Hi Reader,

Another week and another PYCAD edition!

This week I would like to share with you 2 things:

• An important insight that every medical professional should know about AI.
• My experience renting online servers to train my machine learning models, and how you can benefit from it.

Without further ado, let's jump right into it!

Medical Professionals Do NOT know This About AI

When I see medical professionals online posting about how they use AI in their work, it’s usually one of 2 categories:

• Either they’re using this big solution that is FDA approved and that it took months or years to integrate it in their hospitals or clinics.
• Or they’re using general purpose AI tools like ChatGPT to help them with some of their repetitive tasks.

What many of these doctors and radiologists do not know, is that there is somewhere in the middle the possibility to use customized AI tools to solve very specific but painful problems that they have.

I especially noticed this when I had calls with many medical professionals who stumbled upon some of our work online.

I’ll give you a couple of examples about this.

1 - Radiologists need to go through several CT or MRI scans a day where they skim through hundreds of imaging slices and they annotate them for some specific purpose like generating a 3D structure that can be shared with surgeons for pre-surgical planning for example.

2 - Many doctors need to have some estimates of different metrics regarding the patient. Things like: how deep should the implant go inside a patient’s mouth based on the patient’s jaw bone structure? Or what is the estimated volume of the tumor that was found in the liver? …

Both of these can be automated by AI.

​

You may not find in the market a tool that does only that one or 2 things. You might find tools that do this, but they are big and expensive and they do 100 other things.

So if a medical professional just wants this one feature to solve one particular problem, they think such solutions do not exist.

This is where they’re mistaken.

In fact, these solutions can be made and tailored to your specific needs.

I am saying this from experience because we at PYCAD have helped clients do these kind of things before.

And it doesn’t need to be this big app with a ton of features. It can be a simple desktop app with 1 or 2 buttons or it can be an extension inside 3D slicer, etc.

If you're a medical professional and you'd like to know more about our services, then feel free to reach out to me by directly responding to this email. I read every single email!

My Experience Renting Online Servers to Train AI Models

To train my machine learning models for medical imaging purposes, I have rented cloud servers from Google, Amazon, Azure, Lambda, VastAI and RunPod.

Here’s my ranking from best to worst.

• RunPod: very easy to set up and cheap.
• Google: servers not always available but when they are, the cost and performance are there.
• Amazon: servers are available, performance is there, but if you blink you might have a heart attack from the cost.
• Lambda: good prices, servers are not as available as someone would hope, servers have good performance, but it can be a nightmare to give them your money. I mean, I am giving you information about my credit cards and you keep declining them for some stupid reason. Why?
• Azure and VastAI: I couldn’t decide which one is worse so I gave them the same ranking. Why are they at the bottom of the list? Because you can expect some unexpected and very weird behavior.