROC model were 0.778 and 0.853 in differentiating PCa from BPH and in differentiating CS (GS ≥ 7) from Cins (GS ≤ 6 and BPH), both were the highest compared to other metrics. The AUC value of β was significantly higher than that of ADC (P = 0.009) in differentiating CS from Cins, while the differentiation between BPH and PCa did not reach the statistical significance when comparing with ADC (P = 0.089).
Although no significant difference was found in distinguishing PCa from BPH, the metric β derived from FROC model was superior to other diffusion metrics in differentiation between CS and Cins in TZ.
比较多种扩散加权成像模型，包括单指数模型、双指数模型、扩散峰度（DK）模型和分数阶微积分（FROC）模型，以诊断移行区（TZ）前列腺癌（PCa），并区分高级别PCa（ Gleason评分（GS）≥7）病变与TZ中低级别PCa（GS≤6）病变及良性前列腺增生（BPH）的总和。
本研究纳入80例患者的103个病变。计算了九个指标[包括单指数模型得出的表观扩散系数（ADC）；双指数模型的慢扩散系数（D）、快扩散系数（D）和f（快扩散分数）；DK模型的平均扩散率（MD）和平均峰度（MK）；FROC模型的扩散系数（D）、空间分数阶导数（β）和空间指标（μ）]。对这些指标在BPH与PCa病变之间以及临床显著性PCa（CsPCa）（GS≥7，n = 31）与临床非显著性病变（Cins）（GS≤6和BPH，n = 72）之间进行了比较。采用Mann-Whitney U检验和受试者工作特征（ROC）分析进行统计学评估。
FROC模型得出的β在区分PCa与BPH以及区分CS（GS≥7）与Cins（GS≤6和BPH）时，ROC曲线下面积（AUC）值分别为0.778和0.853，均高于其他指标。在区分CS与Cins时，β的AUC值显著高于ADC（P = 0.009），而在与ADC比较时，BPH与PCa之间的区分未达到统计学显著性（P = 0.089）。
尽管在区分PCa与BPH方面未发现显著差异，但FROC模型得出的指标β在TZ中区分CS与Cins方面优于其他扩散指标。
