AI's Diabetes Advice: A Model Comparison Study

In this episode, our hosts take a deep dive into the Nine Sensor Samba, an n=1 experiment that pits a variety of newly released CGM systems against the incumbents, Dexcom and Libre.With 9 sensors, 15 days and 150 fingerpricks, the outcomes aren't quite what you're expecting!

In this episode, our hosts discuss the review of Snaq undertaken over at Diabettech, noting the benefits and frustrations of using it, and the conclusions the reviewer drew.The two articles this is drawn from are:https://www.diabettech.com/ai/snaq-a-diabetes-ai-to-help-you-carb-count-badly/https://www.diabettech.com/diabetes/beyond-snaq-qing-digesting-your-glucose-patterns/

In this episode, our hosts discuss the attempts that LLMs made to analyse a series of images describing outcomes and pump settings, highlighting the benefits and issues that arose in the experiment. Suffice to say that it wasn't all sweetness and light and they won't be replacing healthcare professionals just yet.

Our hosts discuss how Large Language Models manage when presented with images of foods to analyse and determine the carb content. It turns out that they are about as good as humans...

In this episode, the twins discuss the study that Diabettech undertook to see how the various AI models handled facts, tips, etc. There's obviously no irony in this approach...

Our AI hosts discuss the outline of a study comparing how well AI systems give diabetes advice.

With all the information demonstrating that Incretins appears to shield against heart problems, our hosts discuss whether they could help alleviate the complications of diabetes.

A discussion focusing on Medtronic spinning out its Diabetes division, and with Ypsomed doing the same, what it means for the Diabetes Devices industry...

Our favourite hosts discussing the advantages and disadvantages of the "more accurate" and "more data" CGMs on the market at the moment and draw some interesting conclusions. Discussion generated by Google Gemini AI.

In this episode, Mac and Lin discuss infusion site locations and the potential effects on AID systems of changing between abdomen, arms and thighs. it's worth noting that they haven't quite understood that the predominant study referred to in the article used a rapid acting insulin rather than regular human insulin to arrive at the numbers provided. Original article at: https://wp.me/p7O2EL-2uK