Classification of melanocytic lesions in selected and whole-slide images via convolutional neural networks

Steven Hart, William Flotte, Andrew P. Norgan, Kabeer K. Shah, Zachary R. Buchan, Taofic Mounajjed, Thomas J Flotte

Research output: Contribution to journalArticle

Abstract

Whole-slide images (WSIs) are a rich new source of biomedical imaging data. The use of automated systems to classify and segment WSIs has recently come to forefront of the pathology research community. While digital slides have obvious educational and clinical uses, their most exciting potential lies in the application of quantitative computational tools to automate search tasks, assist in classic diagnostic classification tasks, and improve prognosis and theranostics. An essential step in enabling these advancements is to apply advances in machine learning and artificial intelligence from other fields to previously inaccessible pathology datasets, thereby enabling the application of new technologies to solve persistent diagnostic challenges in pathology. Here, we applied convolutional neural networks to differentiate between two forms of melanocytic lesions (Spitz and conventional). Classification accuracy at the patch level was 99.0%-2% when applied to WSI. Importantly, when the model was trained without careful image curation by a pathologist, the training took significantly longer and had lower overall performance. These results highlight the utility of augmented human intelligence in digital pathology applications, and the critical role pathologists will play in the evolution of computational pathology algorithms.

Original languageEnglish (US)
Article number5
JournalJournal of Pathology Informatics
Volume10
Issue number1
DOIs
StatePublished - Jan 1 2019

Fingerprint

Pathology
Neural networks
Artificial Intelligence
Intelligence
Artificial intelligence
Learning systems
Technology
Imaging techniques
Research
Pathologists

Keywords

  • Bioinformatics
  • deep learning
  • dermatology
  • image analysis

ASJC Scopus subject areas

  • Pathology and Forensic Medicine
  • Health Informatics
  • Computer Science Applications

Cite this

Classification of melanocytic lesions in selected and whole-slide images via convolutional neural networks. / Hart, Steven; Flotte, William; Norgan, Andrew P.; Shah, Kabeer K.; Buchan, Zachary R.; Mounajjed, Taofic; Flotte, Thomas J.

In: Journal of Pathology Informatics, Vol. 10, No. 1, 5, 01.01.2019.

Research output: Contribution to journalArticle

@article{8c119e23c2f74e5f93668cc3700dd066,
title = "Classification of melanocytic lesions in selected and whole-slide images via convolutional neural networks",
abstract = "Whole-slide images (WSIs) are a rich new source of biomedical imaging data. The use of automated systems to classify and segment WSIs has recently come to forefront of the pathology research community. While digital slides have obvious educational and clinical uses, their most exciting potential lies in the application of quantitative computational tools to automate search tasks, assist in classic diagnostic classification tasks, and improve prognosis and theranostics. An essential step in enabling these advancements is to apply advances in machine learning and artificial intelligence from other fields to previously inaccessible pathology datasets, thereby enabling the application of new technologies to solve persistent diagnostic challenges in pathology. Here, we applied convolutional neural networks to differentiate between two forms of melanocytic lesions (Spitz and conventional). Classification accuracy at the patch level was 99.0{\%}-2{\%} when applied to WSI. Importantly, when the model was trained without careful image curation by a pathologist, the training took significantly longer and had lower overall performance. These results highlight the utility of augmented human intelligence in digital pathology applications, and the critical role pathologists will play in the evolution of computational pathology algorithms.",
keywords = "Bioinformatics, deep learning, dermatology, image analysis",
author = "Steven Hart and William Flotte and Norgan, {Andrew P.} and Shah, {Kabeer K.} and Buchan, {Zachary R.} and Taofic Mounajjed and Flotte, {Thomas J}",
year = "2019",
month = "1",
day = "1",
doi = "10.4103/jpi.jpi_32_18",
language = "English (US)",
volume = "10",
journal = "Journal of Pathology Informatics",
issn = "2229-5089",
publisher = "Medknow Publications and Media Pvt. Ltd",
number = "1",

}

TY - JOUR

T1 - Classification of melanocytic lesions in selected and whole-slide images via convolutional neural networks

AU - Hart, Steven

AU - Flotte, William

AU - Norgan, Andrew P.

AU - Shah, Kabeer K.

AU - Buchan, Zachary R.

AU - Mounajjed, Taofic

AU - Flotte, Thomas J

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Whole-slide images (WSIs) are a rich new source of biomedical imaging data. The use of automated systems to classify and segment WSIs has recently come to forefront of the pathology research community. While digital slides have obvious educational and clinical uses, their most exciting potential lies in the application of quantitative computational tools to automate search tasks, assist in classic diagnostic classification tasks, and improve prognosis and theranostics. An essential step in enabling these advancements is to apply advances in machine learning and artificial intelligence from other fields to previously inaccessible pathology datasets, thereby enabling the application of new technologies to solve persistent diagnostic challenges in pathology. Here, we applied convolutional neural networks to differentiate between two forms of melanocytic lesions (Spitz and conventional). Classification accuracy at the patch level was 99.0%-2% when applied to WSI. Importantly, when the model was trained without careful image curation by a pathologist, the training took significantly longer and had lower overall performance. These results highlight the utility of augmented human intelligence in digital pathology applications, and the critical role pathologists will play in the evolution of computational pathology algorithms.

AB - Whole-slide images (WSIs) are a rich new source of biomedical imaging data. The use of automated systems to classify and segment WSIs has recently come to forefront of the pathology research community. While digital slides have obvious educational and clinical uses, their most exciting potential lies in the application of quantitative computational tools to automate search tasks, assist in classic diagnostic classification tasks, and improve prognosis and theranostics. An essential step in enabling these advancements is to apply advances in machine learning and artificial intelligence from other fields to previously inaccessible pathology datasets, thereby enabling the application of new technologies to solve persistent diagnostic challenges in pathology. Here, we applied convolutional neural networks to differentiate between two forms of melanocytic lesions (Spitz and conventional). Classification accuracy at the patch level was 99.0%-2% when applied to WSI. Importantly, when the model was trained without careful image curation by a pathologist, the training took significantly longer and had lower overall performance. These results highlight the utility of augmented human intelligence in digital pathology applications, and the critical role pathologists will play in the evolution of computational pathology algorithms.

KW - Bioinformatics

KW - deep learning

KW - dermatology

KW - image analysis

UR - http://www.scopus.com/inward/record.url?scp=85063861699&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063861699&partnerID=8YFLogxK

U2 - 10.4103/jpi.jpi_32_18

DO - 10.4103/jpi.jpi_32_18

M3 - Article

VL - 10

JO - Journal of Pathology Informatics

JF - Journal of Pathology Informatics

SN - 2229-5089

IS - 1

M1 - 5

ER -