首页|University of Utah Reports Findings in Cancer (A pipeline for evaluation of mach ine learning/AI models to quantify PD-L1 immunohistochemistry)

University of Utah Reports Findings in Cancer (A pipeline for evaluation of mach ine learning/AI models to quantify PD-L1 immunohistochemistry)

扫码查看
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News - New research on Cancer is the subject of a report. According to news reporting originating from Salt Lake City, Utah, by NewsRx correspondents, research stated, “Immunohistochemistry (IHC) is used t o guide treatment decisions in multiple cancer types. For treatment with checkpo int inhibitors, PD-L1 IHC is used as a companion diagnostic.” Our news editors obtained a quote from the research from the University of Utah, “However, the scoring of PD-L1 is complicated by its expression in cancer and i mmune cells. Separation of cancer and non-cancer regions is needed to calculate tumor proportion scores (TPS) of PD-L1, which is based on the percentage of PD-L 1 positive cancer cells. Evaluation of PD-L1 expression requires highly experien ced pathologists and is often challenging and time consuming. Here we used a mul ti-institutional cohort of 77 lung cancer cases stained centrally with the PD-L1 22C3 clone. We developed a four-step pipeline for measuring TPS that includes t he co-registration of H&E, PD-L1 and negative control (NC) digital slides for exclusion of necrosis, segmentation of cancer regions and quantificat ion of PD-L1+ cells. As cancer segmentation is a challenging step for TPS genera tion, we trained DeepLab V3 in the Visiopharm software package to outline cancer regions in PD-L1 and negative control (NC) images and evaluated the model perfo rmance by mean intersection over union (mIoU) against manual outlines. Only 14 c ases were required to accomplish an mIoU of 0.82 for cancer segmentation in hema toxylin stained NC cases. For PD-L1 stained slides, a model trained on PD-L1 til es augmented by registered NC tiles achieved an mIoU of 0.79. In segmented cance r regions from whole slide images, the digital TPS achieved an accuracy of 75 % against the manual TPS scores from the pathology report. Major reasons for algor ithmic inaccuracies include the inclusion of immune cells in cancer outlines and poor nuclear segmentation of cancer cells.”

Salt Lake CityUtahUnited StatesNor th and Central AmericaCancerCyborgsEmerging TechnologiesHealth and Medic ineMachine LearningOncology

2024

Robotics & Machine Learning Daily News

Robotics & Machine Learning Daily News

ISSN:
年,卷(期):2024.(MAY.14)