Our Work
This is where you will find some information about some of the latest projects we are working on within the company.
01
Fundus Ageing Clock
Fundus images have the advantage of being easy to acquire, fast, relatively non-invasive and without the use of ionising radiation.
Fundus photographs are commonly taken by opticians or eye specialists. These images are used to diagnose a variety of diseases (e.g. diabetes or macular degeneration).
​
Taking a fundus image, we are able to calculate your biological age based on our propriety deep learning algorithms on large population-based data (~49,000 subjects) with associated clinical data, and long-term follow-up.
Our fundus ageing clock was able to achieve an MAE of ~4.2 years.
We also validated our model against 10 diseases and demonstrated that accelerated agers were highly associated with these diseases such as stroke, myocardial infarction, diabetes, and these features were present even before the onset of diseases.
Currently, we are under the process of performing external validation on local cohorts in Hong Kong.
02
Body Composition Profiling
Body Composition Profiling – Opportunistic Quantification from CT and MRI scans
​
Body composition is an important indicator of overall health. Knowing your body fat percentage, muscle mass, and bone density can help you assess your risk for chronic diseases such as diabetes, heart disease, and osteoporosis.
​
Knowing your body composition also allows you to monitor changes in your body. Tracking your body composition over time can help you monitor changes in your body, such as changes in muscle mass or body fat percentage. This can be especially helpful if you are trying to lose weight or gain muscle.
​
Vast amount of quantitative information can be derived from cross-sectional imaging studies such as CT and MRI scans, but most of which currently are not reported in the radiology reports, such as organ volume, visceral fat, muscle volume, and bone density.
​
With the advent of deep learning segmentation models, these can be done accurately and reliably giving more information on these imaging modalities.
Here at Snowhill Science, we have developed proprietary automated segmentation pipeline that can be applied to CT and MRI scans, and generate automated reports so that these quantitative information can accompany traditional radiology reports.
03
Wearable Activity Tracker
Wearable activity-tracking devices are in common use these days, with the advent of smartwatches and smartphones enabling us to monitor our activity on a daily basis.
​
With this lies an enormous opportunity for health screening and optimisation. We have a large amount of scientific data that shows that inactivity or sedentary lifestyles are increasing, which comes with increased risk of developing chronic illnesses.
​
Here at Snowhill Science, we are working to develop proprietary software that will not only allow you to track but also able to gain insights and risk stratify your individual risk for various chronic diseases. We leverage large population data for association and predictive analysis using machine learning.