Advantage of whole-mount histopathology in prostate cancer: current applications and future prospects

Magnetic Resonance Imaging: By registering with WMH, some features on Magnetic Resonance images exhibit discernible association with PCa characteristic and distribution, therefore can be utilized in PCa detection and classification. Studies have demonstrated the efficacy of Prostate Imaging Reporting and Data System (PI-RADS) in PCa diagnosis with WMH registration, with most achieving significant accuracy [6‐8]. There are also correlations between WMH-obtained tumor feature and specific indicator in Magnetic Resonance Imaging (MRI) sequences. Lower apparent diffusion coefficient (ADC) value exhibits a correlation with higher cancerous degree, thus indicating the ability to identify tissue malignancy [9‐11]. For dynamic contrast enhanced (DCE) images, malignant region correlates with early enhancement [10], and time-to-peak (TTP) outperformed ADC in low-grade tumor detection [9]. There are also conflicts between studies about the performance of ADC in different regions that needs further considerations [11, 12].

Positron emission tomography: Co-registration of Positron emission tomography (PET) image and WMH is essential to provide verification standard for PET tracers. With the WMH serving as the ground truth, studies have investigated the efficacy of ⁶⁸Ga-PSMA [16‐19], ⁶⁸Ga-RM2 [20] and ¹⁸F-DCFPyL PSMA [21] in PCa diagnosis. Since the co-registered WMH images can provide a precise histopathologic standard for PET verification, the utilization of MRI as a verification standard for PET becomes unnecessary, thereby enabling researchers to directly compare the efficacy of PET with MRI. Two study included reported similar performance between PET/CT and MRI [16, 20], while three confirmed better performance of PET/CT than MRI [17‐19] in PCa detection. It’s worth noting that the performance of PET/CT combined with MRI exhibited significantly higher detection rate of cancerous region comparing to solely MRI [16, 18, 20], indicating that PET may improves the definition of tumor extent. Considering the extensive development and exploration of MRI features [22, 23], PET can serve as an important complement to MRI, which implicates that the integration of PET and MRI have the potential to improve the current clinical guidelines. More details of the above studies are available in Table 1. Further investigations can be conducted to improve PET imaging, with WMH serving as the gold standard for histopathologic validation.

Ultrasound: Ultrasound (US) has played an important role in prostatic targeted-biopsy, and researchers are now striving to enhance its diagnostic efficacy for PCa. In our included studies, both sector-based registration [24, 25] and complete spatial registration [6] were implemented. With the help of WMH, the combination of Artificial Neuron Network and basic transrectal US could achieve an accuracy of 78% [24]. Besides basic US, Micro-ultrasound (Micro-US) [25] and ultrasound shear wave elastography (USWE) [6] have also shown satisfactory performance with WMH serving as ground truth for validation. These studies indicated the potential for improving US performance with WMH.

Efficacy of AI radiologists: The usage of DL largely improved the accuracy of AI radiologist in detection [26, 27] and classification [28, 29] of clinically significant PCa (csPCa). These models have considerably enhanced diagnostic accuracy for both junior and senior radiologists [26]. With the WMH images as ground truth, DL models achieved satisfying classification performance, which reached AUC of 0.75 [28] in detecting csPCa. Notably, one model detected 18% of lesions ignored by radiologists [28], which demonstrated the importance of WMH in establishing AI radiologist models.

Robustness of AI radiologists: Gunashekar et al. [30] demonstrated that the DL models focus on tumor lesions on WMH ground truth using Class Activation Map. Study by Taimen et al. [31] demonstrated the consistent repeatability of U-Net based DL models in csPCa detection on ADC maps with WMH for validation. Moreover, the ultimate efficacy of DL models is also affected by the labeling strategy employed [32], suggesting that WMH can potentially offer a standardized labeling framework to enhance the robustness of AI radiologists in the future.

GTV calculation with MRI: With WMH as reference standard of tumor volume, MRI demonstrated a clear tendency to significantly underestimate the tumor volume [33, 34], especially in cases with smaller size and low Gleason scores [35]. Among all the sequences in MRI, DCE showed best performance in volume estimation tasks, while the underestimation remains significant comparing to WMH-generated ground-truth [34]. Despite the persistent issue of tumor volume underestimation on MRI, Kramer et al. [36] developed a potential solution by using WMH assistance to expand the margin of MRI-delineated tumors by 2 mm, which has shown promising results in improving volume estimation accuracy while avoiding overestimation concerns.

GTV calculation with PET: In addition to MRI, PET is also a potent technique for GTV estimation for its excellent sensitivity. In a study by Kranz-Rudolph et al. [33], ⁶⁸Ga-PSMA-11-PET/CT significantly outperformed MRI in detection specificity, and overcame the underestimation problem at the same time. Moreover, in a study by Kostyszyn et al. [37], a convolutional neural network (CNN) algorithm was utilized and demonstrated strong agreement in GTV estimation between ⁶⁸Ga-PSMA and ¹⁸F-PSMA PET images, along with high sensitivities and specificities. With the assistance of WMH, these studies exhibited the superiority of PET in GTV calculation, indicating its substantial value in guiding precise treatment planning in the future.

Neoadjuvant therapy: With WMH providing detailed assessment of pathologic treatment response, Golan et al. [38] evaluated and reported the satisfying outcome in safety and efficacy of lutetium-177 prostate-specific membrane antigen, a radioactive agent in neoadjuvant therapy. WMH can also be employed for investigating the effectiveness of medical imaging in therapeutic evaluation. MRI, for instance, has demonstrated satisfying efficacy in post-treatment residual tumor estimation [39] with the validation of WMH images. Chen et al. [40] also demonstrated a better performance of ⁶⁸Ga-PSMA-11 PET/CT than prostate specific antigen (PSA) response in evaluating pathological response to neoadjuvant therapy using WMH as the reference standard. These studies collectively highlight the potential of WMH in facilitating the assessment following neoadjuvant therapy.

Focal therapy: Takeda et al. [41] validated the efficacy of the focal therapy inclusion criteria, who retrospective reviewed the WMH sections from all 98 patients meeting the criteria, and extensive PCa was detected in 39 patients, thus highlighting the current limitations of these inclusion. Moreover, Sood et al. [42] developed a novel precision prostatectomy approach, and tested its inclusion criteria through retrospective examination of WMH sections obtained from radical prostatectomy (RP), in which they proved the feasibility of this precision prostatectomy protocol. With the aid of WMH, more reliable treatment planning for focal therapy can be developed, thereby reducing its risk of recurrence and improving prognostic life quality.

AI pathologist: AI pathologists were designed to identify pathologic features directly on histopathologic images. According to a study conducted by Yamamoto et al. [43], a DL-based AI pathologist exhibited superior predictive capabilities compared to human pathologists in recurrence prognosis. This highlights the enormous potential of AI in extracting valuable information from WMH images. Furthermore, studies have found tissue component maps (TCMs) derived from WMH can improve the efficiency of AI pathologists in PCa detection [44] and classification [45]. The above studies indicated that AI pathologist have the potential to identify pathologic features from WMH images automatically in the future, which can facilitate pathological examination and provide gold standard for validation purpose.

AI radiologist: Other than employing AI pathologists for label generation, studies also involved training AI models with co-registered WMH and MRI. In these studies, AI radiologist was designed to learn from the relationship between WMH and MRI features, thus enabling the identification and correlation of unobservable features on WMH and medical images, which significantly improved the performance of DL models in tumor classification [44, 46]. These studies proved the potential of WMH to serve not only as the ground truth, but also as learning material for AI radiologists.

Spatial distribution inspection: WMH provides a convenient approach to explore the spatial etiology of PCa on prostatic specimens. Research conducted by Zhou et al. [47] discovered that PCa primarily originates from the areas between the apex above 0.5–0.8 cm and the middle slice, and that the inferior half of the prostate had higher positive rate than superior half, offering valuable guidance for the protocol of systematic biopsy and conventional histopathologic sampling procedure.

Feature-location correlation: Distinguishing the PCa in different regions on WMH, such as peripheral zone (PZ) and transition zone (TZ), can result in more precise diagnosis and more accurate prediction of the tumor feature. In the PZ, ADC has been demonstrated as an effective predictor for PCa, while its efficacy diminishes to varying degrees in the TZ [48]. Other studies on PI-RADS score [49] and MRI-based AI radiologist [29] have found that the diagnosis performance the in PZ was found to be superior to that in the TZ. These findings suggested that the intraprostatic location provided by WMH may offer important implications for clinical PCa detection.

Periurethral PCa locating: The precise location of periurethral PCa is plays a crucial role in treatment planning for focal therapy. WMH provides legible and comprehensive spatial information of periurethral PCa lesions and urethral lumen, serving as a high-quality histopathologic reference. In two separate studies, MRI method was assessed by the precise location obtained from WMH sections, demonstrating substantial [50] and moderate [51] accuracy, respectively, in identifying periurethral PCa (with distance < 5 mm to urethral lumen). By using WMH as the gold standard, further advancements can be achieved to improve treatment planning for focal therapy.