JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Identification of histopathological classification and establishment of prognostic indicators of gastric adenocarcinoma based on deep learning algorithm

  • 0Department of Ultrasound Imaging, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430101, Hubei, China.
Medical Molecular Morphology +

|

August 1, 2024

|

August 1, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

This study developed a deep learning model to accurately predict gastric cancer subtypes from whole-slide images, achieving over 90% accuracy. The model aids in clinical diagnosis and prognosis by analyzing immune infiltration and patient outcomes.

Area Of Science

  • Oncology
  • Computational Pathology
  • Artificial Intelligence in Medicine

Background

  • Gastric adenocarcinoma (STAD) is a significant global health concern.
  • Accurate pathological classification is crucial for effective STAD treatment and prognosis.
  • Whole-slide images (WSIs) offer rich data for computational analysis.

Purpose Of The Study

  • To develop and validate a deep learning (DL) model for predicting pathological subtypes of STAD using WSIs.
  • To assess the model's generalization ability through external validation.
  • To explore DL-extracted features for insights into immune infiltration and patient prognosis.

Main Methods

  • Utilized 356 STAD histopathological WSIs from The Cancer Genome Atlas (TCGA) for training, validation, and testing (8:1:1 split).
  • Incorporated 80 external H&E-stained STAD WSIs for validation.
  • Employed the CLAM tool for WSI segmentation and applied a DL algorithm for classification.

Main Results

  • The DL model achieved over 90% accuracy in predicting histopathological subtypes of STAD.
  • External validation confirmed the model's generalization capability.
  • DL features revealed differences in immune infiltration and patient prognosis between subtypes.

Conclusions

  • The DL model accurately predicts STAD pathological classification, offering clinical diagnostic value.
  • The model's extracted features provide insights into STAD biology and patient outcomes.
  • A nomogram combining DL-signature, gene-signature, and clinical features serves as a prognostic classifier for clinical decision-making.

Keywords: