[Intl_DxMedPhys] FW: Diagnostic Physics work and AI

[Alisa Walz-Flannigan]

MP’s not ignoring PACS or informatics now or in the past:

AAPM Imaging Informatics Subcommittee<https://urldefense.com/v3/__https://www.aapm.org/org/structure/default.asp?committee_code=IISC__;!!KGKeukY!3pPa1u5IApVNHHsbbLSXhk_jlO5pbM84ApgWIs5iysr8fbuHt2XjTQHkkZS1ZZ1X9FT80RA_35RF67IRPAZELURUPgXZjn9hI9l4dfCfxgsMuQ$ >
I’ve passed leadership baton to active and engaged physicists and encourage folk with ideas and energy to contact them with your ideas AND your energy. Here’s one endeavor I helped shepherd some years back:

Interoperability Assessment for the Commissioning of Medical Imaging Acquisition Systems<https://urldefense.com/v3/__https://www.aapm.org/pubs/reports/RPT_248.pdf__;!!KGKeukY!3pPa1u5IApVNHHsbbLSXhk_jlO5pbM84ApgWIs5iysr8fbuHt2XjTQHkkZS1ZZ1X9FT80RA_35RF67IRPAZELURUPgXZjn9hI9l4dfDAH1EBPQ$ >

AAPM Data science committee<https://urldefense.com/v3/__https://www.aapm.org/org/structure/default.asp?committee_code=DSC__;!!KGKeukY!3pPa1u5IApVNHHsbbLSXhk_jlO5pbM84ApgWIs5iysr8fbuHt2XjTQHkkZS1ZZ1X9FT80RA_35RF67IRPAZELURUPgXZjn9hI9l4dfCNQkiU2g$ > and all the subcommittees

And there are very active imaging physicists involved in SIIM and RSNA and ACR in the noted areas of concern. And physicists getting their CIIP, and physicists helping evaluate and test imaging systems for which the knowledge base required is the subject of this thread.


-Alisa



From: Jerry Thomas <jat1946jeep at gmail.com>
Sent: Monday, March 24, 2025 11:25 AM
To: Alisa Walz-Flannigan <walzflanniga at wisc.edu>
Cc: Mary Ellen Jafari <MaryEllen.Jafari at kp.org>; Nicole Ranger <nicoletranger at yahoo.com>; troyxzhou at icloud.com; intl_dxmedphys_wd_osu_list at lists.osu.edu
Subject: Re: [Intl_DxMedPhys] FW: Diagnostic Physics work and AI

To all:

     Mary Ellen has raised a very important issue that we all need to pay close attention to.   Troy's perspective is important as well.  Since I started working / designing PACS I have realized that we are in a rapidly changing world and that technology will replace the "old way" of doing things as well as many of the "old horses" in the field if we don't continue to grow, learn and embrace technology.  The number of persons going into Dx Medical physics may not be adequate to replace the "old gray mares" that are either retiring or getting ready to do so.  AI / computerization of our work effort can be considered as a "force multiplier" for Dx physics rather than a tool to eliminate/replace Dx Physicists.  From the vendor's perspective AI can completely replace the need for most if not all of the work Dx physics currently performs.  If we are doing "satisfy the regulator" physics, which is minimal performance measurements or are we active in the clinical practice of Dx radiology?   I feel our role needs to be one of active involvement in daily clinical practice rather than performing routine performance testing to meet regulatory requirements.  AI can and most likely will assume much of that work.  Our role should include developement of both testing algorithms and phantoms with targets to measure performance of our equipment in greater detail and faster than is the "accepted norm" -- an example would low contrast performance of fluoro machines.  Most of the LC measurements are subjective, poorly tested and meaningless when it comes to being able to quantitatively assess machine performance.  LC measurements in fluoro should be the basis for knowing when an II requires replacement -- few, if any, facilities use the LC degradation of these units as an indicator of the need to upgrade / replace an ii based system.  This being said -- II's are on the way out so the focus now needs to be with the digital receptor's performance.

To Troy's points, my concern is the needed skill set and interest to QC AI programs and results.  When I started developing PACS for DOD I encouraged MPs to become more active with PACS and its capabilities.  For the most part MP has ignored PACS as something that is not of interest and/or is just a "computer system and storage device".   If that mind set remains then AI will be a smart computer algorithm that does medical physics.

