Predicting Body Heat: A New Way with AI

Trying to figure out the temperature inside someone's body just by feeling their skin. That's what scientists are doing with a technique called 3D thermal tomography. They're using a special kind of artificial intelligence, called a 3D convolutional neural network (CNN), to predict how hot or cold it is inside our bodies. This isn't easy, especially since our bodies have tiny temperature differences. So, the scientists tested their CNN under different conditions, like adding noise or changing the background temperature. They even used a special math rule called the heat equation to help the AI learn better. The results? The AI did great with small objects, like a 10 cm ball. But when things got bigger or noisier, the AI struggled, especially in deeper parts of the object. However, when the scientists added more rules based on physics, the AI got much better at finding hot spots deep inside. This shows that mixing deep learning with real-world rules can make non-invasive thermal imaging more reliable, even in tough situations.