Comparison of prostate volume measured by transabdominal ultrasound and MRI with the radical prostatectomy specimen volume: a retrospective observational study

Accurate measurement of PV was essential in the evaluation and management of prostate disease. For patients with benign prostatic hyperplasia, PV has been a powerful tool for the purpose of assessing symptom severity, predicting complications, and selecting appropriate treatments (medication or surgery) [10]. In terms of PCa, PV has been demonstrated to be effective in screening and risk stratification, especially when combined with PSA.

The best way to assess the accuracy of various volume measurements was to compare the results to the actual volume, which was equivalent to measuring the volume of the prostate specimen after RP. Formalin-fixed pathological specimens have been used for weighing instead of fresh surgical specimens in several studies [11, 12]. On the one hand, formalin fixation leads to tissue dehydration, which reduces the actual in vivo volume. On the other hand, pathological specimens may lead to an overestimation, as seminal vesicles and prostatic fat are usually not removed from the prostate. In this study, the volume of the specimen was measured immediately after RP using water displacement with removal of the periprostatic fat, seminal vesicles and vas deferens, thereby reducing the risk of under- or overestimation of the volume.

Historically, Planimetry-based assessment of PV was considered to be the closest to in vivo prostate size [13]. However, it was time-consuming, cumbersome, required special software, and was therefore not widely used in daily clinical practice. Although less accurate than Planimetry, ultrasonography using the ellipsoid formula has been widely used due to its speed, radiation-free nature, and cost-effectiveness. A systematic review showed a favorable correlation between the PV obtained by TRUS and surgical specimens, ranging from 0.70 to 0.90 [14]. In addition, we have also shown that PV measured by TRUS correlates well with TAUS, MRI, and specimen volume in our TRUS subgroup (Supplementary Table 1). However, TRUS was an invasive imaging modality that could cause discomfort and anxiety, particularly in patients with anal diseases such as hemorrhoid, anal fissure, and anal fistula. In fact, TAUS was typically the preferred choice for patients with lower urinary tract symptoms and was more commonly used to measure prostate dimensions. Furthermore, it was a non-invasive method that was well tolerated by the patient.

Previous studies showed a strong positive correlation and agreement between TAUS and TRUS, indicating that TAUS was an excellent surrogate for TRUS [15, 16]. However, few had explored the relationship between TAUS and fresh RP specimens and their relationship was not well defined. In the present study, TAUS overestimated specimen volume, but they were strongly associated and concordant when using the easily applicable ellipsoid formula. Varkarakis et al. also reported that TAUS overestimated the fresh specimen volume (4.61cm3), but the correlation was not reported [17]. Problems in measuring the longitudinal dimension have been suggested as a possible reason for the inaccuracy of the PV measured by TAUS, especially in larger prostates and when the bladder expansion was over or under full [18].

With the spread and improvement of MRI techniques, its higher spatial resolution, better soft-tissue contrast, and more complex computational capabilities made it superior in contouring the prostate, providing more precise and repeatable PV analysis. PI-RADS v2.1 aimed to standardize PV estimation and recommended routine reporting of PV based on MRI, by manual or automated segmentation or ellipsoid formula [9]. However, manual segmentation should be performed by an experienced radiologist or a trained non-radiologist and this approach was neither time-saving nor cost-effective [19]. ​Some types of automated segmentation have proven to be time efficient for accurate PV measurements, but require much more economy and generality [20]. It is worth mentioning that artificial intelligence is increasingly used in radiology, especially in prostate imaging. Deep learning-based prostate segmentation appears to be superior to traditional segmentation, and relevant studies have examined the feasibility of applying automated segmentation based on deep learning algorithm [21]. Nevertheless, the application of such models is mainly limited to academic research rather than clinical use.

Previous studies investigated the accuracy of the PV measured with the ellipsoid formula on MRI and discovered a high degree of association between the ellipsoid formula and the reference (manual planimetry or prostatectomy specimen). A prospective study included 21 patients who had undergone RP and found that PV measured on MRI using the ellipsoid formula had an excellent correlation coefficient with the volume of fresh RP specimen (r = 0.92) [22]. Bezinque et al. reported an excellent correlation between the PV calculated by the ellipsoid formula and MRI-R3D (manual segmentation by a radiologist) (ICC = 0.90), indicating that MRI using the ellipsoid formula provided an accurate measurement of PV [19]. In conclusion, PV estimation on MRI using the ellipsoid formula was a rapid technique with reasonable accuracy and reproducibility, and its general availability made it feasible for routine clinical use [23]. As in previous studies, our results demonstrated that the specimen volume for the entire cohort was highly associated with and underestimated by the volume measured by MRI [24, 25]. ​.

Several studies reported that the direction and magnitude of the difference was volume dependent. ​Matthews et al. compared PV measured by TRUS to specimen volume from 100 men diagnosed with PCa who underwent radical retropubic prostatectomy and reported that TRUS overestimated specimen volume for volumes less than 30cm3 and increasingly underestimated specimen volume for volumes greater than 30cm3 [26]. A similar study found that MRI appeared to overestimate and underestimate PV when the specimen volume was less than 35 cm3 and greater than 35 cm3, respectively [27]. However, this association was not found in TRUS in their study. Accordingly, the present study also explored whether the difference was volume-dependent. No statistically significant correlation was found between the difference and the specimen volume (P = 0.193) in the TAUS group. In the MRI group, we discovered that the direction and magnitude of the difference varied with specimen volume. In other words, if the specimen volume was < 39 ml, MRI overestimated the specimen volume; if the specimen volume was > 39 ml, MRI underestimated the specimen volume. In summary, MRI had a tendency to overestimate the smaller prostates but underestimate the larger ones.

Currently, many clinical risk-stratified prediction models and nomograms incorporate PV as a key predictor [28,29,30]. Although our data and previous studies demonstrated the superiority of MRI in measuring PV, capacity and resource limitations posed a challenge in delivering prebiopsy MRI for all men with suspected PCa. Therefore, we examined the relationship between TAUS-based and MRI-based PV and confirmed that they were highly associated and concordant, and the linear regression equation was established. ​However, the significance of such a conversion result needs to be further validated.

This retrospective study is not without limitations. First, the reproducibility of the volume measurements obtained could be limited by factors such as the inaccuracy of the inherent limitation of the ellipsoid formula used in this study, which assumed that the prostate had an ellipsoid-like shape that did not exist in fact. The shape of the prostate was highly variable and irregular, so any fixed formula that didn’t take shape into account was prone to error. Second, although all imaging tests were performed within 3 months prior to RP, the PV may have changed during this period due to tumor growth, which would affect the accuracy of the comparison between results. Third, the single-centric retrospective study design is another limitation of this study. Due to the limited sample size of this study, our findings need to be further verified in a well-designed, large-sample prospective study.

留言 (0)

沒有登入
gif