In this episode of the HIPAA Vault Show, Adam & Gil dive into the transformative world of AI in medical imaging. They explore how AI, particularly machine learning and deep learning algorithms, is revolutionizing the field of radiology, pathology, and other imaging-intensive specialties.
From enhanced disease detection and improved efficiency to standardization and predictive analytics, AI is reshaping the landscape of medical diagnostics. However, as with any new technology in healthcare, the implementation of AI raises important questions about data privacy and HIPAA compliance. Adam and Gil discuss the main concerns healthcare providers should be aware of, such as data security, de-identification, audit trails, and patient consent.
They provide a comprehensive overview of the steps providers can take to ensure they’re using AI responsibly and in compliance with HIPAA regulations. The episode also looks to the future, exploring the exciting possibilities of more sophisticated AI models, AI-assisted image acquisition, and the integration of AI imaging analysis with other clinical data. The hosts emphasize the importance of approaching AI implementation thoughtfully and systematically, with HIPAA compliance as a cornerstone of the strategy. This episode is a must-listen for healthcare professionals interested in leveraging the power of AI in medical imaging while navigating the crucial waters of patient data protection and regulatory compliance.
For more information on HIPAA Vault, visit our website!
Do you have any remaining questions, requests, or just want to chat with us? Email us at podcast@hipaavault.com!
Transcript
Adam
Hello, and welcome to the HIPAA Vault show, where we explore all things HIPAA compliance and cloud technology. I’m your host, Adam Zeineddine.
Gil
And I’m Gil Vidals. I’m joining Adam to dive into today’s exciting topic.
Adam
Yeah. And today we’re exploring the transformative world of AI in medical imaging and how it’s reshaping the landscape of diagnostics while navigating the crucial waters of HIPAA compliance.
Gil
Yeah, that’s right, Adam. AI is making waves in everything from radiology, pathology and other imaging intensive fields. But as with any new technology in healthcare, it comes with some challenges, particularly when it comes to protecting the patient data, which, at HIPAA Vault, is something we really focus on.
Adam
Yeah. So let’s start with the basic skill. How exactly is AI changing the game in medical imaging?
Gil
So there are several different things that I want to go over with you today, Adam, and in particular, we’re going to talk about machine learning and deep learning algorithms. These are really revolutionizing what’s happening with medical imaging, and it’s pretty exciting. So there’s enhanced detection. I think it’s the first one we should talk about, where AI can identify subtle abnormalities that might be missed by the human eye and that could lead to earlier detection of a problem like cancer. Obviously, the patient would be wanting early detection. And then there’s improved efficiency because AI can quickly analyze a whole large data set, large volume of images. You know, think about it, the doctors, at the end of the day, you know, it’s beer 30, he needs to head out the door like any other employee. And meanwhile, these machines, I just keep cranking it out.
Gil
And then standardization, the AI is going to be more consistent versus having, if you have ten radiologists, you know, they all might do it a little differently. Well, AI is going to be more standard and then predictive analytics. Some AI models are trained to predict future health outcomes, so they can do this based on the current imaging data they’re looking at and maybe make some predictions that a physician wouldn’t necessarily make. So I think those are some of the areas that we’re going touch on today.
Adam
Yeah, that’s fascinating. Could you give us, like, a specific example where this is applied in real world healthcare settings?
Gil
Sure, sure. So let’s look at breast cancer screening. AI is being used to analyze the mammograms, and they’re able to flag certain areas that may be a concern, and then the radiologist can review that further. So this kind of a model where you have AI taking the first look could be considered a second pair of eyes. So the AI does the first scan and pass on radiologist. He can then confirm or deny what’s being found by the AI. And I think that’s good to have a second pair of eyes.
Adam
Yeah, I can see how that would be incredibly valuable. But it also raises some concerns, I imagine, about data privacy and HIPAA compliance. So what would those concerns be? First of all, for healthcare providers to be aware of when implementing AI in the imaging process.
Gil
Yeah, you’re right. AI offers the tremendous benefits, but it has some unique challenges from a HIPAA compliance standpoint. So the main concerns, I think, include data security, data de identification, audit trails, and patient consent. So starting with the bottom one, patient consent, there may be additional consent requirements. When using AIH, you may have to consider that. Say we use AI, do we need to get the patient’s permission in order to do the diagnostics with AI and then audit trails? Healthcare providers have to maintain the records of how AI is being used in patient care. So you have to audit that, or you have to keep the audit trail of paperwork and then data de identification. If patient data is being used to train ionization AI models, it needs to be properly de identified.
Gil
So you have to deidentify, so you can’t tell whose data that is. And then data security. There’s a ton of data, you know, these images, and that has to be stored somewhere. Storing vast amount of data becomes challenging. It’s not necessarily easy to store terabytes of data, and that has to be properly secure. So I say those are the four areas of concern when it comes to HIPAA compliance.