Mary Ellen is correct -- we are on the cusp of a HUGE change in the way we do our work.  We need to embrace AI, learn how to "CONTROL" it and use it as a "FORCE MUTIPLIER" for the work we currently do.  I anticipate that in 10 years the role of the Dx MP will be drastically different than our current role.  Those who fail to become actively integrated into clinical practice will be left behind.

Jerry Thomas

On Mon, Mar 24, 2025 at 9:02 AM Alisa Walz-Flannigan via Intl_dxmedphys_wd_osu_list <intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>> wrote:
We all need to continually evolve our toolkits, and our training programs. I have a ton to learn too, and I understand too well the time limits we all face with this. This last week was the NIIC-RAD course, that’s one option for tool-up (our

We all need to continually evolve our toolkits, and our training programs. I have a ton to learn too, and I understand too well the time limits we all face with this.

This last week was the NIIC-RAD course, that’s one option for tool-up (our institution participated). There are SIIM annual meeting courses (and WORKSHOPS!!), AAPM-SIIM joint sessions, and?
I’d be very happy to hear from others on what you’ve found beneficial (articles, links, projects, etc).

What have you found beneficial?


-Alisa



From: Mary Ellen Jafari <MaryEllen.Jafari at kp.org<mailto:MaryEllen.Jafari at kp.org>>
Sent: Sunday, March 23, 2025 5:45 PM
To: Alisa Walz-Flannigan <walzflanniga at wisc.edu<mailto:walzflanniga at wisc.edu>>; Nicole Ranger <nicoletranger at yahoo.com<mailto:nicoletranger at yahoo.com>>; troyxzhou at icloud.com<mailto:troyxzhou at icloud.com>
Cc: intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>
Subject: RE: [Intl_DxMedPhys] FW: Diagnostic Physics work and AI

Hi Alisa, I think people are misunderstanding my viewpoint.  I completely agree with your comments below.

I am not sure that everyone fully appreciates that the changes that are happening with radiologists and AI assistance in image interpretation are also going to happen (or are already happening) to diagnostic medical physicists. We need to be ready to evolve with the technology and be in the forefront of this change rather than being left out.  We also need to make sure that our role is not cut out by vendors.

Mary Ellen


Mary Ellen Jafari, MS, DABR (D,N), MRSE, MRSO, FACR, FAAPM, CIIP
Chief Physicist and Regional Radiation Safety Officer

Kaiser Permanente
Southern California Permanente Medical Group
Medical Imaging Technology and Informatics
199 S. Los Robles Ave., Pasadena CA 91101
626-430-1231 (office mobile)
Maryellen.Jafari at kp.org<mailto:Maryellen.Jafari at kp.org>

From: Intl_dxmedphys_wd_osu_list <intl_dxmedphys_wd_osu_list-bounces at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list-bounces at lists.osu.edu>> On Behalf Of Alisa Walz-Flannigan via Intl_dxmedphys_wd_osu_list
Sent: Sunday, March 23, 2025 3:36 PM
To: Nicole Ranger <nicoletranger at yahoo.com<mailto:nicoletranger at yahoo.com>>; troyxzhou at icloud.com<mailto:troyxzhou at icloud.com>
Cc: intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>
Subject: Re: [Intl_DxMedPhys] FW: Diagnostic Physics work and AI

Mary Ellen I think it would be excellent to have more and better continuous quality improvement tools, Tools to monitor for transducer dropout, failed detector elements, or damage to readout electronics, would improve the quality control that

Caution: This email came from outside Kaiser Permanente. Do not open attachments or click on links if you do not recognize the sender.

________________________________
Mary Ellen

I think it would be excellent to have more and better continuous quality improvement tools,

Tools to monitor for transducer dropout, failed detector elements, or damage to readout electronics, would improve the quality control that could be done (instead of waiting 6 months or a year for a physicist to show up who may or may not catch the problem).
Or tools that provide better feedback to technologists on positioning or manual modifications for acquisition or postprocessing, would be great too (anyone involved with EQUIP, or looking at centering in review of strange dose CT exams ?)

I don’t know if I see this as replacing ‘bread and butter” activities, but they could be tools to help us be more useful in our work, and “findings” would need human qc and evaluation. And I don’t see us throwing away our dosimeters, but how we evaluate AEC should evolve as AEC operates differently, for example.

