AXiR Engine® realizes Disease Risk Inference at high accuracy by visualization of infection/disease risk and health impediments like patterned grouping or digitalization and repeating the learning of analysis, reference and feedback of heath condition and population data between healthy and pre-disease state people. AXiR Engine® is designed for helping people to get much better healthiness and hybrid type AI engine with perspective of healthiness and disease risk. Healthiness is built by cell metabolism and its level going-down. AXiON Research propose to all of you that “Health Promotion Program” fully optimized for only you can increase QoL referencing health position, lifestyle, its habit, eating one, sleeping quality, exercise and stress condition by adding multi-and-high dimension analysis like multi-parameter analysis, views and correlation from professional people. It handles static data and dynamic data like blood tests and today’s vital data.
of multiple input
Presentation of multiple
options and precision
testing by a medical doctor
team, along with
advice on disease
Malignant neoplasms (cancer, tumors)
Cardiovascular and cerebrovascular diseases
Risk analysis of
and health analysis
organization or group
Spot measurement and
The AXiR Engine® analyzes the process of non-diseased individuals (healthy and pre-disease) shifting towards specific diseases or improving to a healthy state through big data analysis (static analysis), identifying their health positions. If needed, it predicts and estimates immune levels and cellular metabolic levels through vital data, body composition, and medical interviews/health Q&A. For those who want to know their cancer risk and severe infectious disease risk based on their current lifestyle habits, the system traces their conditions for several weeks to months (up to 1-2 years max.) and provides suggestions and recipes for improving health. In particular, multiple traces are used to verify the predicted results through a validation engine and optimize the input replicas using an optimization engine, improving the prediction accuracy through feedback to the Replica Generator (RG-III™). This process dramatically enhances prediction accuracy and evolves the system for future improvements through self-learning processes.
(For early-stage cancer risk prediction, we challenge to predict risks by analyzing data that indicates the cellular state, such as miRNA, Exosomes, and ct-DNA, for a few percent of interested individuals. If you are interested, please contact us.)
It is known that blood data, vital data, and body composition, which indicate various parameters related to immunity and metabolism in the body, show some degree of correlation with the level of health status and disease risk. As the amount of data increases, it is expected that the predictive accuracy of static analysis by big data analysis will improve. However, additional information is needed to fine-tune the predictive accuracy for the relative position and changes in the analysis targets (disease risk prediction estimation) for each health position, if necessary. In particular, if there are clear facts or predictions related to new infectious diseases, diseases, or cancers, correlation processing is carried out in consideration of this. This is especially true for processing in high-risk areas for diseases. There is a report that 70-90% of current lifestyle is related to the risk of diseases such as cancer and infectious diseases. Similar observations have been made for aging and dementia, which need to be monitored.
The Inference Engine is designed to be particularly effective in areas with high disease risk. It provides disease risk scenarios based on information from medical books and papers, as well as the relationship between health and cellular metabolism, immune function (intestinal bacteria), blood vessels, autonomic nerves, hormones from the brain, muscles, bones, lymph, heart, and other organs. The disease risk heat map, transition map indicating disease risk changes, and health Q&A related to lifestyle habits are especially useful when approaching several representative disease risk areas that require attention. This allows for improved prediction accuracy and effectiveness of individual recipes (options) even when the amount of information is limited, which is a challenge for AI processing (Deep Learning). In the future, it is planned to link these individual recipes (options) with the "Health Improvement Promotion Program".
The AI Engine analyzes heatmaps representing health status and disease risk. Some representative disease risks differ in the data items to be analyzed. Each parameter of the disease risk heatmap is arranged in a two-dimensional layout, with standardized scores and correlations among items represented by gradient patterns, enabling the correlation with diseases to be determined to some extent based on this position. In particular, when the area with a strong tendency for disease is present, the results of the inference engine and the AI engine will almost coincide.
As you move away from the disease risk area, you move towards a healthier state, but it becomes technically difficult for the inference engine to demand consistency between this difference and the actual results. The AI engine plays a significant role when moving away from the disease risk area. Just as humans can understand emotions such as joy, anger, sorrow, and fear, as well as anxiety and worry, from facial expressions, the AI engine enables analysis through correlation matrices from health information and disease risk information. In particular, for behavior change towards better health, it is essential to link changes on a Quarterly, Monthly, Weekly, and Daily basis through vital data, body composition, and questionnaires (Q&A), promoting health improvement and disease risk prediction and providing support. Additionally, if there are concerns about the immune system or physical condition, adding biological information such as miRNA can help contribute to the early estimation of cancer risk by referencing the interrelationships between normal cells, immune cells, and cancer cells, and performing dynamic analysis (short-term change tracing). In the future, we will gradually implement tests and trials for supporters and applicants, verifying the effectiveness.
