Meta Researcher Links Neural Networks to Advanced Scanner Signal Fusion

People think too small

You gather extremely high fidelity signals by running the machine slower or at higher sampling rates. Then you can use those scans as a target prediction from a faster more practical scan of the same patient. You can effectively do a 1 hour scan in 10 minutes. This technology works right now for MRI, it’s not vaporware, we built it. It’s used everyday in clinical practice already.

You can even take MRI and CT of the patients during data gathering and use them as your target during training, essentially allowing you to predict the outcome of those scans from the raw signals coming out of the ultrasound! It won’t be perfect but far beyond what we currently do.

@aaron_defazio Kinda unclear where ML would sit in the acquisition stack though right? Is there some kind of analogy from accelerated MR to how ultrasound gets aquired?

You can use an approach where during training you gather extremely high fidelity signals by running the machine slower or at higher sampling rates. Then you can use those scans as a target prediction from a faster more practical scan of the same patient. You can effectively do a 1 hour scan in 10 minutes. This technology works right now for MRI, it’s not vaporware, we built it. It’s used everyday in clinical practice already.

You can even take MRI and CT of the patients during data gathering and use them as your target during training, essentially allowing you to predict the outcome of those scans from the raw signals coming out of the ultrasound!

People think too small

@aaron_defazio You’d be working on shit data with ultrasound, stick with MRIs

@aaron_defazio If the data is anonymized it can be collected. That is what apple does with its watch data. Next!!!! These fear mongers will literally invent any reason to try and attack this technology.

@aaron_defazio Can you sign a waiver or be paid to share your health data? Free scans in exchange of health data?