On January 27, 2020 NantHealth, Inc. (NASDAQ: NH), a next-generation, evidence-based, personalized healthcare company and NantOmics, LLC, the leader in molecular analysis, reported a novel artificial intelligence platform for aiding pathologists in image-based lung cancer subtyping at the Society for Imaging Science and Technology’s International Symposium on Electronic Imaging 2020 (Press release, NantHealth, JAN 27, 2020, View Source [SID1234553572]). This novel machine vision software platform accurately subtypes lung cancer pathology and achieves high concordance with analysis performed by trained medical pathologists.
Schedule your 30 min Free 1stOncology Demo!
Discover why more than 1,500 members use 1stOncology™ to excel in:
Early/Late Stage Pipeline Development - Target Scouting - Clinical Biomarkers - Indication Selection & Expansion - BD&L Contacts - Conference Reports - Combinatorial Drug Settings - Companion Diagnostics - Drug Repositioning - First-in-class Analysis - Competitive Analysis - Deals & Licensing
Schedule Your 30 min Free Demo!
An initial report of the AI technology was presented at the Sixth American Association for Cancer Research (AACR) (Free AACR Whitepaper) and the International Association for the Study of Lung Cancer (IASLC) International Joint Conference. The study entitled, "Tumor-infiltrating lymphocytes (TILs) found elevated in lung adenocarcinomas (LUAD) using automated digital pathology masks derived from deep-learning models" concluded that despite lower overall TMB (tumor mutation burden) and lymphocyte levels, there exists a subset of lung cancers with very high infiltrating lymphocyte counts.
Derived from deep-learning models, together, the findings demonstrate a novel AI-based method for subtyping lung cancer pathologies which impacts treatment options for patients and improved methods of identifying tumor infiltrating white cells found elevated in lung cancer.
"Accurately identifying and quantifying tumor-infiltrating white cells is extremely important for prognosis and treatment decisions in this era of personalized medicine, yet it currently requires manual review of whole slide images by medically trained pathologists, and incurs significant delays and cost," explains Dr. Patrick Soon-Shiong, MD, Chairman and CEO of NantHealth. "Our goal was to develop a scalable remote cloud-based diagnostic imaging system, a NORAD of pathology diagnosis so to speak. To accomplish this, machine vision of digitally transmitted images of tumor tissue would facilitate a scalable cloud-based infrastructure, with an image patch-based, automated system to classify cancers by their immune status."
Non-small cell lung cancer (NSCLC) is the most common form of lung cancer, which is further classified as 40 percent adenocarcinoma (Adeno), 30 percent squamous cell carcinoma (Squamous) and the remainder, large cell carcinoma1. As analyses show that lung adenocarcinomas (LUAD) receive slightly more survival benefit from anti-PD1 therapy than squamous-cell lung carcinomas (LUSC), which have a higher TMB, a team of researchers explored whether lymphocyte distribution in the tumor microenvironment may give a rational explanation for the different responses to immuno-oncology agents independent of TMB.
"By focusing on classifying regions detected as tumorous, we achieved identification of adenocarcinomas versus squamous cell carcinomas in non-small-cell lung cancers with an approximate accuracy rate of 86 percent," explained Soon-Shiong. "With highly accurate tumor-region and lymphocyte detection, oncologists may better treat their patients with adeno versus squamous-based therapies and the use of immunotherapies may result in better outcomes."
Study Design:
The system was trained and tested on 876 subtyped NSCLC gigapixel-resolution diagnostic whole slide images (WSI) from 805 patients obtained from The Cancer Genome Atlas (TCGA) sources. Samples were randomly split into training (711 WSIs from 664 patients) and testing (165 WSIs from 141 patients) sets.
Findings show that NantOmics and NantHealth’s fully-automated histopathology subtyping AI method outperforms other algorithms reported in literature for diagnostic WSIs. The system also generated maps of (tumor) regions-of-interest within WSIs, providing novel spatial information on tumor organization.
Details of the oral presentation at the IS&T International Symposium on Electronic Imaging 2020 outlined below:
Title: "Pathology image-based lung cancer subtyping using deep-learning features and cell-density maps"
Authors: Mustafa I. Jaber, Christopher W. Szeto, Bing Song, Liudmila Beziaeva, Stephen Benz, Patrick Soon-Shiong, and Shahrooz Rabizadeh
Session and Number: Image Processing: Algorithms and Systems XVIII (IPAS-064)
Location: Hyatt Regency San Francisco Airport, Burlingame, CA
Date and Time: January 27, 2020 at 4:10 PM