GenomiCare unveiled the results of PD-L1 immunohistochemistry images in outcome prediction at ASCO 2020

Apr 7, 2020

Recently, the summary of the results of GenomiCare PD-L1 immunohistochemistry (IHC) images in outcome prediction was included in the ASCO 2020 (56th Annual Meeting of the American Society of Clinical Oncology), entitled “Application of Automated Segmentation and Classification of PD-L1 Immunohistochemistry Images in Cancer Treatment Prediction”. This marks the first appearance of GenomiCare’s “AI+Medical” application results at the world-class, high-level and professional clinical oncology conference.

Previously at the World Artificial Intelligence Conference 2019 (WAIC 2019), the application scenarios of GenomiCare “AI+Medical” debuted, drawing the attention of many visitors from healthcare sectors and well-known enterprises at home and abroad.

AI-assisted image interpretation was one of the highlights displayed by GenomiCare at WAIC 2019, which is also the area covered by the summary of results included in the 2020 ASCO. Various examinations often produce a large number of images in the course of a patient’s treatment. For interpretation of such images, physicians are required to fully leverage their wealth of clinical experience and professional knowledge to avoid false positive and false negative errors. However, these repetitious tasks create unnecessary burnout for physicians.

By introducing AI to, for example, the interpretation of PD1 IHC images, GenomiCare puts into place the most advanced model in the field of computer image segmentation and classification. With adjustments made based on clinical practices, AI can be trained with over 20,000 images marked by physicians in the database so as to assist physicians in image interpretation, leading to significant improvement in the efficiency of image interpretation.

Background Therapies targeting immune checkpoints, such as the programmed cell death-1 (PD-1) / programmed cell death ligand 1 (PD-L1) pathway, have become successful in treating several cancer types, particularly those with high PD-L1 expression. Patient response rates, however, are variable, demanding a more sensitive companion diagnostic test than the current method of manual scoring of PD-L1 immunohistochemistry images. The low predictive value of manual PD-L1 scoring is at least partially compounded by the pathologist-to-pathologist or hospital-to-hospital variation. To eliminate the variation caused by human factors, we developed an alternative PD-L1 scoring method based on automated image analysis and assessed its predictive and prognostic values.

Methods Archival or fresh tumor biopsies were analyzed for PD-L1 expression by immunohistochemistry. Digital images were automatically scored as PD-L1 positive or negative using a custom algorithm implemented in Python and Tensorflow, which uses a fine-tuned, ImageNet pre-trained, Inception-ResNet-V2 model as the binary classifier along with a modified U-net image segmentation model to determine the final prediction. Supervised machine learning was used to control for heterogeneity introduced by PD- L1 positive inflammatory cells in the tumor microenvironment. Samples from 256 patients were collected and used in algorithm training and verification by random assignment with an 8:2 split ratio.

Results In total, 10,000 images with balanced PD-L1 positive and negative distributions were obtained. A highly significant correlation of PD-L1 scores was found between pathologist-based consensus reading and automated analysis using our custom algorithm (R=0.97, p<0.0001). The automated analysis reached 0.94 sensitivity and 0.96 specificity. Additionally, it showed excellent reproducibility (R=1.0, p<0.0001) in independent trials executed on different workstations, which is far superior to the manual scoring performed by human examiners.

Conclusions Our automated PD-L1 evaluation algorithm significantly reduces scoring variability. It may help to identify cancer patients who can have optimal response to PD-1/PD-L1 related immune therapy and therefore facilitate patient stratification in clinical practice.

AI-assisted image interpretation is only one of the applications and services enabled by the medical-scenario-based AI and database of GenomiCare. There are many more powerful functions such as scoring and screening assisted by NGS data mutation, standardized input of clinical information, and the decision tree model combining NGS and clinical information. By tapping deep into the application of AI in tumor diagnosis and treatment, they can truly benefit patients; help improve physician efficiency, and drive the development of precision oncology in China.

GenomiCare has also deepened the application of AI in clinical and translational oncology. With the integration of international standard CLIA / CAP clinical laboratory services, the in-depth combination of real-world molecular + clinical data services, the open and interactive SaaS on cloud, and the unique data-driven intelligent algorithm, GenomiCare launched its Data Intelligence service, offering a powerful and flexible platform for oncology drug development for biopharmaceutical companies. Clinical validation and exploration can be conducted based on clients’ interest to help realize the full potential of clinical data easily and swiftly.

At present, GenomiCare’s Data Intelligence has received preliminary validation in many aspects such as the discovery of new biomarkers and drug targets, the exploration of mechanism of action, clinical trial design and prediction, drug combination strategy, the accurate stratification of patients, and the expansion of new indications. We look forward to closer partnerships with more leading biological and pharmaceutical companies in the future to jointly promote the development of innovative anti-cancer drugs.