Digital breast tomosynthesis versus digital mammography: integration of image modalities enhances deep learning-based breast mass classification
To evaluate the impact of utilizing digital breast tomosynthesis (DBT) or/and full-field digital mammography (FFDM), and different transfer learning strategies on deep convolutional neural network (DCNN)-based mass classification for breast cancer.
We retrospectively collected 441 patients with both DBT and FFDM on which regions of interest (ROIs) covering the malignant, benign and normal tissues were extracted for DCNN training and validation. Experiments were conducted for tasks in distinguishing malignant/benign/normal: (1) classification capabilities of DBT vs FFDM and the role of transfer learning were validated on 2D-DCNN; (2) different strategies of combining DBT and FFDM and the associated impacts on classification were explored; (3) 2D-DCNN and 3D-DCNN trained from scratch with volumetric DBT were compared.
2D-DCNN with transfer learning outperformed that without for DBT in distinguishing malignant (ΔAUC = 0.059 ± 0.009, p < 0.001), benign (ΔAUC = 0.095 ± 0.010, p < 0.001) and normal tissue (ΔAUC = 0.042 ± 0.004, p < 0.001) (paired samples t test). 2D-DCNN trained on DBT (with transfer learning) achieved higher accuracy than those on FFDM (malignant: ΔAUC = 0.014 ± 0.014, p = 0.037; benign: ΔAUC = 0.031 ± 0.006, p < 0.001; normal: ΔAUC = 0.017 ± 0.004, p < 0.001) (independent samples t test). The 2D-DCNN employing both DBT and FFDM for training achieved better performances in benign (FFDM: ΔAUC = 0.010 ± 0.008, p < 0.001; DBT: ΔAUC = 0.009 ± 0.005, p < 0.001) and normal (FFDM: ΔAUC = 0.005 ± 0.003, p < 0.001; DBT: ΔAUC = 0.002 ± 0.002, p < 0.001) (related samples Friedman test). The 3D-DCNN and 2D-DCNN trained from scratch with DBT only produced moderate classification.
Transfer learning facilitates mass classification for both DBT and FFDM, and DBT outperforms FFDM when equipped with transfer learning. Integrating DBT and FFDM in DCNN training enhances mass classification accuracy for breast cancer.
• Transfer learning facilitates mass classification for both DBT and FFDM, and the DBT-based DCNN outperforms the FFDM-based DCNN when equipped with transfer learning.
• Integrating DBT and FFDM in DCNN training enhances breast mass classification accuracy.
• 3D-DCNN/2D-DCNN trained from scratch with volumetric DBT but without transfer learning only produce moderate mass classification result.
KeywordsBreast Mammography Deep learning Neural network (computer) Classification
Area under the ROC curve
Digital breast tomosynthesis
Deep convolutional neural network
Double transfer learning
Full-field digital mammography
Mixture of DBT&FFDM
Picture archiving and communication system
Positive predictive value
Recurrent neural network
Receiver operating characteristic
Region of interest
Single transfer learning
Visual geometry group
We gratefully acknowledge all the members of Department of Radiology, Nanfang Hospital, for continuous assistance. In particular, we would like to thank Dr. Weiguo Chen for his advice during the project.
This study has received funding by the National Natural Science Foundation of China (81874216 and 81728016), the National Key Research and Development Program of China (2017YFC0112900).
Compliance with ethical standards
The scientific guarantor of this publication is Professor Linghong Zhou.
Conflict of interest
The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.
Statistics and biometry
No complex statistical methods were necessary for this paper.
Written informed consent was waived by the Institutional Review Board.
Institutional Review Board approval was obtained.
• performed at one institution
- 8.Lång K, Andersson I, Rosso A, Tingberg A, Timberg P, Zackrisson S (2016) Performance of one-view breast tomosynthesis as a stand-alone breast cancer screening modality: results from the Malmo breast Tomosynthesis screening trial, a population-based study. Eur Radiol 26:184–190PubMedCrossRefGoogle Scholar
- 14.Skaane P, Bandos AI, Gullien R et al (2013) Prospective trial comparing full-field digital mammography (FFDM) versus combined FFDM and tomosynthesis in a population-based screening programme using independent double reading with arbitration. Eur Radiol 23:2061–2071PubMedPubMedCentralCrossRefGoogle Scholar
- 21.Wei J, Chan HP, Sahiner B et al (2011) Computer-aided detection of breast masses in digital breast tomosynthesis (DBT): improvement of false positive reduction by optimization of object segmentation. In: SPIE medical imaging 2011, Lake Buena Vista, Florida, United States, 796311:1–6Google Scholar
- 25.Kim DH, Kim ST, Baddar WJ, Ro YM (2015) Feature extraction from bilateral dissimilarity in digital breast tomosynthesis reconstructed volume. In: 2015 IEEE international conference on image processing (ICIP), Quebec City, Quebec, Canada, 4521–4524Google Scholar
- 31.Fotin SV, Yin Y, Haldankar H, Hoffmeister JW, Periaswamy S (2016) Detection of soft tissue densities from digital breast tomosynthesis: comparison of conventional and deep learning approaches. In: SPIE medical imaging 2016, San Diego, California, United States, 97850X:1–6Google Scholar
- 32.Kim DH, Kim ST, Ro YM (2016) Latent feature representation with 3-D multi-view deep convolutional neural network for bilateral analysis in digital breast tomosynthesis. In: 2016 IEEE international conference on acoustics, speech and signal processing (ICASSP), Shanghai, China, 927–931Google Scholar
- 37.Simonyan K, Zisserman A (2014) Very deep convolutional networks for large-scale image recognition. In: arXiv e-prints. Available via https://arxiv.org/abs/1409.1556v6. Accessed 10 Apr 2015
- 38.Tran D, Bourdev L, Fergus R, Torresani L, Paluri M (2014) Learning spatiotemporal features with 3D convolutional networks. In: arXiv e-prints. Available via https://arxiv.org/abs/1412.0767v4. Accessed 7 Oct 2015
- 39.Perez L, Wang J (2017) The effectiveness of data augmentation in image classification using deep learning. In: arXiv e-prints. Available via https://arxiv.org/abs/1712.04621v1. Accessed 13 Dec 2017
- 40.Fitzpatrick JM, Sonka M (2000) Handbook of medical imaging: volume 2. Medical image processing and analysis. SPIE, Bellingham, WashingtonGoogle Scholar
- 41.Seeram E (2010) Digital radiography: an introduction, 1st edn. Delmar Learning, Clifton Park, New YorkGoogle Scholar
- 42.Gonzalez RC, Woods RE (2017) Digital image processing, 4th edn. Pearson, HobokenGoogle Scholar
- 44.Reiser I, Bian J, Nishikawa RM, Sidky EY, Pan X (2009) Comparison of reconstruction algorithms for digital breast tomosynthesis. In: arXiv e-prints. Available via https://arxiv.org/abs/0908.2610v1. Accessed 01 Aug 2009
- 45.D’Orsi CJ, Sickles EA, Mendelson EB, Morris EA (2013) ACR BI-RADS® atlas: breast imaging reporting and data system, 5th edn. American College of Radiology, RestonGoogle Scholar