Applications of artificial intelligence and deep learning in colorectal cancer surgery – Correspondence

Abstract

Dear Editor, The advent of advanced computer technologies has significant benefits for numerous fields and where the adoption of technology increases, there is inevitably savings in time and efficiency. This is particularly the case with the application of Artificial Intelligence (AI) and machine learning technology and their impact on colorectal surgical procedures. AI technologies are already playing a significant role in medicine by supporting clinical decision-making, imaging and diagnosis, drug discovery and genomics [1]. As AI began to first be introduced into surgical procedures, the technologies were based upon improving imaging and navigation. Further techniques also began to be developed that resulted in novel ways to detect features during the pre-operative planning stages of a procedure. As a result, there is a gradual transition in the practice of surgery with significant developments and advances in imaging technologies, navigation and the introduction of surgical robots [2] over the past 20-years. When we consider the practice of the diagnosis and surgical intervention of colorectal cancer. From a contemporary stance, a surgeon must operate safely on a patient using fluorescence during endoscopic examination, to interpret and identify different types of tissues. As with all visual observations, it can often be problematic to distinguish tissue features that are inside the human body. This is particularly the case when viewing these structures that are often in motion through an endoscopic camera. The initial challenge of using AI to classify cancerous and distinguish between health and diseased tissue is in being able to use real-time vision data. This needs to be transformed into observations that can be analysed, interpreted and then subsequently used to train an AI neural network. Subsequently, this data can be used to categories a tissue type and map the morphology of the tumour structure. The end objective is to transform data into something that is able to precisely identify diseased tissues and differentiate it from the healthy tissue. Deep learning convolutional neural networks have previously been used to determine mortality and postoperative bleeding after cardiac surgery and to predict renal failure in real time. As a result of using this new technology, this has resulted in significantly improved outcomes in comparison to that of standard clinical reference tools. This brave new field of enabling machines to identify and classify the observed biological properties is called computer vision-based learning. This in conjunction with convolutional neural network-based machine learning is a very powerful technology in surgery. As shown, Fig. 1 is a diagram that explains from a practical context how this process works when identifying and mapping a tumour structure. The advances in AI and computer vision can be used to transform fluorescence into a powerful clinical tool.Fig. 1.: A practical vision-based convolutional neural network developed by the authors using the python programming language. This has been used to demonstrate the means in which machine learning is used for identifying regions of cancerous tissue in the colon. A boundary map and grid is automatically placed around the cancerous tissue in real time using this technique.The process starts with training data, which is used from the context of (i) object detection, (ii) categorisation and (iii) classification. A convolutional neural network sees real-time endoscopic video as grids of pixels. Each of these pixels is represented using a combination of red, green, and blue, which subsequently have differing degrees of intensity. The image is subsequently changed to grayscale so each set of pixels is compared to that of the pixels used in the training data [3]. This is carried out to determine and find the changes in the tissue due to disease. This process involves analysing variations in tissues based upon past observations made by surgeons tog