Development and validation of a Radiopathomics model based on CT scans and whole slide images for discriminating between Stage I-II and Stage III gastric cancer

Yang Tan ¹, Li-Juan Feng ², Ying-He Huang ³

⁰Department of Pathology, The First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi, China.

Bmc Cancer +

|

March 23, 2024

View abstract on PubMed

Summary

A novel radiopathomics model integrating CT scans and pathology slides accurately predicts gastric cancer stage. This artificial intelligence approach shows promise for improved pathological staging and clinical decision-making in gastric cancer patients.

Area Of Science

Artificial intelligence in oncology
Radiomics and pathomics integration
Gastric cancer staging

Background

Accurate pathological staging is crucial for gastric cancer treatment.
Traditional methods rely solely on postoperative pathological examination.
Integrating preoperative imaging with pathological data may enhance staging accuracy.

Purpose Of The Study

To develop and validate an artificial intelligence (AI) radiopathological model.
To predict pathological staging (Stage I-II vs. Stage III) of gastric cancer.
To utilize preoperative CT scans and postoperative HE-stained slides for model construction.

Main Methods

A cohort of 202 gastric cancer patients was analyzed (141 training, 61 validation).
Pathological features were extracted from HE slides; radiomic features from CT scans.
Machine learning algorithms (SVM, LR, NaiveBayes) were used to build pathological, radiomic, and radiopathological models, with performance assessed by ROC curve and DCA.

Main Results

The optimal pathological model (SVM) achieved an AUC of 0.949 (training) and 0.777 (validation).
The radiomic model utilized 6 SVM-selected features from 1834 radiomic features.
The best performing model, SVM_radiopathomics, combined radiomic and pathomic features, achieving an AUC of 0.953 (training) and 0.851 (validation) with excellent clinical utility.

Conclusions

The AI-driven radiopathomics model demonstrates superior performance in distinguishing between Stage I-II and Stage III gastric cancer.
Combining pathological and radiomic features offers a powerful approach for predicting gastric cancer pathological staging.
This study validates the feasibility of using preoperative CT and pathological slides for AI-based gastric cancer staging.

Keywords:

Computed tomography Gastric Cancer Machine learning Pathomics Radiomics Radiopathomics Whole slide image