Enhancing the diagnostic capacity of [18F]PSMA-1007 PET/MRI in primary prostate cancer staging with artificial intelligence and semi-quantitative DCE: an exploratory study

Abstract

Background: To investigate the ability of artificial intelligence (AI)-based and semi-quantitative dynamic contrast enhanced (DCE) multiparametric MRI (mpMRI), performed within [¹⁸F]-PSMA-1007 PET/MRI, in differentiating benign from malignant prostate tissues in patients with primary prostate cancer (PC).

Results: A total of seven patients underwent whole-body [¹⁸F]-PSMA-1007 PET/MRI examinations including a pelvic mpMRI protocol with T2w, diffusion weighted imaging (DWI) and DCE image series. Conventional analysis included visual reading of PET/MRI images and Prostate Imaging Reporting & Data System (PI-RADS) scoring of the prostate. On the prostate level, we performed manual segmentations for time-intensity curve parameter formation and semi-quantitative analysis based on DCE segmentation data of PC-suspicious lesions. Moreover, we applied a recently introduced deep learning (DL) pipeline previously trained on 1010 independent MRI examinations with systematic biopsy-enhanced histopathological targeted biopsy lesion ground truth in order to perform AI-based lesion detection, prostate segmentation and derivation of a deep learning PI-RADS score. DICE coefficients between manual and automatic DL-acquired segmentations were compared. On patient-based analysis, PET/MRI revealed PC-suspicious lesions in the prostate gland in 6/7 patients (Gleason Score-GS ≥ 7b) that were histologically confirmed. Four of these patients also showed lymph node metastases, while two of them had bone metastases. One patient with GS 6 showed no PC-suspicious lesions. Based on DCE segmentations, a distinction between PC-suspicious and normal appearing tissue was feasible with the parameters fitted maximum contrast ratio (FMCR) and wash-in-slope. DICE coefficients (manual vs. deep learning) were comparable with literature values at a mean of 0.44. Further, the DL pipeline could identify the intraprostatic PC-suspicious lesions in all six patients with clinically significant PC.

Conclusion: Firstly, semi-quantitative DCE analysis based on manual segmentations of time-intensity curves was able to distinguish benign from malignant tissues. Moreover, DL analysis of the MRI data could detect clinically significant PC in all cases, demonstrating the feasibility of AI-supported approaches in increasing diagnostic certainty of PSMA-radioligand PET/MRI.

Keywords: 18F-PSMA-1007; AI; DCE; PET/MRI; Primary staging; Prostate cancer.

Fig. 1

A 77-year-old patient with an initial diagnosis of PC (Gleason score 9) referred for initial staging with PET-MRI. His PSA level at the time of examination was 18.9 ng/ml. Distinct T2w hypointensity of the prostate (A) with marked diffusion restriction with signal increase in the high b-values (B) and low ADC values in the ADC-map (C) predominantly in the left hemisphere of the prostate with extension over the midline to the right side and with extracapsular extension compatible with a PIRADS score of 5. The AI-based PI-RADS equivalent (DL-PIRADS) is in good agreement with the MRI findings and also shows a very high probability score for PCa (DL-PIRADS 5, D). Correspondingly, [¹⁸F]-PSMA-1007 PET (fused T2w, (E)) shows an intense [¹⁸F]-PSMA-1007 accumulation in the prostate gland. Quantitative DCE data based on manual segmentation showed a steeper wash-in slope for tumor-suspicious lesion (TSL, red) compared to perilesional tissue (PLT, yellow) and normal appearing tissue (NAT, green). The corresponding fitted maximum (*, intersection of wash-in and wash-out slope) was 4.2 for TSL, 2.2 for PLT and 1.0 for NAT, allowing a clear discrimination between them. All 3 curves showed an increasing curve pattern (F). Whole-body [¹⁸F]-PSMA-1007 PET imaging showed the primary tumor, multiple iliac, retroperitoneal, mediastinal and cervical lymph node metastases, as well as bone metastases in the sacral bone and the thoracic spine (G)

Fig. 2

A 64-year-old patient with an initial diagnosis of PC (Gleason score 9) referred for initial staging with PET-MRI. His PSA level at the time of examination was 6.8 ng/ml. MRI showed a PC suspicious lesion in the anterior peripheral and adjacent transitional zone in the predominatly right-sided prostate apex with a diameter of 16 mm, showing T2w hypointensity (A), marked diffusion-restriction with signal enhancement in the high B-values (B) and signal decrease in the ADC-map (C), compatible with PIRADS 5 score. The AI probability score is in good agreement with the MRI findings, which also show a very high probability score for PC (DL-PIRADS 5, D). PET-MRI (fused T2w, E) shows intense [¹⁸F]-PSMA-1007 accumulation in this lesion. Quantitative DCE data based on manual segmentation showed a steeper wash-in-slope for the tumor-suspicious lesion (TSL, red) compared to the perilesional tissue (PLT, yellow) and normal appearing tissue (NAT, green), which showed comparable wash-in slopes. The washout slope for TSL was slightly negative. PLT and NAT showed an increasing curve pattern. The corresponding fitted maximum (*, intersection of wash-in and wash-out slopes) was 1.5 for TSL, 1.1 for PLT and 1.0 for NAT, allowing a clear distinction between TSL versus PLT and NAT (F). Whole-body [¹⁸F]-PSMA-1007 PET imaging (not shown) showed no lesions consistent with metastases. Histology confirmed an anterior, predominantly right-sided prostate carcinoma (digitally merged whole-mount prostate slide, hematoxylin and eosin staining, G)

Fig. 3

A 70-year-old patient with an initial diagnosis of PC (Gleason score 9) referred for initial staging with PET-MRI. His PSA level at the time of examination was 63.0 ng/ml. MRI showed a lesion suspicious for PC mainly in the posteromedial/-lateral peripheral zone with infiltration in the seminal vesicles (not shown) with T2w hypointensity (A) and diffusion restriction with increase in high b-value (B) and decrease in ADC map (C). The AI probability score is in good agreement with the MRI findings and also shows a high probability score for PC (DL-PIRADS 4, D). Quantitative data of DCE based on manual segmentations showed a steeper wash-in slope for tumor-suspicious lesion (TSL, red) compared to perilesional tissue (PLT, yellow) and normal appearing tissue (NAT, green). The corresponding fitted maximum (*, intersection of wash-in and wash-out slope) was 3.3 for TSL, 1.4 for PLT and 0.9 for NAT, enabling a clear distinction between TSL, PLT and NAT. All three curves showed an increasing curve pattern (F). Whole-body [¹⁸F]-PSMA-1007 PET imaging revealed the primary tumor as well as multiple PSMA-avid iliac lymph node metastases and osseous metastases in the sacral bone, thoracic spine and left scapula (G)

Fig. 4

Schematic representation of DCE features for quantitative differentiation of malignant and normal prostate tissue conditions. The wash-in and wash-out slopes of an intensity curve (tumor or normal tissue) can be used to determine each “fitted maximum” (intersection of the dashed lines, shown here with red and green asterisks). The ratio of these fitted maximum intensity values is called “fitted maximum contrast ratio” (fMCR)