Comparison of Image Normalization Methods for Multi-Site Deep Learning

被引:6
|
作者
Albert, Steffen [1 ]
Wichtmann, Barbara D. [2 ]
Zhao, Wenzhao [3 ,4 ]
Maurer, Angelika [2 ]
Hesser, Juergen [3 ,4 ,5 ,6 ]
Attenberger, Ulrike I. [2 ]
Schad, Lothar R. [1 ]
Zoellner, Frank G. [1 ]
机构
[1] Heidelberg Univ, Mannheim Inst Intelligent Syst Med MIISM, Med Fac Mannheim, Comp Assisted Clin Med, D-68167 Mannheim, Germany
[2] Univ Hosp Bonn, Dept Diagnost & Intervent Radiol, D-53127 Bonn, Germany
[3] Heidelberg Univ, Mannheim Inst Intelligent Syst Med MIISM, Med Fac Mannheim, Data Anal & Modeling Med, D-68167 Mannheim, Germany
[4] Heidelberg Univ, Cent Inst Sci Comp IWR, D-69120 Heidelberg, Germany
[5] Heidelberg Univ, Cent Inst Sci Comp IWR, CZS Heidelberg Ctr Model Based, D-69120 Heidelberg, Germany
[6] Heidelberg Univ, Cent Inst Comp Engn ZITI, D-69120 Heidelberg, Germany
来源
APPLIED SCIENCES-BASEL | 2023年 / 13卷 / 15期
关键词
normalization; MRI; medical imaging; RECTAL-CANCER; HARMONIZATION;
D O I
10.3390/app13158923
中图分类号
O6 [化学];
学科分类号
0703 ;
摘要
In this study, we evaluate the influence of normalization on the performance of deep learning networks for tumor segmentation and the prediction of the pathological response of locally advanced rectal cancer to neoadjuvant chemoradiotherapy. The techniques were applied to a multicenter and multimodal magnet resonance imaging data set consisting of 201 patients recorded at six centers. We implemented and investigated six different normalization methods (setting the mean and standard deviation, histogram matching, percentiles, combining percentiles and histogram matching, fixed window and an auto-encoder with adversarial loss using the imaging parameters) and evaluated their impact on four deep learning tasks: tumor segmentation, prediction of treatment outcome, and prediction of sex and age. The latter two tasks were implemented as a reference test. We trained a modified U-Net with different normalization methods in multiple configurations: on all images, images from all centers except one, and images from a single center. Our results show that normalization only plays a minor role in segmentation, with a difference in Dice of less than 0.02 between the best and worst performing networks. For the prediction of sex and treatment outcomes, the percentile method combined with histogram matching works best for all scenarios. The biggest difference in performance, depending on the normalization method, occurs for classification. In conclusion, normalization is especially important for small data sets or for generalizing to different data distributions. The deep learning method was superior to the classical methods only in a minority of cases, probably due to the limited amount of training data.
引用
收藏
页数:13
相关论文
共 50 条
  • [21] MULTI-SITE TRIALS: MODELS, METHODS AND OUTCOMES.
    Wyatt, Gwen
    Sikorskii, Alla
    ONCOLOGY NURSING FORUM, 2015, 42 (02) : E123 - E123
  • [22] Multi-Site Diagnostic Classification of Schizophrenia Using Discriminant Deep Learning with Functional Connectivity MRI
    Zeng, Ling-Li
    Wang, Huaning
    Hu, Panpan
    Yang, Bo
    Pu, Weidan
    Shen, Hui
    Chen, Xingui
    Liu, Zhening
    Yin, Hong
    Tan, Qingrong
    Wang, Kai
    Hu, Dewen
    EBIOMEDICINE, 2018, 30 : 74 - 85
  • [23] A Hierarchical Bayesian Model for Multi-Site Diffeomorphic Image Atlases
    Hromatka, Michelle
    Zhang, Miaomiao
    Fleishman, Greg M.
    Gutman, Boris
    Jahanshad, Neda
    Thompson, Paul
    Fletcher, P. Thomas
    MEDICAL IMAGE COMPUTING AND COMPUTER-ASSISTED INTERVENTION - MICCAI 2015, PT II, 2015, 9350 : 372 - 379
  • [24] A Comparison of the Deep Learning Methods for Solving Seafloor Image Classification Task
    Rimavicius, Tadas
    Gelzinis, Adas
    INFORMATION AND SOFTWARE TECHNOLOGIES (ICIST 2017), 2017, 756 : 442 - 453
  • [25] Application and extension of structural reliability methods to multi-site damage
    Millwater, H
    Wu, J
    Burnside, H
    Mear, M
    ICAF 97: FATIGUE IN NEW AND AGEING AIRCRAFT, VOLS I AND II: VOL I: LECTURE PAPERS, VOL II: POSTER PAPERS, 1997, : 671 - 682
  • [26] Benefits and Challenges of Multi-site Trials: Models, Methods, and Outcomes
    Wyatt, Gwen
    PSYCHO-ONCOLOGY, 2015, 24 : 249 - 249
  • [27] Semi-quantitative methods for multi-site deglycosylation events
    Du, Jonathan
    Roberts, Blaine
    Sundberg, Eric
    Deredge, Daniel
    GLYCOBIOLOGY, 2021, 31 (12) : 1733 - 1734
  • [28] Methods in Multi-site Trials of Family-based Interventions
    Duncan, Greg J.
    EMERGING METHODS IN FAMILY RESEARCH, 2014, 4 : 245 - 260
  • [29] Wind speed forecasting using multi-site collaborative deep learning for complex terrain application in valleys
    Qiao, Dalei
    Wu, Shun
    Li, Ge
    You, Jiaxing
    Zhang, Juan
    Shen, Bilong
    RENEWABLE ENERGY, 2022, 189 : 231 - 244
  • [30] Multi-site stimulation
    Chen, Rex Chin-Hao
    Atry, Farid
    Brodnick, Sarah K.
    Novello, Joseph
    Suminski, Aaron
    Pisaniello, Jane
    Williams, Justin
    Pashaie, Ramin
    NEURAL IMAGING AND SENSING, 2017, 10051