Development of an automated combined positive score prediction pipeline using artificial intelligence on multiplexed immunofluorescence images

Recent Advancement of PD-L1 Detection Technologies and Clinical Applications in the Era of Precision Cancer Therapy

“…The clinical application of the technology could greatly reduce the burden on pathologists and make the task of scoring CPS less time-consuming and easier to perform. However, the technique still has some shortcomings and cannot effectively identify some artifacts, such as extrusion artifacts and PD-L1 staining necrosis 89 .…”

“…Numerous radionuclides have been utilized in the labeling of ligands for the production of various tracers. Each radionuclide has a different half-life: 68 Ga has a half-life of 68 min, 18 F has a half-life of 109.8 min, 99m Tc has a half-life of 6 h, 64 Cu has a half-life of 12.7 h, 111 In has a half-life of 2.8 d, 89 Zr has a half-life of 3.27 d, 124 I has a half-life of 4.18 d, and 125 I has a half-life of 60.1 d. Long half-life radionuclides such as 111 In and 89 Zr, which have similar half-lives to the biological half-lives of mAbs, are beneficial for imaging; however, long half-life radionuclides have drawbacks, such as delayed clearance, long imaging times, and excessive radiation doses to healthy organs [122]. Consequently, it is essential to choose the proper radionuclide when manufacturing tracers.…”

“…This finding encouraged further research and development of molecular PET imaging techniques to better assess PD-L1 status and treatment response in oncology patients [125]. Later, several novel 89 Zr-labeled tracers emerged, including [ 89 Zr]Zr-DFO-PD-L1 mAb [126], 89 Zr-DFO-6E11 [127], and [ 89 Zr]DFO-Anti-PDL1 [128], demonstrating that 89 Zr-labeled mAb-based tracers have some clinical translation potential. In addition, Li et al reported a humanized full-length anti-PD-L1 mAb probe labeled with 124/125 I: 124/125 I-CS1001, and their results demonstrated the feasibility of its use for monitoring and assessing PD-L1 expression in tumors [129].…”

Pros and cons of artificial intelligence implementation in diagnostic pathology

“…Histopathology published by John Wiley & Sons Ltd., Histopathology, 84, 924-934. cancer, 22,[47][48][49] and breast cancer grading, 50,51 counting Ki67-positive cells, 52 and therapeutic targets such as ER, PR, HER2, and PD-L1. 53 Scoring of HER2 IHC staining intensity, crucial for the decision on HER2 inhibition in breast cancer, is done more accurately by AI-supported pathologists compared to non-AI-assisted pathologists. 54 For prostate needle biopsies, variability in Gleason grading is higher in general pathologists than in expert urogenital pathologists.…”

“…Artificial intelligence has shown great promise in making pathology diagnostics more reproducible among pathologists. Examples include AI‐supported mitoses counting, 26,27,46 Gleason grading of prostate cancer, 22,47–49 and breast cancer grading, 50,51 counting Ki67‐positive cells, 52 and therapeutic targets such as ER, PR, HER2, and PD‐L1 53 …”