AI in endoscopy: A new superhero or an old villain?

In the rapidly evolving landscape of technology, artificial intelligence (AI) has emerged as a transformative force, reshaping industries and redefining possibilities. As its influence continues to grow, the medical field stands at the forefront of this revolution, particularly in areas like endoscopy, where AI's potential to enhance precision and efficiency is becoming increasingly  evident. The previous two sentences were generated by ChatGPT, one of the most famous AI resources nowadays. With the right commands, called prompts, we can achieve long and complex tasks that would normally take hours within minutes. There are countless activities that can now be streamlined or even totally performed by AI tools, ranging from trip itineraries to PhD theses. Also, several examples in the medical field have been taking place: early detection of malignancies, quick remote patient assistance, robots operating during surgery, etc. Of course, the endoscopy field would not be on this list, and brand new studies are coming to improve both theoretical and technical skills. However, specialists are worried about how far AI might go. Are there any possible harms AI could bring to humanity? Should we define limits before things are out of our control?

If you have never seen before the video based on an article by Igor van Gemert about the 10 stages of AI, I strongly advise you to do it. We can barely understand the strength it has and how helpful it can be for endoscopic procedures if used wisely. AI-aided systems can improve adenoma rate detection by almost 25%, according to a recent meta-analysis, and deep-learning algorithms are leading to better detection of early esophageal and gastric cancers. Capsule endoscopy imaging requires a great deal of time to be correctly interpreted, which can be bypassed with AI-reading systems, helping to identify small changes that even a trained gastroenterologist would not notice. Specific details that usually cannot be precisely defined by human eyes, such as villous atrophy in celiac disease or initial flares in inflammatory bowel diseases, are now easily detected with AI technology. What about computer high-tech systems that can teach you advanced endoscopic techniques with endless classes on virtual reality platforms without the risk of adverse events, giving simultaneous feedback on your performance? And have you ever imagined an algorithm that takes a look at the pictures of a procedure and elaborates a report within seconds? These scenarios are already a reality with AI and machine-learning tools. 

But it is not always rainbows and butterflies, and the increasing presence of AI in people’s lives has raised some important ethical issues. First, the internet is a powerful tool, but it is also filled with misinformation. Considering AI systems learn from pre-established knowledge, it is possible that they use non-representative training data, leading to misinterpretations and unfair decisions. The UNESCO website provided a guideline with recommendations on the ethics of AI, bringing some nice examples to illustrate a few dilemmas. Type on your computer, “List the greatest leaders of all time." How many women are described? Or how many black individuals appear on this list? This might affect AI reliability as well, and defining some filters in addition to preparing regular system audits can help to avoid biases. Particularly in the medical scenario, data security is another major concern since there is an unbridled collection of several private characteristics (have you ever wondered where our facial recognition is stored?). Besides, the more we set tasks for AI systems, the more we need to clearly define accountability—if AI makes a wrong diagnosis, who would be responsible for that mistake? Finally, in the global warming era, we have to consider the impact of AI on our environment. The creation and development of such complex algorithms are highly costly and require a great quantity of energy, increasing gas emissions and the production of electronic waste. 

An AI-based routine is a way with no return, and several challenges are waiting for us. AI will not set aside effort and work; it will help us speed up complex tasks and, thus, provide greater self-development. AI and machine learning are not required when endoscopists face obvious changes, but when there is doubt about subtle alterations or therapeutic options. Nevertheless, the current problems we have to deal with remain the same as those we have always faced, but with a fancier technological appearance. Humanity has already fought against old villains in the past, such as misinformation, career adaptation, and data protection, and getting out of our comfort zone is necessary if innovation is a goal. With the correct approach, AI has everything it takes to become our newest superhero.