Figure - PMC

Skip to main content

An official website of the United States government

Here's how you know

Here's how you know

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

View full-text article in PMC

. 2022 Apr 13;14(8):1964. doi: 10.3390/cancers14081964

Search in PMC
Search in PubMed
View in NLM Catalog
Add to search

© 2022 by the authors.

Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

PMC Copyright notice

Communicator-based clean-up can be transferred to CNNs using a different input size. (a) Selection of the normalized 299 × 299 × 3 tissue patches based on the classification of the communicator CNNs using a 224 × 224 × 3 crop of the image tiles is illustrated (Spot: Scale bar = 300 μm, Patch: Scale bar = 60 μm). (b) Heatmap of the performance difference between the inceptionnetv3 trained on data with and without the data clean-up via the communicators is shown. (c) Visualization of the classified validation data by the baseline or cleaned-up inceptionv3 net. Healthy lymph node (HLN, yellow), healthy pancreas (HP, red), pancreatic ductal adenocarcinoma (PDAC, blue), background (BG, grey) and adipose tissue (ADI, cyan) classification is illustrated (scale bar = 2 mm). (d) Pooled classification, as determined using a baseline and cleaned inceptionv3 network from healthy pancreas (HP) (n = 3) and pancreatic ductal adenocarcinoma (PDAC) images (n = 15), is illustrated. (e) Average of classification, as determined using a baseline and cleaned inceptionv3 network of images showing healthy lymph nodes (HLN) (n = 5) and lymph nodes with pancreatic ductal adenocarcinoma metastasis (LNPM) (n = 6), is presented (ADI= adipose tissue, BG = background).

Communicator-based clean-up can be transferred to CNNs using a different input size. (a) Selection of the normalized 299 × 299 × 3 tissue patches based on the classification of the communicator CNNs using a 224 × 224 × 3 crop of the image tiles is illustrated (Spot: Scale bar = 300 μm, Patch: Scale bar = 60 μm). (b) Heatmap of the performance difference between the inceptionnetv3 trained on data with and without the data clean-up via the communicators is shown. (c) Visualization of the classified validation data by the baseline or cleaned-up inceptionv3 net. Healthy lymph node (HLN, yellow), healthy pancreas (HP, red), pancreatic ductal adenocarcinoma (PDAC, blue), background (BG, grey) and adipose tissue (ADI, cyan) classification is illustrated (scale bar = 2 mm). (d) Pooled classification, as determined using a baseline and cleaned inceptionv3 network from healthy pancreas (HP) (n = 3) and pancreatic ductal adenocarcinoma (PDAC) images (n = 15), is illustrated. (e) Average of classification, as determined using a baseline and cleaned inceptionv3 network of images showing healthy lymph nodes (HLN) (n = 5) and lymph nodes with pancreatic ductal adenocarcinoma metastasis (LNPM) (n = 6), is presented (ADI= adipose tissue, BG = background).