Current and future applications of AI in pathology for gastrointestinal cancers
OA Version
Citation
Abstract
The field of pathology has been shifting toward digitalization following the advent of whole-slide image (WSI) scanners. This is helpful for keeping records of samples; however, it also generates an enormous amount of image data that could be used to promote better patient care. Sorting through this data would be incredibly taxing for human pathologists, but it can be done with artificial intelligence (AI).Interest in AI has exploded these past few decades. This has lead to more efficient and precise algorithms being developed almost yearly. Following this trend, researchers have also begun to delve deeper into how AI can be used in medicine. There are many types of AI, including machine learning (ML) and deep learning (DL) with sub-groups including artificial neural networks (ANNs), residual neural networks (RNNs) and convolutional neural networks (CNNs) etc. RNNs and CNNs are by far the most popular DL networks being used in medicine. However, for pathology, CNNs are the more popular form with the most popular architectures being AlexNet/ImageNet, VGG, InceptionNet, ResNet, DenseNet and EfficientNet. The theory behind CNNs was taken from research on animal visual systems and its structure lends itself particularly well to image analysis. This is ideal for pathology as there is a large amount of image data generated.
This thesis will focus on how these popular CNN models have been tested in the context of gastrointestinal (GI) cancers. GI cancers account for a significant portion of annual cancer incidence and mortality rates. This is due to early stages of most GI cancers being non-specifically symptomatic or asymptomatic. Thus, there is a need for more accurate and efficient pathology screening and diagnostic methods for patients at risk. CNN models tested for GI cancer diagnosis or screening have shown great promise and tended to fare just as well if not better than human pathologists. However, there are still many logistical and ethical concerns regarding the use of AI in pathology. Further research into these areas could greatly increase the efficiency and diagnostic accuracy of pathology labs.
Description
2024
License
Attribution-NonCommercial-NoDerivatives 4.0 International