Current Developments of Artificial Intelligence in Digital Pathology and Its Future Clinical Applications in Gastrointestinal Cancers

Abstract

The implementation of DP will revolutionize current practice by providing pathologists with additional tools and algorithms to improve workflow. Furthermore, DP will open up opportunities for development of AI-based tools for more precise and reproducible diagnosis through computational pathology. One of the key features of AI is its capability to generate perceptions and recognize patterns beyond the human senses. Thus, the incorporation of AI into DP can reveal additional morphological features and information. At the current rate of AI development and adoption of DP, the interest in computational pathology is expected to rise in tandem. There have already been promising developments related to AI-based solutions in prostate cancer detection; however, in the GI tract, development of more sophisticated algorithms is required to facilitate histological assessment of GI specimens for early and accurate diagnosis. In this review, we aim to provide an overview of the current histological practices in AP laboratories with respect to challenges faced in image preprocessing, present the existing AI-based algorithms, discuss their limitations and present clinical insight with respect to the application of AI in early detection and diagnosis of GI cancer.

Keywords: algorithms; artificial intelligence; cancer diagnosis; computational pathology; deep learning; digital pathology; gastrointestinal tract; histopathology; machine learning; whole-slide imaging.

Figure 1

Tissues obtained by endoscopic biopsies are fixed in 10% neutral buffered formalin for 12–24 h. The fixed tissue undergoes dehydration, clearing and impregnation with molten paraffin wax by automatic tissue processors. The tissue is subsequently embedded in a paraffin wax block with proper orientation so tissue sections (3–5 µm thick) can be cut with a microtome. The tissue section is manoeuvred onto a glass slide, stained and mounted with a coverslip to protect and preserve the section. The glass slide is digitized using high-throughput WSI scanners to create a virtual slide to allow for remote diagnosis and large-scale computational pathology.

Figure 2

Overview of AI techniques and algorithm development in computational pathology for GI cancers. AI is a concept that mimics human intelligence with respect to learning and problem solving. Deep learning is a subset of machine learning; both are techniques used for the development of AI to study the patterns or relationships in WSIs. Deep learning-based techniques are capable of automatic feature extraction, whereas machine-learning-based techniques require manually designed features. Existing software, algorithms and network architectures discussed in this study are summarized in the figure.

Figure 3

Representative H&E-stained sections of different pathologies along the GI tract. (A) Helicobacter-pylori-associated gastritis. Abundant curved rods lining the surface epithelium with underlying mixed inflammatory infiltrate. (B) Moderate number of plasma cells (green arrow), indicating chronic inflammation, and neutrophils (yellow arrow), indicating active inflammation. (C) Gastric adenocarcinoma. Irregular angulated glands lined by tumour cells with enlarged hyperchromatic nuclei, moderate nuclear pleomorphism and frequent mitotic figures. The background stroma is inflamed and desmoplastic. (D) Normal oesophageal squamous epithelium with normal surface maturation. (E) Oesophageal squamous epithelium with high-grade dysplasia involving full thickness of the epithelium. The dysplastic cells exhibit enlarged hyperchromatic nuclei, marked nuclear pleomorphism, loss of polarity and lack of surface maturation. No stromal invasion is observed. (F) Oesophageal squamous cell carcinoma. Irregular nests of tumour cells infiltrating in desmoplastic stroma. The tumour cells exhibit enlarged hyperchromatic nuclei, marked nuclear pleomorphism and frequent mitotic figures. Squamous pearls (yellow arrow) are observed. (G) Oesophageal squamous epithelium with high-grade dysplasia involving full thickness of the epithelium. The dysplastic cells exhibit enlarged hyperchromatic nuclei, marked nuclear pleomorphism, loss of polarity and lack of surface maturation. No stromal invasion is observed. (H) Oesophageal adenocarcinoma. Complex cribriform glands lined by tumour cells with enlarged hyperchromatic nuclei, moderate nuclear pleomorphism and frequent mitotic figures. (I) Oesophageal adenocarcinoma. Poorly differentiated areas with predominantly solid nests observed. (J) Colonic tubular adenoma. Crowded colonic crypts with low-grade dysplasia. The dysplastic cells exhibit pseudostratified, elongated hyperchromatic nuclei. (K) Colon adenocarcinoma. Signet ring cells (arrow) with intracellular mucin that displaces the nucleus to the side. (L) Colon adenocarcinoma. Cribriform glands with extensive tumour necrosis (arrow).