In my last article, I opened the topic of the contribution of artificial intelligence in the field of medicine. According to many experts, it is one of the most fruitful areas where the fastest advancements can be made. Some of the possibilities I mentioned that the application of this technology will bring are: rapid and accurate diagnostics, improvement of doctor-patient relationships, improvements in logistics, and reduction of administrative costs, and in a very important segment of scientific research.
Currently, the largest market for artificial intelligence in healthcare is North America, which makes up almost 50% of the global market. Projections are that it will grow from about 14.6 billion dollars in 2023 to over 102 billion dollars by 2028. This should not be surprising considering that healthcare is facing huge problems, with a growing number of patients and too few doctors to treat them.
Artificial intelligence can help solve some of these problems, offering advancements ranging from more efficient diagnoses to safer treatments. On one hand, the driver of this development and growth will be the creation of large data sets that will combine different sources, from social media data, readings of existing records, all new data obtained from devices, billing data, or fresh biomedical data. Of course, for these reasons, the quality of medical data will be important, which is still difficult to collect or access when they exist, and the industry’s employees are not yet focused on the process of collecting and storing data at all stages. On the other hand, the limitation will be the lack of AI-educated workforce who should know basic concepts like cognitive computing, machine learning, and machine intelligence, or image recognition.
Large corporations are already in the game (like Microsoft and Johnson & Johnson) and a good example is the collaboration between GE and NVIDIA who have already installed over 500,000 devices for collecting and processing medical images, around the world. The software is made to recognize organs and identify lesions in some areas, through available scans, and based on them automatically generates reports that help in the efficiency and precision of diagnoses. Hospitals in Europe and the US are already using some systems to help hospitalized patients and enable them to easily schedule examinations. The good news is that there is increasing acceptance of new technologies, especially EMR (Electronic Medical Records) by institutions, which helps in collecting the aforementioned data. I was pleased to see that there are startups in Croatia that are solving some of the challenges in this area — such as the startup Meddox, an application that allows storage (and centralization) of medical findings to always be available.
One of the very promising areas is at the intersection of AI and genetics. It is believed that in the next decade a large part of the world’s population will be offered complete genome sequencing, either at birth or in adulthood. Genome sequencing could generate 100–150 GB of data and provide a powerful tool for precision medicine. The connection of genetic and phenotypic information is still ongoing. The current clinical system should be redesigned to take advantage of such data and their benefits. Healthtech companies like Deep Genomics analyze samples in huge genetic data sets and EMRs, linking them with respect to disease markers. They use these correlations to identify therapeutic targets, either existing or new candidates, with the aim of developing individualized genetic drugs. They apply AI at every step of drug discovery and development, including target discovery, lead compound optimization, toxicity assessment, and innovative trial designs. Many hereditary diseases result in symptoms without a specific diagnosis, while the interpretation of whole genome data still presents a challenge due to many genetic profiles. Precision medicine can enable methods to improve the identification of genetic mutations based on complete genome sequencing and the use of AI.
Significant acceleration could also be achieved in the field of drug discovery and development, which I mentioned last time. Drug discovery and development are extremely long, expensive, and complex processes that often take more than 10 years, from the identification of molecular targets to approval and market placement. Each failure in this process has a large financial impact, and most drug candidates actually fail during development and never reach the market. In addition, there is an increasing number of regulatory obstacles and difficulties in the continuous discovery of drug molecules that are significantly better than currently available. This makes the process of drug innovation challenging and inefficient, with a high price of new drugs that manage to reach the market. In recent years, there has been a significant increase in the number of data on the activity of drug compounds and biomedical data. This is due to increased automation and the introduction of new experimental techniques, including speech-to-text based on the hidden Markov model and parallel synthesis. However, it is necessary to mine the data of large chemical sets in order to effectively classify potential drug compounds, and machine learning techniques have shown great potential.
And with that, we have only scratched the basic areas of applications. Perhaps in a new column, I will mention some of the projects related to health monitoring, wearable devices, and other devices used in medicine that are managed by artificial intelligence.