Can Deep Learning Predict PD-L1 Status A Fine-Tuned InceptionV3 Convolution Neural Network Architecture Accurately Determines PD-L1 Status in Colonic Adenocarcinoma

Abstract

PD-L1 immunohistochemical staining is becoming increasingly important in the oncological treatment of various neoplasms. Interestingly , colonic adenocarcinomas show PD-L1 dependent morphological changes (medullary phenotype and increased tumor infiltrating lymphocytes) rendering them potentially amenable to analysis by a Convolutional Neural Network (CNN). Previously, we have shown that CNNs have the potential to reduce costs and turnaround time associated with immunohistochemical (IHC) staining. Herein, we demonstrate that a fine tuned InceptionV3 CNN architecture is able to accurately and confidently determine PD L1 status from H and E images.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 10, 2020
Accession Number
AD1101322

Entities

People

  • Andrew L. Walls
  • Devin R. Broadwater
  • Nathaniel E. Smith

Organizations

  • 59th Medical Wing

Tags

DTIC Thesaurus Topics

  • Adenocarcinoma
  • Air Force
  • Artificial Intelligence Computing
  • Artificial Intelligence Software
  • Biological Staining And Labeling
  • Computing System Architectures
  • Convolutional Neural Networks
  • Deep Learning
  • Department Of Defense
  • Neoplasms
  • Network Architecture
  • Neural Networks
  • Probability
  • Probability Distributions
  • Test Sets
  • Training

Fields of Study

  • Medicine

Readers

  • Immunology
  • Neural Network Machine Learning.
  • Oncology and Biomarker-Based Cancer Detection.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks