Purpose This study proposes an automated prostate cancer (PC) lesion characterization method based on the deep neural network to determine tumor burden on 68 Ga-PSMA-11 PET/CT to potentially facilitate the optimization of PSMA-directed radionuclide therapy. Methods We collected 68 Ga-PSMA-11 PET/CT images from 193 patients with metastatic PC at three medical centers. For proofof-concept, we focused on the detection of pelvis bone and lymph node lesions. A deep neural network (triple-combining 2.5D U-Net) was developed for the automated characterization of these lesions. The proposed method simultaneously extracts features from axial, coronal, and sagittal planes, which mimics the workflow of physicians and reduces computational and memory requirements. Results Among all the labeled lesions, the network achieved 99% precision, 99% recall, and an F1 score of 99% on bone lesion detection and 94%, precision 89% recall, and an F1 score of 92% on lymph node lesion detection. The segmentation accuracy is lower than the detection. The performance of the network was correlated with the amount of training data. Conclusion We developed a deep neural network to characterize automatically the PC lesions on 68 Ga-PSMA-11 PET/CT. The preliminary test within the pelvic area confirms the potential of deep learning methods. Increasing the amount of training data should further enhance the performance of the proposed method and may ultimately allow whole-body assessments.

Purpose 68 Ga-PSMA-11 PET/CT is commonly performed at 1 h post injection (p.i.). However, various publications have demonstrated that most prostate cancer (PC) lesions exhibit higher contrast at later imaging. The aim of this study was to compare the Bcommon^protocol of 68 Ga-PSMA-11 PET/CT with a modified protocol. Methods In 2017, we used the following scanning protocol for 68 Ga-PSMA-11 PET/CT in patients with recurrent PC: acquisition at 1 h p.i. without further preparations. From 2018, all scans were conducted at 1.5 h p.i. In addition, patients were orally hydrated with 1 L of water 0.5 h p.i. and were injected with 20 mg of furosemide 1 h p.i. Both protocols including 112 patients (2017) and 156 (modified protocol in 2018) were retrospectively compared. Rates of pathologic scans, maximum standardized uptake values (SUVmax), and tumor contrast (ratio lesion-SUVmax/background-SUVmean) as well as average standardized uptake values (SUVmean) of urinary bladder were analyzed. Results Both tumor contrast and tracer uptake were significantly (p < 0.001) higher in the novel protocol. Although statistically not significant, the rates of pathologic scans were also higher in the modified protocol: 76.3% vs. 68.8% for all PSA values including 38.9% vs. 25.0% for PSA < 0.5 ng/ml and 60.0% vs. 56.7% for PSA > 0.5-≤ 2.0 ng/ml. Average SUVmean of the urinary bladder was significantly (p < 0.001) lower with the modified protocol. Conclusions The modified protocol, which includes a combination of delayed image acquisition at 1.5 h p.i., hydration, and furosemide resulted in higher tumor contrast and seems to have the potential to increase the rates of pathological scans, especially at low PSA levels.

Distinct thresholds of the relative aBSI could help to better assess disease progression in mPCa patients. Manual corrections of the BSI values are not required in most cases. The aBSI could serve as a useful additional parameter for therapy monitoring in mPCa patients in the future.

During angiogenesis, single endothelial cells (EC) specialize into tip cells that guide vessel sprouting towards growth factor gradients and instruct the adjacent vessel stalk. The balance between tip and stalk cells is regulated by endothelial Notch signalling through the expression of Notch ligand Delta-like 4 (Dll4) in tip cells, which suppresses a tip cell fate in adjacent stalk cells. Here we show, using genetic reporter and conditional deletion strategies, that myeloid cells regulate tip cell numbers and Dll4 expression via the Notch ligand Dll1 during vascular development in the retina. Dll1 is selectively expressed by a subpopulation of retinal myeloid cells, which progressively localizes to the sprouting vascular network. Conditional, myeloid-specific deletion of Dll1 impairs endothelial Dll4 tip-stalk gradient resulting in an increase of endothelial tip cells and EC filopodia, accompanied by an increase in vascular density and branching. In vitro , co-culture of human EC with monocyte-derived macrophages induced Dll1 upregulation in macrophages and Dll4 upregulation and an endothelial tip cell signature in EC. Furthermore, culturing human EC on recombinant DLL1 induced endothelial Dll4 expression and a tip cell program, indicating that changes are Dll1-dependent. Thus, myeloid cells regulate tip cell fate and angiogenesis through expression of Notch ligand Dll1.