SYMBIOSIS OF MEDICAL TECHNOLOGIES AND ARTIFICIAL INTELLIGENCE: NEW OPPORTUNITIES IN ONCOLOGY
Keywords:OECI General Assembly
The OECI General Assembly (June 15–17, 2022, Valencia), during which the scientific conference “Artificial Intelligence: A Tool In Modern And Future Oncology” was held, clearly defined the vector of interests of the oncology for the coming years. A bright outburst of interest in artificial intelligence (AI) outside the scientific teams of cybernetic research and engineering laboratories took place against the background of numerous scientific and popular scientific publications, a number of world bestsellers and fantasy films.
My fascination with the prospects of AI capabilities was formed after studying and analyzing a number of scientific publications and recently reading “AI Superpowers: China, Silicon Valley and The New World Order” by Kai-Fu Lee. Familiarity with the possibilities of AI allows us to realize that the surge of interest in its application in the medical field is due to the hope for the emergence of a “magic wand” that is capable of ensuring the emergence of new humanistic elements for the development of personalized medicine in the era of innovative technologies.
The desire to create a virtual product to meet the fantastic needs of medicine permeates almost the entire history of the development of civilization. Even in prehistoric myths, fairy tales and legends, there was a desire to endow created objects or images with the qualities of intelligent subjects. Classical philosophical treatises of Aristotle, R. Descartes and others did not miss this topic; through their vision of a system of mechanical perception of thinking, they described the functional activity of a person. The originator and founder of modern cybernetics and science of control and communication, Norbert Wiener noted that both biological and mechanical control systems are aimed at the implementation of rational purposeful behavior, which in turn ensures the minimization of errors through the system of training and accumulated experience.
Today, it can be said without a doubt that AI, after a difficult and winding path between fantastic optimism and excessive skepticism, has become an everyday part of our lives. Its intervention in the general human consciousness happened quickly and imperceptibly. Recently, this process has acquired a revolutionary character and has captured almost all spheres of human activity.
A clear idea was formed in public conscience that innovations in the field of AI are able to provide the modern service market with new creative developments. It became clear that AI products based on the application of innovative technologies are capable of changing our lives in the near future. Thanks to AI, today it is possible to decipher the structure of any human protein and predict its change in the pathological process.
At the beginning of the 21st century, world leaders in the field of information technology initiated the creation of a number of large-scale projects aimed at solving key problems of modern medicine, including existing and projected challenges in oncology.
The birth of the symbiosis of biological and cybernetic sciences, aimed at processing information with the help of simulation programs, began to be perceived as both an object for research and a subject of the process of learning about the functioning of living things. Such capabilities allowed humanity to simultaneously process large volumes of structured and unstructured databases, generate significantly larger volumes of scientific information necessary for extrapolation of individual features and search for the closest analogue. In general, digital image processing in many areas of human activity has become one of the main directions of scientific and technical progress and is used to increase the efficiency and reliability of decisions made.
At the start of this process, we should not only learn to set a specific task, but also try to understand the range of parameters and algorithms of the process in order to prevent choosing the wrong vector, which will deprive us of the desire to move forward.
The combination of the capabilities of AI tools along with the quantitative and qualitative identification of molecular and structural-functional features of cellular components can be a good example of a successful symbiosis in optimizing the diagnosis of diseases and the choice of treatment.
However, up-to-day, among the evidence-based methods of medical and biological examinations, cytomorphological studies remain the basis of cancer diagnostics.
Solving the problem of improving the quality of research of morphological, histological, and immunohistochemical preparations can occur using computer data processing methods during the analysis of medical images, in particular computer aided diagnosis systems (CAD), which help specialists to interpret medical images. The imaging methods in X-ray, MRI, and ultrasound diagnostics, which have been actively developing since the late 70s, provide a significant additional amount of information that a medical professional analyzes and comprehensively evaluates in a short time. With the advent of whole-slide imaging and machine learning algorithms, CAD also has great potential for use in digital pathology for standard stained preparations. CAD systems process digital images of the entire slide and highlight areas of interest to offer the resulting data to support the decision made by the professional. The use of mathematical methods, computer data processing during the analysis of medical images allow solving the problems of accurate morphological and histological assessment of changes; automation of the processes of morphological research allows to significantly increase the efficiency of research work and achieve more accurate results.
Today, with the use of digital analysis of histological and cytological images, extremely diverse tasks in the biological and medical fields are solved, and despite the existing limitations (the difficulty of obtaining a significant number of identical-quality histological specimens, the high variability of most histological structures), the automation of the microphoto processing allows for objective assessment of images, increased speed of information processing and the accuracy of detecting changes in cancer, and thereby expand the possibilities of histopathological research.
In recent years, numerous experimental and clinical studies have shown that the difficulties of diagnosis, prognosis and treatment of cancer patients are associated with ambiguous clinical manifestations of the tumor process and differences in the morphogenetic characteristics of tumors caused by the etiopathogenetic diversity of neoplasias.
