Artificial Intelligence Improves the Accuracy in Histologic Classification of Breast Lesions

António Polónia, MD, PhD; Sofia Campelos, MD; Ana Ribeiro, MD; Ierece Aymore, MD; Daniel Pinto, MD; Magdalena Biskup-Fruzynska, MD; Ricardo Santana Veiga, MD; Rita Canas-Marques, MD; Guilherme Aresta, MEng; Teresa Araújo, MEng; Aurélio Campilho, PhD; Scotty Kwok, MSc; Paulo Aguiar, PhD; Catarina Eloy, MD, PhD


Am J Clin Pathol. 2021;155(4):527-536. 

In This Article

Abstract and Introduction


Objectives: This study evaluated the usefulness of artificial intelligence (AI) algorithms as tools in improving the accuracy of histologic classification of breast tissue.

Methods: Overall, 100 microscopic photographs (test A) and 152 regions of interest in whole-slide images (test B) of breast tissue were classified into 4 classes: normal, benign, carcinoma in situ (CIS), and invasive carcinoma. The accuracy of 4 pathologists and 3 pathology residents were evaluated without and with the assistance of algorithms.

Results: In test A, algorithm A had accuracy of 0.87, with the lowest accuracy in the benign class (0.72). The observers had average accuracy of 0.80, and most clinically relevant discordances occurred in distinguishing benign from CIS (7.1% of classifications). With the assistance of algorithm A, the observers significantly increased their average accuracy to 0.88. In test B, algorithm B had accuracy of 0.49, with the lowest accuracy in the CIS class (0.06). The observers had average accuracy of 0.86, and most clinically relevant discordances occurred in distinguishing benign from CIS (6.3% of classifications). With the assistance of algorithm B, the observers maintained their average accuracy.

Conclusions: AI tools can increase the classification accuracy of pathologists in the setting of breast lesions.


Digital pathology (DP) has been implemented in several pathology departments around the world.[1–6] One of the main advantages of using whole-slide images (WSI) is the potential for implementing computer-aided diagnosis (CAD) tools that may improve the evaluation of tissue morphology, both quantitatively and qualitatively, adding robustness to image diagnosis.[7] The subjectivity in the appreciation of the histologic features of breast pathology sometimes leads to lower than desired interobserver concordance rates, with borderline cases usually as the reasons for disagreements.[8–10] The misclassification of breast diseases may result in under- or overtreatment with important consequences to patients' health.

CAD tools can potentially provide a complementary and objective assessment of histologic features, improving both sensitivity and specificity of the pathologic diagnosis without increasing the workload of pathologists. In recent years, an explosion of studies have been published reporting high accuracy levels of automatic classification of histology images by machine learning algorithms in several disease models.[11–18] In fact, the use of artificial intelligence in DP is seen as the third revolution of pathology, following the introduction of immunohistochemistry (IHC) and molecular pathology.[19]

Our group recently organized an image analysis challenge (Breast Cancer Histology [BACH]) as part of the 15th International Conference on Image Analysis and Recognition (ICIAR 2018) that aimed at the automatic classification of breast tissue histology using H&E-stained microscopy photographs and WSIs.[20] Remarkably, the best performing methods achieved performances similar to those of human experts.

The purpose of the current work is to compare the classification accuracy of the best algorithms of the BACH challenge with the accuracy of a larger group of human observers (including pathologists and pathology residents). In addition, we also aimed to assess whether the output from the algorithms could be used by the observers to improve their classification accuracy.