So the direct costs, the licensing costs and so forth, and the indirect costs is like the maintenance that I spoke to earlier.

Yeah, maybe just piggybacking on that, Dr. Wald, I mean, If you think about it now from kind of the provider's perspective here, who in many cases are the ones that are evaluating these types of algorithms to provide to their clinicians and whatnot, this becomes a real challenge for them.

As we've already mentioned, all the different hurdles to make sure that these technologies are fully integrated into that routine clinical workflow, evaluated properly, combine all that now and bringing now the topic of reimbursement back into the discussion here, the fact that you have now over 1,500 of these algorithms that are available to potentially procure, yet there's no consistency in terms of the potential reimbursement associated with this.

Yeah, so without that consistency, whereas maybe some are getting reimbursed, others are not, they're getting reimbursed in the many different ways that Dr. Nicola talked about as well.

This makes it very, confusing, quite frankly, for providers to kind of say, wait, maybe should I really dive into this AI thing and actually try to deploy some of these algorithms? So this is unfortunately having an impact in terms of the adoption of this technology overall.

So even though we all see this impact of this technology from a personal perspective in our personal lives here, as we think about how this technology now has the great potential to help patients here, The adoption of this technology becomes a big challenge going forward because of all these hurdles that both of the clinicians have talked about.

So one policy thing I think would be to address the fact that there really is no great avenue for innovation within the physician fee schedule.

Yeah, maybe piggybacking on Dr. Nicola's comments here, from an industry perspective, we're very concerned, of course, that stifling innovation and kind of these new and emerging technologies and how they can positively impact the patient going forward.

So I think Christophe is exactly right that those types of products are what we're seeing actually get reimbursement, starting in the hospital outpatient perspective system, I think, payment system, I think that that will eventually translate into the physician fee schedule with the Category 1 CPT codes.

Yeah, and I was just going to echo both what Dr. Nicole and Dr. Wald have been talking about here.

From an industry perspective, the algorithms that we're creating now going forward, especially in this clinical space, are really kind of focused on algorithm-based healthcare services.

Yeah, so they're really trying to now create that additional piece of information or clinical information that would be otherwise difficult for that clinician to be able to calculate on their own or discover on their own.

So with these different, these healthcare services now being able to generate kind of that additional clinical information that provides that added value that the clinician is looking for to make that more informed diagnostic decision.

But I just want to open the lens that the current way technology is made and marketed is probably not how that's going to be in three to five years from now.

Now, certainly, I think, Dr. Wald, you're absolutely correct that the technology is continuing to evolve.

And I think as Dr. Nicola kind of mentioned, the FDA is also recognizing that this technology is evolving as well.

So Peter, as more multimodal data, as we were talking about, as more of that comes together, such as the imaging, diagnostics, and the clinical context, what role should AI play in reducing the manual coordination across oncology teams while keeping that clinical oversight really central?

Yeah, I think here, Marianne, this is where artificial intelligence can play a significant role.

By being able to consume, again, that imaging data, that patient history, the pathology data, the diagnostic or laboratory data, even genomic information about that patient.

With the help of AI, it can consume all those different pieces of information about that patient, find correlations within all those different pieces of data, and then drive that more informed kind of diagnostic decision or maybe more personalized treatment decision for that particular patient.

And with the help of a technology like artificial intelligence, not only can help us identify potential treatment options for that patient that fits that individual themselves, but then it can also help us in terms of automatically monitoring how successful that treatment is as well.

And as you mentioned earlier, the ability then to adapt as that treatment is happening.

So from a practical standpoint, as we're For instance, applying radiation therapy to a cancer tumor, that tumor changes over the course of time.

It's about having confidence in that entire workflow and really having that data and infrastructure so the physician can make the final decision with the confidence that they need.