In addition, it is known that the informativeness of biological objects on medical images obtained using optical and electron microscopy is insufficient due to the low contrast of the image of cells and cellular structures, the complexity of the biological organization of tissue structures, the presence in the field of view of various groups of cells, artifacts and significant heterogeneity of tissue as a background. Moreover, the technological equipment of laboratories, the provision of research with high-quality reagents play an important role as well as the human factor (the effectiveness of studying morphological, histological, immunohistochemical preparations, etc. which largely depends on the level of expertise, competence, and experience of the medical staff). It should also be noted that the processing of histological specimens is time-consuming and varies in the quality and types of protocols, methods and reagents in different laboratories, which can complicate the process of identifying morphological characteristics and reduce the accuracy and quality of research.
That is why the improvement of cancer diagnosis and prognosis of the course of the disease largely depend on modern bioinformatics tools with the use of computer automation of image analysis, which is an extremely relevant and dynamically progressive approach. Various types of cluster analysis, learning algorithms of artificial neural network, artificial neural network methods, other mathematical methods as bioinformatics tools, are promising for performing the tasks of analyzing cytological, histological, immunohistochemical, immunocytological images in clinical practice.
Today, in fact, new modern fields of research are emerging, the subject of which is the analysis of cytomorphological images (microphotos), such as digital pathology as image-based environment focused on study and and analysis of histological images, which includes image processing, data extraction (data mining) and database visualization, extraction, retrieval, comparison and management of biomedical knowledge within an array of image collections. In modern clinical practice and scientific laboratories, digital pathology is increasingly becoming a technological requirement of today. The advent of whole-slide imaging, faster networks, and low-cost storage solutions have made it easier for pathologists to manage and share medical digital images for clinical use. In parallel, progress in machine learning allows combining AI and digital pathology, offering new opportunities for accurate diagnosis based on molecular and ultrastructural cytological images.
As for pattern recognition algorithms, a model trained on well-known described examples can be used to classify a new unfamiliar image. Machine learning algorithms applied to automated image classification also require a representative sample of image sets. Due to the high visual variability of histopathological images, a machine learning method usually needs a large number of images from different patients to make successful generalizations.
Unfortunately, AI, having recognized all the key objects in the photo, cannot connect them by correct logic interpretation. Today’s task is the need to unite various “agents” of neural networks, to combine them into a single architectural complex that would reach the level, image and functionality in the imagination of the human intellect. Creating a modular architecture to optimize the functioning of a neural network is a common promising task for biologists and programmers to solve as a joint research team.
The technological successes of such teams will allow AI to go beyond the boundaries of research laboratories and ensure the improvement of the quality of diagnostics and therapy. However, on the way to technology transfer, it is necessary to conduct thorough laboratory and clinical tests to protect potential users from misleading or insufficiently verified results. Organ-and-tissue and intracellular structural-and-molecular diversity requires careful data collection to train an effective and reliable algorithm.
It is worth conducting an in-depth analysis of the identified errors and their differences in frequency and significance. At the same time, we should not forget that the biological system is not a constant, but is significantly dynamic. And here it is worth remembering the popular saying about the devil hiding in the details.
Therefore, a superficial analysis of errors can rapidly change the balance of chances between life and death.
It is this danger that forces a team of specialists to determine the limits of AI application in a timely manner, to identify indicators and protectors that could signal the limitation of its capabilities. AI algorithms provide high reliability only under the conditions of clear verification of quantitative and qualitative indicators of the pathological condition, and if possible, allow finding out their place in the hierarchy and architecture of the development of the malignant process.
Deep fundamental research and analysis of the achieved successes and existing problems with the implementation of AI products obliges us to be careful so that one-sided rhetoric in rose-colored glasses around the phrase “pursuit in the field of artificial intelligence” does not destroy or once again slow down social processes close to Darwin’s evolutionary theory. We must promote scientific and technical symbiosis between representatives of the natural, cybernetic, engineering and technical fields without unnecessary excessive dominant ambitions and in compliance with legal norms and ethical standards. Convergence of ideas and goals, regarding the rapid symbiosis of modern medical technologies and AI in the interests of public values, usually requires new approaches and experiments on the part of the defined road.
That is why specialists of various fields should not remain passive observers. We must seek and find a niche that will allow our ideas about the future of AI to provide solutions to the global problems of carcinogenesis and anticarcinogenesis. We must do everything to reduce the risks of the rapid spread of cancer, and offer the means of prevention, optimization of personalized diagnostics, monitoring of the course of the disease and effective therapy aimed at extending the duration and improving the quality of life of the patient.
An epilogue to the editorial column will be a fitting quote from Steve Jobs in his famous stadium speech on June 12, 2005 to the students of Stanford University: “You can’t connect the dots looking forward; you can only connect them looking backwards. So you have to trust that the dots will somehow connect in your future.”
One can only wish a fair wind to those who move …
On this way, AI, as a director and moderator, will help to combine the interweaving of the hopes of patients and the crazy success of specialists, and will warn against unfortunate mistakes in the area of meeting the needs and capabilities of modern people.
Editor-in-Chief of Experimental Oncology
How to Cite
Copyright (c) 2023 Experimental Oncology
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.