TY - JOUR
T1 - Impact of multi-source data augmentation on performance of convolutional neural networks for abnormality classification in mammography
AU - Hwang, In Chan
AU - Trivedi, Hari
AU - Brown-Mulry, Beatrice
AU - Zhang, Linglin
AU - Nalla, Vineela
AU - Gastounioti, Aimilia
AU - Gichoya, Judy
AU - Seyyed-Kalantari, Laleh
AU - Banerjee, Imon
AU - Woo, Min Jae
N1 - Publisher Copyright:
2023 Hwang, Trivedi, Brown-Mulry, Zhang, Nalla, Gastounioti, Gichoya, Seyyed-Kalantari, Banerjee and Woo.
PY - 2023
Y1 - 2023
N2 - Introduction: To date, most mammography-related AI models have been trained using either film or digital mammogram datasets with little overlap. We investigated whether or not combining film and digital mammography during training will help or hinder modern models designed for use on digital mammograms. Methods: To this end, a total of six binary classifiers were trained for comparison. The first three classifiers were trained using images only from Emory Breast Imaging Dataset (EMBED) using ResNet50, ResNet101, and ResNet152 architectures. The next three classifiers were trained using images from EMBED, Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM), and Digital Database for Screening Mammography (DDSM) datasets. All six models were tested only on digital mammograms from EMBED. Results: The results showed that performance degradation to the customized ResNet models was statistically significant overall when EMBED dataset was augmented with CBIS-DDSM/DDSM. While the performance degradation was observed in all racial subgroups, some races are subject to more severe performance drop as compared to other races. Discussion: The degradation may potentially be due to (1) a mismatch in features between film-based and digital mammograms (2) a mismatch in pathologic and radiological information. In conclusion, use of both film and digital mammography during training may hinder modern models designed for breast cancer screening. Caution is required when combining film-based and digital mammograms or when utilizing pathologic and radiological information simultaneously.
AB - Introduction: To date, most mammography-related AI models have been trained using either film or digital mammogram datasets with little overlap. We investigated whether or not combining film and digital mammography during training will help or hinder modern models designed for use on digital mammograms. Methods: To this end, a total of six binary classifiers were trained for comparison. The first three classifiers were trained using images only from Emory Breast Imaging Dataset (EMBED) using ResNet50, ResNet101, and ResNet152 architectures. The next three classifiers were trained using images from EMBED, Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM), and Digital Database for Screening Mammography (DDSM) datasets. All six models were tested only on digital mammograms from EMBED. Results: The results showed that performance degradation to the customized ResNet models was statistically significant overall when EMBED dataset was augmented with CBIS-DDSM/DDSM. While the performance degradation was observed in all racial subgroups, some races are subject to more severe performance drop as compared to other races. Discussion: The degradation may potentially be due to (1) a mismatch in features between film-based and digital mammograms (2) a mismatch in pathologic and radiological information. In conclusion, use of both film and digital mammography during training may hinder modern models designed for breast cancer screening. Caution is required when combining film-based and digital mammograms or when utilizing pathologic and radiological information simultaneously.
KW - breast cancer
KW - cancer screening (MeSH)
KW - CBIS-DDSM
KW - deep learning—artificial intelligence
KW - EMBED
KW - FFDM—full field digital mammography
KW - mammography
UR - http://www.scopus.com/inward/record.url?scp=85183628259&partnerID=8YFLogxK
U2 - 10.3389/fradi.2023.1181190
DO - 10.3389/fradi.2023.1181190
M3 - Article
C2 - 37588666
AN - SCOPUS:85183628259
SN - 2673-8740
VL - 3
JO - Frontiers in Radiology
JF - Frontiers in Radiology
M1 - 1181190
ER -