Adam
Those are really important considerations. How can healthcare providers ensure they’re using AI responsibly and in compliance with HIPAA?
Gil
That’s a solid question. So some of the key areas that I want to mention are, first of all, conduct a thorough risk assessment before implementing the AI system. Then, number two, you want to ensure that you have robust data encryption and access controls are in place. By access controls, we mean, who accessed the data yesterday at 02:00 you have to look at a log so you know everyone is accessing it. Number three, use HIPAA compliant cloud services for processing. Aih. If you’re not going to do it on Prem.
Adam
Yeah.
Gil
Keep in mind just a quick little tangent on this. In the years past, people and companies thought that on Prem was better, and I can tell you that’s very likely untrue. Today, it’s better to more secure in the cloud than on Prem. Now, that’s debatable. Some people might say, but I would do a lot of reading before you jump to the conclusion that because you have a server in a closet is better than having it in the cloud. Implement strong de identification techniques when using your data. So we had a customer that was doing videoing of surgeries, and in the surgery there’s a whiteboard that has the name of the patient, their age, their weight, all this stuff. Very important that on the video they develop a technique to make sure that whiteboard couldn’t be seen blurring out the information.
Gil
So things like that have to be taken into consideration. You want to, you want to maintain the audit log that we talked about already? We mentioned that. And then train the staff, you want to train your staff on proper HIPAA compliance in the context of using AI, and then work with your AI vendors and prioritize, you know, your HIPAA compliance, but talk to your vendors, you’re going to be looping into this process, see what helpful guidance they can give you. So I think I mentioned the seven. Those are the seven I would consider.
Adam
Well, that’s a really comprehensive approach. Gil, are there anything specific? You know, HIPAA is, health insurance portability and Accountability act is a very broad law. Are there any specific rules that are particularly relevant to AI in medical imaging?
Gil
Yeah, the HIPAA compliance is a framework, and that framework consists of three different areas. One is compliance in the administrative area, one is in the physical, and the other is in the technical safeguards. A lot of the ones we’ve mentioned are on the technical safeguards, like encrypting the data. That would be considered a technical safeguard. The administrative safeguard would be, for example, having logs that you’ve trained your staff that you actually can prove. Okay, I’ve trained my staff. Here’s the name of my employee. They were trained at 02:00 on Wednesday. They cover these modules. That’s administrative. And then physical security would be, for example, access coach into your office, webcams in certain areas. Those are the physical constraints. So those are the, that’s the framework that you want to be thinking about when it comes to HIPAA compliance.
Gil
Again, the administrative, physical and technical, make sure you cover all of those.
Adam
I see. And what about the future of AI in medical imaging? Where are you seeing this field heading?
Gil
This field is so wide open, and it’s hard to prognosticate where this is going to go, but obviously we’re only in the beginning stages. I’ve already heard some reports that certain trained models can do better than a radiologist, like they’ll pit the radiologist against the AI imaging, and the AI imaging scores higher. In other words, it catches the problems and sees the issues. So that’s exciting. And the other area would be AI assisted image acquisition, that is, helping technicians capture the best images. And then integration of AI imaging analysis with other clinical data, where you start merging the AI data set output with your other clinical data so you can see the comprehensive diagnostics and then increased use of federated learning. So that means allowing AI models to be trained across multiple institutions without sharing the raw data.
Gil
So a lot of the information in AI has to do with aggregating data. The more data you have, the better you can trained. So if you’re able to work with other institutions and de identify the data and share it, you’re going to have the ability to train your models with larger data sets. So that would be something in the future that associations or federations might try doing.
Adam
They all sound very promising. Any final thoughts or advice for the listeners who might be considering implementing AI in the imaging processes?
Gil
Yeah, even though it’s a very exciting field, Adam, there’s so much potential to benefit the patient. You don’t want to just dive into it. You want to really do a systematic approach. You want to consider your risk assessment. What am I and what are my goals like specifically, what are you trying to accomplish? Are you trying to do it on a particular budget? You’re trying to have better data for your patients, so on and so forth. You need to outline what those are and do it responsibly because you are dealing with patient data. So you want a clear understanding of your goals and then conduct the risk assessment. I would say, yeah.
Adam
Thanks, Gil, for those great comments and insights to our listeners. Remember that while AI and medical imaging offers exciting possibilities for improved patient care, it’s crucial to maintain the focus on privacy and a HIPAA compliance throughout the implementation process.
Gil
Yeah, it is exciting. Adam, I agree with you. These are exciting days. We live in that science and technology is in that exponential growth phase. So we’re seeing drastic changes, not gradual like we did in the past, but drastic leaps of improvements. Now we’re at that phase in history. So it’s super exciting. It’s just a matter of being organizing your approach and taking safety as your prime concern.
Adam
Definitely. Well, that wraps up this episode of the HIPAA Vault show. If you enjoyed it, please give us a like and don’t forget to subscribe. And thanks for tuning in. Stay secure and stay compliant.