Enhancing Quantitative Image Synthesis Through Pretraining and Resolution Scaling for Bone Mineral Density Estimation from a Plain X-Ray Image

被引:0
作者
Gu, Yi [1 ]
Otake, Yoshito [1 ]
Uemura, Keisuke [2 ]
Takao, Masaki [3 ]
Soufi, Mazen [1 ]
Okada, Seiji [4 ]
Sugano, Nobuhiko [2 ]
Talbot, Hugues [5 ]
Sato, Yoshinobu [1 ]
机构
[1] Nara Inst Sci & Technol, Grad Sch Sci & Technol, Div Informat Sci, Ikoma, Japan
[2] Osaka Univ, Dept Orthopead Med Engn, Grad Sch Med, Suita, Osaka, Japan
[3] Ehime Univ, Dept Bone & Joint Surg, Grad Sch Med, Matsuyama, Ehime, Japan
[4] Osaka Univ, Dept Orthopaed, Grad Sch Med, Suita, Osaka, Japan
[5] Univ Paris Saclay, CentraleSupelec, Gif Sur Yvette, France
来源
SIMULATION AND SYNTHESIS IN MEDICAL IMAGING, SASHIMI 2024 | 2025年 / 15187卷
关键词
Generative adversarial network (GAN); Image-to-image (I2I) translation; Representation learning; Radiography; BMD; REGRESSION;
D O I
10.1007/978-3-031-73281-2_13
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
While most vision tasks are essentially visual in nature (for recognition), some important tasks, especially in the medical field, also require quantitative analysis (for quantification) using quantitative images. Unlike in visual analysis, pixel values in quantitative images correspond to physical metrics measured by specific devices (e.g., a depth image). However, recent work has shown that it is sometimes possible to synthesize accurate quantitative values from visual ones (e.g., depth from visual cues or defocus). This research aims to improve quantitative image synthesis (QIS) by exploring pretraining and image resolution scaling. We propose a benchmark for evaluating pretraining performance using the task of QIS-based bone mineral density (BMD) estimation from plain X-ray images, where the synthesized quantitative image is used to derive BMD. Our results show that appropriate pretraining can improve QIS performance, significantly raising the correlation of BMD estimation from 0.820 to 0.898, while others do not help or even hinder it. Scaling up the resolution can further boost the correlation up to 0.923, a significant enhancement over conventional methods.
引用
收藏
页码:134 / 145
页数:12
相关论文
共 35 条
  • [1] Azampour M.F., 2024, IEEE Transactions on Medical Imaging, P1
  • [2] Brock Andrew, 2021, P MACHINE LEARNING R, V139
  • [3] Buvat I., 2012, Handbook of Particle Detection and Imaging, P1043, DOI [10.1007/978-3-642-13271-141, DOI 10.1007/978-3-642-13271-141]
  • [4] Self-Supervised Learning of Audio Representations From Permutations With Differentiable Ranking
    Carr, Andrew N.
    Berthet, Quentin
    Blondel, Mathieu
    Teboul, Olivier
    Zeghidour, Neil
    [J]. IEEE SIGNAL PROCESSING LETTERS, 2021, 28 : 708 - 712
  • [5] Chen T, 2020, PR MACH LEARN RES, V119
  • [6] Cheng WH, 2021, ACM COMPUT SURV, V54, DOI [10.1145/3552468.3554360, 10.1145/3447239]
  • [7] Deep learning based synthesis of MRI, CT and PET: Review and analysis
    Dayarathna, Sanuwani
    Islam, Kh Tohidul
    Uribe, Sergio
    Yang, Guang
    Hayat, Munawar
    Chen, Zhaolin
    [J]. MEDICAL IMAGE ANALYSIS, 2024, 92
  • [8] Deep learning-enabled medical computer vision
    Esteva, Andre
    Chou, Katherine
    Yeung, Serena
    Naik, Nikhil
    Madani, Ali
    Mottaghi, Ali
    Liu, Yun
    Topol, Eric
    Dean, Jeff
    Socher, Richard
    [J]. NPJ DIGITAL MEDICINE, 2021, 4 (01)
  • [9] Generating Synthetic Labeled Data From Existing Anatomical Models: An Example With Echocardiography Segmentation
    Gilbert, Andrew
    Marciniak, Maciej
    Rodero, Cristobal
    Lamata, Pablo
    Samset, Eigil
    Mcleod, Kristin
    [J]. IEEE TRANSACTIONS ON MEDICAL IMAGING, 2021, 40 (10) : 2783 - 2794
  • [10] MSKdeX: Musculoskeletal (MSK) Decomposition from an X-Ray Image for Fine-Grained Estimation of Lean Muscle Mass and Muscle Volume
    Gu, Yi
    Otake, Yoshito
    Uemura, Keisuke
    Takao, Masaki
    Soufi, Mazen
    Hiasa, Yuta
    Talbot, Hugues
    Okada, Seiji
    Sugano, Nobuhiko
    Sato, Yoshinobu
    [J]. MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION, MICCAI 2023, PT VII, 2023, 14226 : 497 - 507