Treating w/ AI

How can AI predict responses to immunotherapy?

Filed under
Treating w/ AI
AI keeps track of cancer cells overlooked by the human eye.

After studying the PD-L1 expression of 380,000 cancer cells, AI* can predict the result of immunotherapy with 96% accuracy. It’s like using a magnet to find the needle in the haystack, as AI* can identify all cancer cells expressing PD-L1 in a slide of tissue.

Consequently, AI* is better at finding patients responsive to immunotherapy than human doctors. Approximately 50% of the patients previously considered unfit for immunotherapy are found to be actually responsive.¹

AI* can accurately quantify the identified biomarkers like PD-L1, but it can also present itself as a new biomarker by analyzing the distribution of immunocytes within the cancer tissue to predict the response to immunotherapy.

This AI-based biomarker* starts from the scientific premise that for an immunocyte to attack the cancer cells, it has to be close to them.

Conventional pathology techniques and human analytical capability proved unable to accurately identify the distribution of immunocytes.
One tissue slide image, scanned for digital analysis, is a file as large as a single HD movie. To have a file that large dissected and zoomed into each pixel in each frame, would take ages. Immunocytes and cancer cells inside a tumor form a very complex microenvironment, and having them analyzed and quantified is no human job.

The higher the rate of the immune-inflamed pattern, the more effective immunotherapy is on the patient.

AI* can provide an accurate analysis of a tissue slide through deep-learning algorithm.

It first starts by identifying each feature and categorizing the distribution pattern. Immunocytes could be distributed near the cancer cells (immune-inflamed) or far away (immune-excluded) or could exist in few numbers (immune-desert).

The AI* then quantifies the rate of distribution pattern in order to predict responses to immunotherapy. The higher the rate of the immune-inflamed pattern, the more effective immunotherapy is on the patient.

AI discovers 50% more patients eligible for immunotherapy.

A research was conducted where AI* predicted a patient to show effective response to immunotherapy. Immunotherapy was administered to the patient, and it did result in extended survival, proving the accuracy of the AI-based biomarker*.²

If the AI-based biomarker* currently in development is utilized in immunotherapy, it will allow us not only to discover cancer patients who would’ve been missed by the conventional biomarkers like PD-L1, but also allow us to effectively use immunotherapy on more patients by accurately predicting their responses to the treatment.

*For research use
¹ Hyojin Kim, et al. Clinical performance of artificial intelligence-powered annotation of tumor cell PD-L1 expression for treatment of immune-checkpoint inhibitor (ICI) in advanced non-small lung cancer (NSCLC), ASCO 2021
² Jeanne Shen, et al. Artificial intelligence-powered spatial analysis of tumor-infiltrating lymphocytes predicts survival after immune checkpoint inhibitor therapy across multiple cancer types, ASCO 2021
Copy Link
chapter 3
Treating w/ AI
Can't find a topic you want?
Please ask us anything.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.