Beyond the QC scope in healthcare, as Troy mentioned, there is great needs for knowledgeable and engaged people for oversight, governance, and validation.
Troy is this what you refer to by what the EU is doing: Artificial Intelligence in healthcare - European Commission<https://urldefense.com/v3/__https:/health.ec.europa.eu/ehealth-digital-health-and-care/artificial-intelligence-healthcare_en*aicareeu-deployment-of-ai-in-healthcare__;Iw!!KGKeukY!yByUTJxO00XuLlVXOBnXiWLxG-qstsZPLK-9Aym11MMFCmS1ltEubUGu-Qp84Az_Jt2jSj0-9Pk_jbJ3FeU5fuIaT3U_Ve-TOFln0Ufi8t_QiQ$> ?
Even if there are/were (2025 has been rough) resources at the FDA working on validation, the rate of creation of tools + variation in utilization+ variation in inputs wrt to training :  means that practices themselves are going to need to establish process and governance for their own validation; at least until we have better paradigm for data standardization, standards for utilization and tools for validation . And there are folk more knowledgeable than me working on this, see ARCH AI<https://urldefense.com/v3/__https:/www.acr.org/Data-Science-and-Informatics/AI-in-Your-Practice/ARCH-AI__;!!KGKeukY!yByUTJxO00XuLlVXOBnXiWLxG-qstsZPLK-9Aym11MMFCmS1ltEubUGu-Qp84Az_Jt2jSj0-9Pk_jbJ3FeU5fuIaT3U_Ve-TOFln0UcpoSJ6_w$>.

But regardless, I hope in all our work we are focused on making things better and not hung-up on how this impacts our turf (such prognostication is fraught and limiting).
I don’t think work for skilled experts is going to go away, but it should evolve, hopefully to be more effective.

Instead, we could focus on lobbying vendors to provide “for processing” or raw data as standard outputs for all their customers and not just to Microsoft Healthcare AI or whoever the highest bidder is, or keeping locked up for their own private development. This paradigm limits what can be developed and can limit robustness in clinical AI tools as well, or make training and validation that much harder .

And if we don’t want to see tools locked down in licensing subscriptions for 10 different companies, it would be fabulous to have vendor-neutral data , that could setup for proper competition and better product development, that better empower (whether that’s me or a lead tech, I’m not too worried. I see plenty of work around for me to do, and more than happy to share the load with whatever FTE we can find).

Nicole,

I don’t know who the arbiters or poster kids of Med Phys 3.0 are, but I’ve been involved evaluating or commissioning of CAD/AI since I started in medical physics.
And I continue to work in my role on such evaluation, as do others I know who are much more deeply and knowledgeably involved than me (I work with several).
And this topic is relevant to the imaging informatics education that I hope residents get (its on my radar as a program director).

-Alisa


Alisa Walz-Flannigan, PhD (DABR, FAAPM, MRSE)
Program Director, Imaging Physics Residency: Department of Medical Physics
Section Chief, Clinical Imaging Physics: UW Health-Radiology
University of Wisconsin, Madison, WI
walzflanniga at wisc.edu<mailto:walzflanniga at wisc.edu>
608-265-9685







From: Intl_dxmedphys_wd_osu_list <intl_dxmedphys_wd_osu_list-bounces+walzflanniga=wisc.edu at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list-bounces+walzflanniga=wisc.edu at lists.osu.edu>> On Behalf Of Nicole Ranger via Intl_dxmedphys_wd_osu_list
Sent: Sunday, March 23, 2025 3:14 PM
To: troyxzhou at icloud.com<mailto:troyxzhou at icloud.com>
Cc: intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>
Subject: Re: [Intl_DxMedPhys] FW: Diagnostic Physics work and AI

Troy, That is an interesting perspective. I’ve often wondered why proponents of Medical Physics 3. 0 did not propose more involvement by medical physicists in validating and monitoring Mammo CAD systems which were the precursors to more sophisticated
Troy,

That is an interesting perspective.  I’ve often wondered why proponents of Medical Physics 3.0 did not propose more involvement by medical physicists in validating and monitoring Mammo CAD systems which were the precursors to more sophisticated AI approaches in mammography imaging.  This would require a shift in training and necessitate developing consensus guidelines from the community.

Regarding QC and AI, I see a role for more automation in acquisition, analysis, and trend monitoring but the goal is to avoid demonstrable artifacts in clinical images so I would want to focus that effort on QC images which ensuring that clinical AI algorithms are clever enough to be able to detect macro effects introduced by equipment failure such that they would auto-route those images to MPs for investigation and remediation.

One thing is certain, the scope and practice of medical physics is constantly evolving.

