Predicting patient outcome using AI

Written by Alvaro el
Predicting patient outcome using AI

AI in the healthcare industry

Artificial Intelligence (AI) has become one of the most powerful tools for different industries. It is showing a growing interest, driven not only by scientific development but also by the significant capital investment made by the public and private sectors worldwide.

Notably, deep learning and machine learning algorithms have gained prominence in the healthcare sector due to their great potential for supporting the decision-making process.

Machine learning]( (ML) is a branch of artificial intelligence (AI) that focuses on the use of data for algorithm development, seeking to replicate the way humans learn. In this way, the goal is for algorithms to recursively learn the characteristics that allow them to perform an activity and gradually improve the performance they obtain by doing so. However, although it has been shown to be of great help in the health sector for breast cancer detection or new drug discovery, ML is entirely dependent on the ability of a human expert when performing feature extraction, making it cumbersome for the analysis of databases with a great diversity of content.

On the other hand, Deep Learning (DL) is a subset of ML algorithms that is capable of more autonomously finding relevant features for decision making. These algorithms have the advantage of analyzing unstructured data in its raw form, allowing them to be applied in computer vision and natural language processing. Particularly in the case of the healthcare sector, they have been used to support diagnostic image analysis and clinical history review.

Predicting a patient’s outcome with AI

Although a couple of applications that are revolutionizing the healthcare industry have already been mentioned, there is one sector in which AI is attracting particular attention. The predictive capability of AI-based algorithms has been particularly striking in predicting a patient’s outcome when, for example, they enter an emergency room, are diagnosed with a disease, or are sent for referral.

In general terms, the prediction process consists of collecting large amounts of information that may or may not be relevant to the outcome to be predicted. This information, which is already being stored by healthcare systems, may include Electronic Health Records, diagnostic images, or demographic data, among other medical or administrative patient information. With this data, healthcare personnel is responsible for analyzing previously identified and characterized biomarkers related to the outcome to make decisions to preserve the patient’s life. Thus, for example, by understanding the status of the biomarkers, a new pharmacological treatment or the transfer of the patient to a hospital area with greater or lesser care can be defined.

However, in this process of outcome prediction, the data are often diverse and the patients extremely heterogeneous, making this a difficult task even for the best-trained staff. In this regard, AI has demonstrated an enormous capacity to find complex relationships in large volumes of data and to be able to simultaneously and rapidly analyze a large number of variables that allow prediction of outcomes of interest, such as sepsis or mortality. This ability has even surpassed the performance of traditional prediction models used in the clinical setting, as reported by a study published in 2018 in Nature Digital Medicine that sought to predict unexpected readmissions, prolonged hospital stays, and in-hospital deaths.

The Covid-19 case

But this is not the only case. In fact, the power of AI for predictive analytics has been used in diverse applications. These range from machine learning models that accurately predict which patients diagnosed with acute myeloid leukemia will go into remission after treatment of their disease and which will relapse, to algorithms that predict the mortality of patients admitted to Intensive Care Units. Among the latter are the model developed by Arkangel AI in conjunction with the Universidad de la Sabana (Colombia) to predict severe COVID-19 and its mortality based on the interpretation of chest X-ray images and clinical variables. In the last two years, the emergence of the pandemic caused by COVID-19 has strained the resources of the health sector worldwide. With more than 240 million confirmed cases to date and mortality ranging from 2.1 to 55%, SARS-CoV-2 has become a global public health issue that requires all available tools to make the right decisions. In this sense, the development of strategies that allow stratifying patients according to their risk of developing severe COVID-19 and even dying from the disease is vital to define time and resources when providing timely care to patients. The algorithm developed by Arkangel AI and Universidad de la Sabana combines ML and DL tools to analyze clinical data and diagnostic images of patients diagnosed with COVID-19 from the LIVEN COVID initiative organized by several hospitals in Latin America. This algorithm obtained an Area Under the ROC Curve (AUC-ROC) value of 0.92 for predicting ICU admission and 0.81 for predicting mortality in patients admitted to the emergency room. Although the research results are in the process of being published, it is already clear that this will be a tool with great potential at the time of ranking the speed of care given to patients diagnosed with COVID-19 and the type of treatment provided to overcome the disease within hospitals. Finally, it is undeniable that AI applications are gaining ground within the healthcare sector as decision support tools. However, as highlighted by Johan Lundin, Research Director of the Finnish Institute of Molecular Medicine (FIMM) at the University of Helsinki and Professor of Medical Technology at Karolinska Institutet, the future for this type of tool is uncertain. It will be interesting to see what will happen when AI goes beyond what healthcare personnel can do and makes discoveries on its own.

While this happens, in Arkangel AI we would like to invite all agents in the healthcare sector, doctors, nurses, hospitals, and pharmaceutical companies, to get in contact with us to delve into the development of AI tools in search of predicting the outcome that their patients will have and thus improve the decision-making process to always provide the best care and preserve people’s lives.

**If you want to know more about Arkangel Ai software, leave us your (professional information) and one of our agents will contact you to accompany you in a one-on-one onboarding of our technology and advise you on the project you have in mind.


comments powered by Disqus