Jaundice in neonates is caused by elevated levels of bilirubin in the bloodstream. This can cause severe brain damage and even death. The bilirubin levels can rise quickly, therefore they must be measured frequently. Traditionally this is done by taking blood samples, to determine the Total Serum Bilirubin (TSB). This method causes a high strain on the newborn, due to their limited blood volume.
A non-invasive technique is Transcutaneous Bilirubinometry (TcB) measurement. This is done using a point-of-care device. It is not invasive, and cheaper and faster than taking blood samples. TcB devices use two optical paths to determine the bilirubin level.
The downside to TcB measurements is that they are not very accurate. They seem to underestimate the TSB values for light and medium skin colours while overestimating in darker skin colours. The accuracy of TcB measurements decreases for higher levels of serum bilirubin.
Furthermore, the TcB accuracy depends on gestational maturity and measurement of a body location. This means that TcB is not able to completely replace TSB, thus still causing a high strain on the neonate.
A model is desired that can predict the actual TSB values, by using the TcB measurements and some patient characteristics.
A model will be created that can predict the TSB value, using TcBmeasurements and patient characteristics. This is done through machine learning. Multiple programming languages will be considered, such as MATLAB and Python. New programming skills will need to be applied.
The data collection is not part of the scope of this assignment, since a dataset of 101 patients is provided at the start of the assignment.
Different machine learning techniques will need to be researched and applied to develop an optimal model. Statistical analysis will be performed to determine the significance of this model.