Optimizing Vision Transformers for Histopathology: Pretraining and Normalization in Breast Cancer Classification

被引:4
作者
Baroni, Giulia Lucrezia [1 ]
Rasotto, Laura [1 ]
Roitero, Kevin [1 ]
Tulisso, Angelica [2 ]
Di Loreto, Carla [2 ]
Della Mea, Vincenzo [1 ]
机构
[1] Univ Udine, Dept Math Comp Sci & Phys, I-33100 Udine, Italy
[2] Azienda Sanit Univ Friuli Cent, Ist Anat Patolog, I-33100 Udine, Italy
关键词
breast cancer; deep learning; histology; normalization; transformers; DATA-EFFICIENT;
D O I
10.3390/jimaging10050108
中图分类号
TB8 [摄影技术];
学科分类号
0804 ;
摘要
This paper introduces a self-attention Vision Transformer model specifically developed for classifying breast cancer in histology images. We examine various training strategies and configurations, including pretraining, dimension resizing, data augmentation and color normalization strategies, patch overlap, and patch size configurations, in order to evaluate their impact on the effectiveness of the histology image classification. Additionally, we provide evidence for the increase in effectiveness gathered through geometric and color data augmentation techniques. We primarily utilize the BACH dataset to train and validate our methods and models, but we also test them on two additional datasets, BRACS and AIDPATH, to verify their generalization capabilities. Our model, developed from a transformer pretrained on ImageNet, achieves an accuracy rate of 0.91 on the BACH dataset, 0.74 on the BRACS dataset, and 0.92 on the AIDPATH dataset. Using a model based on the prostate small and prostate medium HistoEncoder models, we achieve accuracy rates of 0.89 and 0.86, respectively. Our results suggest that pretraining on large-scale general datasets like ImageNet is advantageous. We also show the potential benefits of using domain-specific pretraining datasets, such as extensive histopathological image collections as in HistoEncoder, though not yet with clear advantages.
引用
收藏
页数:21
相关论文
共 51 条
  • [1] Alotaibi A., 2023, P 2023 1 INT C ADV I, P1
  • [2] BACH: Grand challenge on breast cancer histology images
    Aresta, Guilherme
    Araujo, Teresa
    Kwok, Scotty
    Chennamsetty, Sai Saketh
    Safwan, Mohammed
    Alex, Varghese
    Marami, Bahram
    Prastawa, Marcel
    Chan, Monica
    Donovan, Michael
    Fernandez, Gerardo
    Zeineh, Jack
    Kohl, Matthias
    Walz, Christoph
    Ludwig, Florian
    Braunewell, Stefan
    Baust, Maximilian
    Quoc Dang Vu
    Minh Nguyen Nhat To
    Kim, Eal
    Kwak, Jin Tae
    Galal, Sameh
    Sanchez-Freire, Veronica
    Brancati, Nadia
    Frucci, Maria
    Riccio, Daniel
    Wang, Yaqi
    Sun, Lingling
    Ma, Kaiqiang
    Fang, Jiannan
    Kone, Ismael
    Boulmane, Lahsen
    Campilho, Aurelio
    Eloy, Catarina
    Polonia, Antonio
    Aguiar, Paulo
    [J]. MEDICAL IMAGE ANALYSIS, 2019, 56 : 122 - 139
  • [3] Baroni G.L., 2023, P IM AN PROC ICIAP 2, P15
  • [4] BRACS: A Dataset for BReAst Carcinoma Subtyping in H&E Histology Images
    Brancati, Nadia
    Anniciello, Anna Maria
    Pati, Pushpak
    Riccio, Daniel
    Scognamiglio, Giosue
    Jaume, Guillaume
    De Pietro, Giuseppe
    Di Bonito, Maurizio
    Foncubierta, Antonio
    Botti, Gerardo
    Gabrani, Maria
    Feroce, Florinda
    Frucci, Maria
    [J]. DATABASE-THE JOURNAL OF BIOLOGICAL DATABASES AND CURATION, 2022, 2022
  • [5] Multi-classification of Breast Cancer Histology Images by Using a Fine-Tuning Strategy
    Brancati, Nadia
    Frucci, Maria
    Riccio, Daniel
    [J]. IMAGE ANALYSIS AND RECOGNITION (ICIAR 2018), 2018, 10882 : 771 - 778
  • [6] CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification
    Chen, Chun-Fu
    Fan, Quanfu
    Panda, Rameswar
    [J]. 2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, : 347 - 356
  • [7] Classification of Breast Cancer Histology Image using Ensemble of Pre-trained Neural Networks
    Chennamsetty, Sai Saketh
    Safwan, Mohammed
    Alex, Varghese
    [J]. IMAGE ANALYSIS AND RECOGNITION (ICIAR 2018), 2018, 10882 : 804 - 811
  • [8] Automatic detection of invasive ductal carcinoma in whole slide images with Convolutional Neural Networks
    Cruz-Roa, Angel
    Basavanhally, Ajay
    Gonzalez, Fabio
    Gilmore, Hannah
    Feldman, Michael
    Ganesan, Shridar
    Shih, Natalie
    Tomaszewski, John
    Madabhushi, Anant
    [J]. MEDICAL IMAGING 2014: DIGITAL PATHOLOGY, 2014, 9041
  • [9] Deep Learning for Whole Slide Image Analysis: An Overview
    Dimitriou, Neofytos
    Arandjelovic, Ognjen
    Caie, Peter D.
    [J]. FRONTIERS IN MEDICINE, 2019, 6
  • [10] Dosovitskiy A, 2021, Arxiv, DOI arXiv:2010.11929