Nicole


Sent from my iPhone

On Mar 23, 2025, at 2:55 PM, troyxzhou--- via Intl_dxmedphys_wd_osu_list <intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>> wrote:

Forwarded from the old listserv. Thank you for initiating this insightful discussion on the potential evolution in the role of diagnostic medical physicists with the advent of AI in medical imaging. I appreciate your perspective on the possibility
Forwarded from the old listserv.


Thank you for initiating this insightful discussion on the potential evolution in the role of diagnostic medical physicists with the advent of AI in medical imaging.

I appreciate your perspective on the possibility of AI replacing traditional testing and compliance tasks. However, I see this shift differently. With the recent regulatory advancements in AI, such as the EU's comprehensive regulations, there's a clear emphasis on ensuring the safety, transparency, and HUMAN oversight of AI systems, especially in high-risk medical applications like diagnostic imaging.

As AI becomes more integrated into medical devices, including imaging equipment, there arises a critical need for medical physicists to serve as vigilant watchdogs. Our role will evolve to include rigorous quality assurance of AI algorithms. Are AI-assisted collimation techniques accurately aligning with anatomical regions? Are AI-driven breast lesion detection systems effectively distinguishing true positives while minimizing the risk of false negatives?  These are just few essential questions that require human-driven oversight to maintain diagnostic accuracy and patient safety.

While AI offers promising advancements in medical imaging, it's crucial to proceed thoughtfully, ensuring that rigorous standards and human expertise are upheld to mitigate risks and optimize patient outcomes.

By the way it was written by ChatGPT and QA’d by me 😊

Happy Sunday

Troy




From: Diagnostic Imaging Physics List <DXIMGMEDPHYS at HERMES.GWU.EDU<mailto:DXIMGMEDPHYS at HERMES.GWU.EDU>> On Behalf Of Mary Ellen Jafari
Sent: Sunday, March 23, 2025 3:09 PM
To: DXIMGMEDPHYS at HERMES.GWU.EDU<mailto:DXIMGMEDPHYS at HERMES.GWU.EDU>
Subject: Diagnostic Physics work and AI

Hello Listmembers, the recent discussion over the new CMS eCQM on Excessive Radiation Dose or Inadequate Image Quality for Diagnostic CT made me think there could be a shift coming in the role of diagnostic medical physicists

The “bread and butter” of work done by many diagnostic medical physicists, and certainly consultants, is testing imaging equipment for compliance with regulatory requirement and accreditation standards.  Could thistesting be replaced by using a combination of AI applied to clinical images and remote access by the vendor to monitor imaging equipment?  Our vendors already can “remote in” to do diagnostics and monitor equipment performance.

Vendors would like this because they could charge for the service and more deeply embed their customers with their products.  Employers would like it because they could potentially employ fewer physicists.  I’m not saying it would be either a good thing or a bad thing, change is always happening and is something we need to consider thoughtfully.

Is this something other list members could see potentially happening?

Mary Ellen

Mary Ellen Jafari, MS, DABR (D,N), MRSE, MRSO, FACR, FAAPM, CIIP
Chief Physicist and Regional Radiation Safety Officer

Kaiser Permanente
Southern California Permanente Medical Group
Medical Imaging Technology and Informatics
199 S. Los Robles Ave., Pasadena CA 91101
626-430-1231 (office mobile)
Maryellen.Jafari at kp.org<mailto:Maryellen.Jafari at kp.org>

NOTICE TO RECIPIENT:  If you are not the intended recipient of this e-mail, you are prohibited from sharing, copying, or otherwise using or disclosing its contents.  If you have received this e-mail in error, please notify the sender immediately by reply e-mail and permanently delete this e-mail and any attachments without reading, forwarding or saving them. v.173.295  Thank you.

________________________________

Access the DXIMGMEDPHYS Home Page and Archives<https://urldefense.com/v3/__https:/hermes.gwu.edu/cgi-bin/wa?A0=DXIMGMEDPHYS__;!!KGKeukY!xLBe7GbTZgjh9efkgIaGY-OfqjrCQ2SHGtOfdXTVgya2P3nIN-S7GML12xxtgzTPVoNUE27c7_a-9Sd6Cy9rpkkCVhvj1xKObLuNpRrK$>

Unsubscribe from the DXIMGMEDPHYS List<mailto:DXIMGMEDPHYS-signoff-request at HERMES.GWU.EDU?subject=Unsubscribe>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250324/f3d6ea82/attachment.html>