Mar 19 2021
Anemia is a medical condition that generally occurs when there is a dearth of healthy red blood cells in the human body.
AISACS received a total of five inputs and churned out dosage direction probabilities for erythropoiesis-stimulating agents and iron supplements. It was noted that AISACS sometimes produces “clinically appropriate’ directions that are different from those of physicians. Image Credit: 2021 Okayama University.
Anemia is common in patients who suffer from chronic kidney disease and had to undergo regular hemodialysis—a process that helps 'clean' the blood when the kidneys have impaired functions.
Therefore, iron supplements (ISs) and red blood cell-stimulating agents, known as 'erythropoiesis-stimulating agents' (ESAs), are administered to the patients as part of this process.
However, complications can emerge if the affected patients have a poor response to drugs or have an altered iron metabolism. Added to this, the drugs tend to be costly and impose a huge financial burden on public health.
But today, such patients are growing in numbers and yet a sufficient number of physicians have not been appropriately trained to perform the treatment. As a result, more support systems with intelligent decision-making capabilities are actively sought after.
One alternative is to use artificial intelligence (AI), which appears to be promising but requires a huge dataset. It is also not a viable option because of the different health conditions of patients.
So, what would be the best way to improve this situation? In a new study recently published in the International Journal of Medical Sciences, medical scientists from Japan have attempted to find a solution to this problem.
The researchers developed a novel method—that is, rather than making the AI learn from the intricate physiology of the patient’s body, they used a prediction model built on the decisions of qualified doctors.
We got the idea while contemplating the thought process of seasoned physicians. After all, they do not calculate detailed values of vital reactions in a patient's body when deciding dosages, which means prediction models based on biochemistry are not necessary.
Toshiaki Ohara, Assistant Professor, Okayama University
Ohara is also the lead researcher on the study.
The investigators began their study by gathering clinical data at two hospitals based in Japan and subsequently prepared a couple of datasets for every hospital—one dataset for training the new model, and the other for testing and verifying its predictions.
The researchers simultaneously captured the dosage directions of doctors at both hospitals and considered the reactions for the two drugs, that is, ISs and ESAs, used at the time of hemodialysis.
Depending on these outcomes, the team built an AI-based model known as an 'artificial-intelligence-supported anemia control system' (AISACS). This system received a total of five inputs (four items of dosage history and blood examination) and provided the probabilities of dosage direction for both the drugs as outputs.
To make the training process even more efficient, the researchers also compensated for the time lag between dosage decisions and blood analysis by using “data rectification” to match the examination dates with the decision dates.
To their surprise, the AISACS system demonstrated a high prediction precision with exact classification (directions corresponding with those of doctors) rates of 72% to 87%. But what was more fascinating was that the AISACS system offered “clinically appropriate” classifications at much higher rates of 92% to 97%.
These were classification directions that did not correspond with those of doctors (and were occasionally given ahead of them) but were still believed to be suitable from a medical standpoint.
With these outcomes, the team is hopeful about the future prospects of the AISAC system.
By preventing anemia, our system can help alleviate the burdens on physicians and medical insurance systems. Moreover, it has the potential to share the knowledge and experiences related to medications.
Toshiaki Ohara, Assistant Professor, Okayama University
The new AI-based approach is expected to offer some hope to patients undergoing routine hemodialysis and also to doctors treating them.
Journal Reference:
Ohara, T., et al. (2021) Artificial intelligence supported anemia control system (AISACS) to prevent anemia in maintenance hemodialysis patients. International Journal of Medical Sciences. doi.org/10.7150/ijms.53298.