Tuesday, May 28, 2024

Artificial Intelligence Webinar Questions & Answers


On Monday May 20th, I was honored to present a webinar on the Viva Learning platform.  (Disclosure: The event was underwritten by Philips Sonicare.). The Viva Learning platform allowed for one hour of content and unfortunately I didn't have any time left to do a Q&A with the attendees.

However, before signing off of the event, I copied all of the questions submitted via the Zoom chat feature and saved them.  I told the attendees that I would answer the questions on the blog, so here they are.  If you didn't happen to submit a question, feel free to submit them below in the "Comments" section and I'll do my best to answer them at a future date.

  • Question:  Can "AI data" be used in court?  Is the law keeping up with technology and are insurance companies using AI to adjudicate claims?
    • Answer:  I do some expert witness work in the dental malpractice field.  I haven't come across AI being used in a court of law yet, but I certainly think it will be.   One of the reasons I feel I haven't seen it yet is because, in my experience, it seems to take a really long time to get a case to court.  Cases can take years before an actual trial occurs.  Since AI has really only gone 'mainstream' since 2022-2023, nothing I've worked on has included AI.  That being said, I certainly think in the future AI *will* be seen in court cases.
    • Answer:  Yes, third party payers are beginning to use AI to adjudicate claims.  It doesn't appear to be widespread, but it is already happening.  My concern in this area is currently focused on who is creating the AI network and providing the data.  If a third party payer is using *the same data* that doctors are using, I don't currently see a problem with that.  However, I CAN envision a database being created and used that is trained to deny.  That would be a worst case scenario.  My feelings currently are that third party payers need to justify their systems through an unbiased third party audit.  I'm not aware of that happening yet.

  • Question:  What about DEXIS for CBCT?  I think it is called "DTX"
    • Answer: Yes, DEXIS has a nice product.  It is called "DTX Studio" and it has some really great features.  I have 'tinkered' with DTX Studio as an evaluation and I thought it was a good product.  If you are looking for an AI solution for your practice, I think it should be on your list to take a look at.

  • Question:  Out of curiosity, is DiagnoCAT applying for FDA "clearance" or "approval"?
    • Answer:  I don't have a lot of information on that one.  I have looked at the product and worked with some of their test cases.  They allowed me to have access to a limited number of cases so that I could see what they provide and how to use the product.  It's important for me to state here what I am about to say is SPECULATION.  I *think* they are working on FDA 510(K) Clearance, but I cannot state that as fact.  Probably the best thing to do is to reach out to the company directly, if you are interested is where they are in the process.

  • Question:  Does Pearl distinguish between caries and cervical burnout?
    • Answer:  All the systems being used for radiographic interpretation are looking at the images for 'loss of density' but they are also smart and know *where* to look.  The systems *can* make mistakes in areas such as cervical burnout, which is why I always like to emphasize that these systems DO NOT diagnose.  When performing a clinical exam visually, doctors can evaluate a situation using their own diagnostic data that they gain through the exam process.  We use Pearl in our office and I feel it does a really good job of NOT indicating cervical burnout as caries, but the doctor has the final determination of that fact.

  • Question:  Because systems can detect existing RCT, implants, crowns, etc. Can they connect to the office system (i.e. Dentrix, Open Dental, etc.) to chart what is existing for a patient?
    • Answer:  I don't know of a system that can do that currently, but I know companies are working on that.  The biggest road block right now, is the fact that having this feature will require PMS (Practice Management Software) companies to give the AI companies access to their systems.  Most PMS companies are very protective of their systems for a number or reason, chief among them is security.  I see this as a tremendous benefit, but the legal aspects of allowing this access will probably keep this as a 'want' for a lot offices for at least the near future.

  • Question:  How do you implement Pearl or Overjet into the practice?
    • Answer:  Implementation is pretty simple.  We happen to use Eaglesoft and are still on a Client/Sever setup.  When we started, we allowed Pearl remote access to our server which allowed them to install a piece of software on the server.  Once that was done, Pearl uploaded our images into their system.  Now all we need to do is to start Chrome on the workstations and open a tab to log into Pearl.  The radiographs are quickly uploaded (pretty much as they are taken).  When they are uploaded, the patient is then 'clicked on' via the Chrome tab and the images pop up on the screen.  You can then select or deselect the things you would like to see or hide (caries, calculus, periapical radiolucencies, etc.)  We have a tab open to Pearl on every clinical workstation and doctor workstation every day.  I should also add that all of the data is transferred in a HIPAA compliant way.

  • Question:  Does Pearl do any detection in pediatrics?  And other than Pearl and OverJet, are there other AI diagnostics you use or have heard of?
    • Answer:  The first thing I'd like to state here is that FDA Clearances happen frequently in the AI space so it's probably best to check with the company or check the FDA website with specific questions.  My *current understanding* is that Pearl can be used on ages 12 and above while OverJet can be used in ages 4 and above.   I want to emphasize that I am providing this answer based on my current knowledge.  This info may change as I am typing this!  
    • As far as the companies currently providing AI radiographic analysis I'm aware of the following companies: Peal, OverJet, Denti.ai, DentaXr.ai, VideaHealth, DTX Studio, Apteryx XV Web (which uses Pearl), adravision, DentrixDetect, and DiagnoCAT.  This list is probably not inclusive as these are the ones that I thought of as I typed this.  The area of dental AI for radiographic interpreation is a constantly growing and changing industry.

  • Question:  Can you talk about cost?
    • Answer:  While I don't normally like discussing financials, I can give an overview here.  Just so everyone understands, my personal feeling is that if you treat people to the best of your ability and use the best tools you can, a doctor will make a comfortable income.  However, I realize that doctors constantly 'walk the tightrope' between expenses and production.  I often say that you can be the best doctor on the planet, but if you can't stay in business, your skills won't help many people.  So with that in mind, currently the costs associated with AI seem to be under $500  per month the last time I checked.  If you factor in that a 2-surface posterior composite restoration averages around $250, the math is pretty clear that it doesn't take long before the office can see ROI on the investment.

  • Question:  How often is the AI software updated?
    • Answer:  In the 'big picture' every time a user makes a correction to the info they are provided the system learns from that.  So, at least in theory, the systems are being updated constantly.  For major changes, in my experience those are pushed out fairly regularly, maybe once a quarter or so.  The nice thing about these systems being Cloud Based is that when updates happen, the user doesn't have to do or install anything.  If an update has happened over night, the next time you open the tab in your browser, those changes are there for you.

  • Question:  Any HIPAA concerns with uploading patient radiographs to AI?
    • Answer:  I feel that the concerns with the connections between an office and the AI company are no greater than for any other data an office is using when it comes to The Cloud... that includes credit card processing and electronic claims processing, just to name a couple more.  Now don't get me wrong... security should *ALWAYS* be a concern when dealing with patient data and I'm glad you asked that question, because it shows you are concerned about that aspect of things.  There are a lot of people in this industry that have made it their life's work to keep data secure and I trust them to do so.

No comments:

Post a Comment