The Verification Engine is used to check and record whether there is a more optimal option among past predictions and proposal candidates when selecting the best one from the two disease risk prediction estimates provided by the Inference Engine (knowledge-based) and the AI Engine (DL type), as well as from the proposals for the "Health Improvement Promotion Program" participants and users. In particular, it will function as an engine to increase the system's reliability by playing a feedback role in the event of a GAP in future prediction estimates and their results. Ideally, this part should be fully automated in the future, but for now, we will adopt a method of improving completion by adjusting it based on the input of experts, designers, stakeholders, and concerned parties.
We have developed RG-Ⅲ™(Replica Generator Ⅲ™) as a software module for expanding data in big data analysis. If there is information that can be used for replica generation by using the Verification Engine's feedback and that can potentially improve prediction accuracy, the Optimization Engine will notify RG-Ⅲ™ of this information. Although we aim for automation in the future, for the time being, our policy is to have experts, designers, and stakeholders intervene and check, minimizing errors as much as possible and increasing accuracy.
The Final Judgment Module performs filtering by prioritizing the Inference Engine (knowledge-based) if the health position of the subject in the disease risk heatmap is in an area where a shift (transition or movement) to a specific disease risk is highly expected. However, as the quality and quantity of data from the AI Engine (DL-based) become sufficient, we would like to gradually change the method to prioritize the AI side. This is because AI has difficulty applying to cases or events with extremely small amounts of learning data, like Black Swans, and there remains uncertainty in high-accuracy predictions of disease risk. For the same reason, we would like to make similar judgments when expanding the target disease risk areas if the learning data is scarce. On the other hand, if there is enough data available and judgments are made based on specific markers or numerical criteria, we believe that AI should be prioritized in areas where clear disease tendencies are not visible.
By utilizing our unique technology such as replica generator and 2D correlation visualization data of health and biological information, AXiR Engine® is able to generate necessary big data from a small amount of data, allowing for accurate identification of health positions and prediction of disease risks with high accuracy. (Patent pending)
Through our unique design, we measure actual data and convert it into deviation values and shift them using correlation coefficients. We then use a two-dimensional correlation matrix to create a heatmap and estimate the correlation distance and health distance between each disease and the target through image processing. This enables us to visualize the "position" and "risk" in the pre-disease area. By inputting health information and biometric data that show changes over time, we can predict disease risk in the future and accelerate the application of this technology.
High-speed patternization and visualization of pre-disease levels The concept of "scientifically approaching health" is considered as the development of solutions to the problem of how to scientifically approach the intermediate process from a healthy state to a state of illness. Therefore, we aim to analyze the trends before feeling suddenly ill or experiencing discomfort during a health checkup or while sleeping at night. We hypothesize that the signs of illness have been prepared long before the onset of the disease. We are challenging ourselves to visualize this process, and an individually optimized "health improvement program" will efficiently utilize this visualization technology. By analyzing the invisible continuous path from a healthy state to a state of illness, we model individual levels of pre-disease to enable rapid pattern recognition and visualization.
AXiR Engine® uses a hybrid system consisting of two engines, an inference engine and an AI engine, to obtain highly accurate results. This is because health data and biological data are complex and simply increasing the amount of data does not necessarily lead to a better understanding of their interactions. The human body has redundant functions and backup systems that can quickly activate in case of emergency, making it difficult to predict health outcomes accurately. By using both supervised and unsupervised learning techniques and verifying and optimizing the results with dedicated modules, AXiR Engine® is able to generate highly accurate predictions of disease risks and health status.
We believe that each person's daily lifestyle has a strong correlation with their health status in the coming weeks, months, and even years. On the other hand, in AI processing, we do not adopt the position that accuracy increases as data increases when predicting specific diseases or health levels by simply inputting vast amounts of data. Each data point has a high likelihood of having a correlation individually, and we believe that it is meaningless to collect data without finding correlated data. On the other hand, when making complex predictions from data that increases every day, we want to adopt a common method. By analyzing and predicting using the same technology for both biometric data that doesn't change much over a certain period of time and health-related data that changes in a few hours to a few days, we can utilize heat maps (static) and transition maps (dynamic), as well as dynamic spider charts (dynamic) to predict changes over long periods (years to months) and short periods (hours to days, weeks) with high accuracy. We believe that this technology is an effective method for promoting "health improvement programs," even for behavior change.
By adding inputs such as disease labeling and measurement data, the AI engine updates its proprietary definition files, enabling early detection of diseases other than specific diseases, identification of cancer factors, and assessment of immune status against infectious diseases. It also allows for predicting correlations and results that may appear random but differ individually. This technology is expected to be applied in various fields beyond health management.