Artificial Intelligence in the Pathology of Gastric Cancer

Prognostic prediction of gastric cancer based on H&E findings and machine learning pathomics

AIM

In this research, we aimed to develop a model for the accurate prediction of gastric cancer based on H&E findings combined with machine learning pathomics.

METHODS

Transcriptome data, pathological images, and clinical data from 443 cases were retrieved from TCGA (The Cancer Genome Atlas Program) for survival analysis. The images were segmented using the Otsu algorithm, and features were extracted using the PyRadiomics package. Subsequently, the cases were randomly divided into a training cohort of 165 cases and a validation cohort of 69 cases. Features selected via minimum Redundancy - Maximum Relevance (mRMR)- recursive feature elimination (RFE) screening were used to train a model using the Gradient Boosting Machine (GBM) algorithm. The model's performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), calibration curves, and decision curves. Additionally, the correlation between the Pathomics score (PS) and immune genes was examined.

RESULTS

In the multivariate analysis, heightened infiltration of activated CD4 memory T cells was strongly associated with improved overall survival (HR = 0.505, 95 % CI = 0.342-0.745, P < 0.001). The pathomic model, exhibiting robust predictive capability, demonstrated impressive AUC values of 0.844 and 0.750 in both study cohorts. The Decision Curve Analysis (DCA) unequivocally underscored the model's exceptional clinical utility. In a subsequent multivariate analysis, heightened infiltration of the PS also emerged as a significant protective factor for overall survival (HR = 0.506, 95 % CI = 0.329-0.777, P = 0.002).

CONCLUSION

The pathomic model based on H&E slides for predicting the infiltration degree of activated CD4 memory T cells, along with integrated bioinformatics analysis elucidating potential molecular mechanisms, offers novel prognostic indicators for the precise stratification and individualized prognosis of gastric cancer patients.

Artificial intelligence-based real-time histopathology of gastric cancer using confocal laser endomicroscopy

There has been a persistent demand for an innovative modality in real-time histologic imaging, distinct from the conventional frozen section technique. We developed an artificial intelligence-driven real-time evaluation model for gastric cancer tissue using confocal laser endomicroscopic system. The remarkable performance of the model suggests its potential utilization as a standalone modality for instantaneous histologic assessment and as a complementary tool for pathologists' interpretation.

Real-time Histological Evaluation of Gastric Cancer Tissue by Using a Confocal Laser Endomicroscopic System

BACKGROUND/AIM

The need for instant histological evaluation of fresh tissue, especially in cancer treatment, remains paramount. The conventional frozen section technique has inherent limitations, prompting the exploration of alternative methods. A recently developed confocal laser endomicroscopic system provides real-time imaging of the tissue without the need for glass slide preparation. Herein, we evaluated its applicability in the histologic evaluation of gastric cancer tissues.

MATERIALS AND METHODS

A confocal laser endomicroscopic system (CLES) with a Lissajous pattern laser scanning, was developed. Fourteen fresh gastric cancer tissues and the same number of normal gastric tissues were obtained from advanced gastric cancer patients. Fluorescein sodium was used for staining. Five pathologists interpreted 100 endomicroscopic images and decided their histologic location and the presence of cancer. Following the review of matched hematoxylin and eosin (H&E) slides, their performance was evaluated with another 100 images.

RESULTS

CLES images mirrored gastric tissue histology. Pathologists were able to detect the histologic location of the images with 65.7% accuracy and differentiate cancer tissue from normal with 74.7% accuracy. The sensitivity and specificity of cancer detection were 71.9% and 76.1%. Following the review of matched H&E images, the accuracy of identifying the histologic location was increased to 92.8% (p<0.0001), and that of detecting cancer tissue was also increased to 90.9% (p<0.001). The sensitivity and specificity of cancer detection were enhanced to 89.1% and 93.2% (p<0.0001).

CONCLUSION

High-quality histological images were immediately acquired by the CLES. The operator training enabled the accurate detection of cancer and histologic location raising its potential applicability as a real-time tissue imaging modality.