Deep learning detected histological differences between invasive and non-invasive areas of early esophageal cancer

Abstract

The depth of invasion plays a critical role in predicting the prognosis of early esophageal cancer, but the reasons behind invasion and the changes occurring in invasive areas are still not well understood. This study aimed to explore the morphological differences between invasive and non-invasive areas in early esophageal cancer specimens that have undergone endoscopic submucosal dissection (ESD), using artificial intelligence (AI) to shed light on the underlying mechanisms. In this study, data from 75 patients with esophageal squamous cell carcinoma (ESCC) were analyzed and endoscopic assessments were conducted to determine submucosal (SM) invasion. An AI model, specifically a Clustering-constrained Attention Multiple Instance Learning model (CLAM), was developed to predict the depth of cancer by training on surface histological images taken from both invasive and non-invasive regions. The AI model highlighted specific image portions, or patches, which were further examined to identify morphological differences between the two types of areas. The 256-pixel AI model demonstrated an average area under the receiver operating characteristic curve (AUC) value of 0.869 and an accuracy (ACC) of 0.788. The analysis of the AI-identified patches revealed that regions with invasion (SM) exhibited greater vascularity compared with non-invasive regions (epithelial). The invasive patches were characterized by a significant increase in the number and size of blood vessels, as well as a higher count of red blood cells (all with p-values <0.001). In conclusion, this study demonstrated that AI could identify critical differences in surface histopathology between non-invasive and invasive regions, particularly highlighting a higher number and larger size of blood vessels in invasive areas.

Keywords: artificial intelligence; epithelium; esophageal cancer; submucosa; surface histomorphology.

FIGURE 1

Difference in artificial intelligence (AI) accuracy by image size. (A) Three different image sizes (128, 256, and 512 pixels) were used for AI training. Comparison of AI inference using test images. (B) Average values for the area under the curve (AUC) and (C) average accuracy (ACC) estimations, using three image sizes. ROC curves generated 10 times for each image size (D).

FIGURE 2

Representative examples of patch images extracted by artificial intelligence. (A) Non‐invasive area. (B) Invasive area.

FIGURE 3

Morphological analysis of blood vessels. (A) Estimation of the number of red blood cells (RBCs) (left) and blood vessel area (right). Comparison of (B) the number of RBCs and (C) blood vessel area in regions with (submucosal, SM) and without (epithelial, EP) tumor invasion.

FIGURE 4

Morphological analysis of tumor‐cell nuclei. (A) Description of the measurements used to calculate the area of tumor nuclei area, maximum and minimum diameters, and circumference. Comparison of characteristics in cases without (EP) and with (SM) invasion. (B) Area of tumor nuclei. (C) Maximum diameter. (D) Minimum diameter. (E) Circumference.

FIGURE 5

Inference for lamina propria mucosa (LPM) and muscularis mucosa (MM) cases using the learning model with the highest accuracy. (A) The probability of being determined as epithelial (EP, without invasion) in each depth. (B) Most noteworthy patches are divided between submucosal (SM) and epithelial (EP).