EUVIMEDEUVIMED entdecken
Vergleichsstudie

Cell comparative learning: A cervical cytopathology whole slide image classification method using normal and abnormal cells

Jian Qin, Yongjun He, Yiqin Liang, Lanlan Kang, Jing Zhao, Bo Ding

Computerized Medical Imaging and Graphics · 2024

DOI: 10.1016/j.compmedimag.2024.102427 ↗

Kurz gefasst

Vergleichsstudie in der Zeitschrift Computerized Medical Imaging and Graphics, 2024. Verfasst von Jian Qin und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science and Technology, Anhui University of Technology, Maanshan, China. Electronic address: qinjian@ahut.edu.cn. Hauptthemen laut MEDLINE: Uterine Cervical Neoplasms. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 43 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Heilongjiang Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Automated cervical cancer screening through computer-assisted diagnosis has shown considerable potential to improve screening accessibility and reduce associated costs and errors. However, classification performance on whole slide images (WSIs) remains suboptimal due to patient-specific variations. To improve the precision of the screening, pathologists not only analyze the characteristics of suspected abnormal cells, but also compare them with normal cells. Motivated by this practice, we propose a novel cervical cell comparative learning method that leverages pathologist knowledge to learn the differences between normal and suspected abnormal cells within the same WSI. Our method employs two pre-trained YOLOX models to detect suspected abnormal and normal cells in a given WSI. A self-supervised model then extracts features for the detected cells. Subsequently, a tailored Transformer encoder fuses the cell features to obtain WSI instance embeddings. Finally, attention-based multi-instance learning is applied to achieve classification. The experimental results show an AUC of 0.9319 for our proposed method. Moreover, the method achieved professional pathologist-level performance, indicating its potential for clinical applications.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Uterine Cervical NeoplasmsHumansFemaleImage Interpretation, Computer-AssistedMachine LearningCervix UteriDiagnosis, Computer-Assisted

Stichwörter der Autor:innen

Cervical cytopathology imageDeep learningMultiple instance learningWhole slide images classification

Autor:innen und Institutionen

  • School of Computer Science and Technology, Anhui University of Technology, Maanshan, China. Electronic address: qinjian@ahut.edu.cn.
  • School of Computer Science and Technology, Harbin Institute of Technology, Harbin, China. Electronic address: heyongjun@hit.edu.cn.
  • School of Computer Science and Technology, Harbin University of Science and Technology, Harbin, China.
  • College of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin, China.

Förderung

  • National Natural Science Foundation of China
  • Natural Science Foundation of Heilongjiang Province

Zitieren

APA

Qin, J., He, Y., Liang, Y., Kang, L., Zhao, J., & Ding, B. (2024). Cell comparative learning: A cervical cytopathology whole slide image classification method using normal and abnormal cells. Computerized Medical Imaging and Graphics. https://doi.org/10.1016/j.compmedimag.2024.102427

Vancouver

Qin J, He Y, Liang Y, Kang L, Zhao J, Ding B. Cell comparative learning: A cervical cytopathology whole slide image classification method using normal and abnormal cells. Computerized Medical Imaging and Graphics. 2024. doi:10.1016/j.compmedimag.2024.102427

BibTeX
@article{Qin2024,
  author = {Qin, Jian and He, Yongjun and Liang, Yiqin and Kang, Lanlan and Zhao, Jing and Ding, Bo},
  title = {Cell comparative learning: A cervical cytopathology whole slide image classification method using normal and abnormal cells},
  journal = {Computerized Medical Imaging and Graphics},
  year = {2024},
  doi = {10.1016/j.compmedimag.2024.102427}
}
Kostenlos starten →