Bloods.AI - Blood Spectroscopy Classification Challenge

From painful sticks to seamless scans, bloods.ai aims to revolutionise the blood analysis field with the help of artificial intelligence. With the help of data science enthusiasts all around the world, the company’s new competition has the goal of building a machine learning model that would classify specific levels of compounds in the blood sample from spectroscopic data. This means that blood analysis can be done without even piercing the skin, as the content of the blood can be analyzed through light spectroscopy with the help of machine learning that is trained on large numbers of previous samples. Since there are more than 4000 compounds that can be found in human blood, the goal requires very high precision that the competition host aims to find its answer through artificial intelligence.

The data for the competition was collected by analyzing the donated samples 60 times per sample through the methodology called a Near Infrared Spectroscopy. This gathered data will fall between the range of 950-1350 nanometers as it has the highest penetration power through the skin, which will ideally be applied onto the fingertip. This method will direct a beam of light to the skin(some of which would be partially absorbed and/or reflected) and will gather the data on a range of wavelengths to measure the amount of energy absorbed at each level. All light that is reflected is classified into an array of 170 intensities per scan. This will also account for humidity and temperature to see whether these affect the outcome of the process. The biggest issue with this methodology is the risk of overfitting, as each of the samples have very precise data values that can easily push the model to overfit to its training data. 

We are hoping to be one of the pioneers on this new technology with our submissions and help AI pave the way. Have more questions? Feel free to reach us out!

Previous
Previous

Task Mate Kenyan Sign Language Classification Challenge

Next
Next

Jigsaw Rate Severity of Toxic Comments