Affordable Access

Access to the full text

Visualizing “featureless” regions on mammograms classified as invasive ductal carcinomas by a deep learning algorithm: the promise of AI support in radiology

Authors
  • Ueda, Daiju1
  • Yamamoto, Akira1
  • Takashima, Tsutomu1
  • Onoda, Naoyoshi1
  • Noda, Satoru1
  • Kashiwagi, Shinichiro1
  • Morisaki, Tamami1
  • Tsutsumi, Shinichi2
  • Honjo, Takashi1
  • Shimazaki, Akitoshi1
  • Goto, Takuya1
  • Miki, Yukio1
  • 1 Graduate School of Medicine, Osaka City University, 1-4-3 Asahi-machi, Abeno-ku, Osaka, 545-8585, Japan , Osaka (Japan)
  • 2 Osaka Prefectural Hospital Organization Osaka Habikino Hospital, 3-7-1 Habikino, Habikino city, Osaka, 583-8588, Japan , Osaka (Japan)
Type
Published Article
Journal
Japanese Journal of Radiology
Publisher
Springer Singapore
Publication Date
Nov 16, 2020
Volume
39
Issue
4
Pages
333–340
Identifiers
DOI: 10.1007/s11604-020-01070-9
Source
Springer Nature
Keywords
License
Yellow

Abstract

PurposeTo demonstrate how artificial intelligence (AI) can expand radiologists’ capacity, we visualized the features of invasive ductal carcinomas (IDCs) that our algorithm, developed and validated for basic pathological classification on mammograms, had focused on.Materials and methodsIDC datasets were built using mammograms from patients diagnosed with IDCs from January 2006 to December 2017. The developing dataset was used to train and validate a VGG-16 deep learning (DL) network. The true positives (TPs) and accuracy of the algorithm were externally evaluated using the test dataset. A visualization technique was applied to the algorithm to determine which malignant findings on mammograms were revealed.ResultsThe datasets were split into a developing dataset (988 images) and a test dataset (131 images). The proposed algorithm diagnosed 62 TPs with an accuracy of 0.61–0.70. The visualization of features on the mammograms revealed that the tubule forming, solid, and scirrhous types of IDCs exhibited visible features on the surroundings, corners of the masses, and architectural distortions, respectively.ConclusionWe successfully showed that features isolated by a DL-based algorithm trained to classify IDCs were indeed those known to be associated with each pathology. Thus, using AI can expand the capacity of radiologists through the discovery of previously unknown findings.

Report this publication

Statistics

Seen <100 times