Rethink Imaging

EP 35 • January 8, 2026

Teaching Radiologists for an AI-First World

Featured Guest

Dr. Ali Tejani, MD

Neuroradiology Fellow • UCSF Department of Radiology

Listen Now43 min

Also on:

Let's Talk

Beyond dose

Ask us what your imaging data can actually show you.

Start a Conversation

Share This Episode

Subscribe to Rethink Imaging

Dr. Ali Tejani, a neuroradiology fellow at UCSF, has spent his career building the AI education he wished existed when he started. As a resident at UT Southwestern, Tejani co-founded the Imaging Informatics and Business Intelligence track, a curriculum that began as monthly sessions and expanded to twice a month on trainee demand, with an after-hours advanced group covering large language models, bias, and AI monitoring. He now chairs the SIIM Members-in-Training Committee and co-directs the National Imaging Informatics Course, the joint RSNA and SIIM program that pairs flipped classrooms with hands-on model-building workshops.

The conversation gets specific about what has to change. Tejani argues that any residency without AI content is underpreparing its graduates, who will walk into practices where AI already sits in the workflow. He makes the case for retiring the hot-seat case conference in favor of sessions where being wrong is safe, explains why two-hour lectures lose today’s trainees while short, interactive content lands, and lays out how to introduce AI early without de-skilling residents: clear framing, read the images yourself first, then check the model, all practiced in sandboxes that never touch patient care.

Host

Chris St. John

Host, Rethink Imaging • Imalogix

Featured Guest

Dr. Ali Tejani, MD

Neuroradiology Fellow • UCSF Department of Radiology

Watch the Episode

Key Takeaways

Tejani calls it a disservice when residencies do not teach AI: graduates now land directly in practices where AI is in the workflow, without the long adjustment runway earlier generations had. What every radiologist needs is not coding or model math but informed, critical use: knowing whether a tool is tuned for sensitivity, where it fails, and how automation bias can pull you into agreeing with a wrong result.
The I2BIT track at UT Southwestern started as monthly curriculum content, expanded to twice a month on trainee demand, and added an after-hours advanced group. The month ChatGPT broke through in 2022, that group ran a focused session on how large language models could change radiology, surfacing use cases that later showed up in journals and vendor products.
The National Imaging Informatics Course (RSNA and SIIM) shows what hybrid teaching looks like in practice: flipped classroom discussions, asynchronous video with three or four take-home questions instead of fifty, and workshops where experts walk trainees through building AI models in Google Collab notebooks and prompt engineering.
His self-described hot take: the traditional hot-seat case conference should go. Collaborative case sessions where mistakes are normalized, with attendings admitting their own errors, get more residents volunteering for cases and learning from being wrong instead of hiding from it.
On de-skilling, timing and framing both matter. Introduce AI early but tell trainees to read the images first and confirm with the model second, run orientations on every deployed tool, and give residents sandboxes where they can hit the pitfalls in a controlled setting that never touches patient care.

Full Transcript

Rethink Imaging Podcast Transcript

Guest: Ali Tejani

Host: Chris St. John

CHRIS ST. JOHN 00:00:45 Welcome back to Rethink Imaging. Today I am joined by Dr. Ali Tejani, a Neuroradiology fellow at UCSF whose work focuses on training and supporting the next generation of imaging professionals. Ali created the Imaging Informatics and Business Intelligence track during his residency at UT Southwestern, currently chairs SIIM’s Members and Training Committee, and co-directs the National Imaging Informatics Course. He’s published on AI literacy in radiology and the role of social media in medical education. His passion for teaching goes back to medical school, where he won an Excellence in Tutoring Award. Today, we’re going to talk about how radiologists learn, how continuing education may evolve, and what it means to prepare for the future. Ali, thank you so much for joining us today.

ALI TEJANI 00:01:26 Thank you for having me. It’s a pleasure to be here and, of course, always great to connect with others in our field to talk more about the current and future status of radiology education.

CHRIS ST. JOHN 00:01:39 So can we just start with you a little bit? What drew you into radiology, and when did education itself become such a core component of your work?

ALI TEJANI 00:01:50 I wasn’t always thinking about radiology. I was initially headed towards a life in neurology as a sports neurologist in headache medicine. Working with a lot of our teams, though, I started to find an inclination towards imaging and found myself heading down to the reading room in our hospital to learn more about our patients’ scans, what that means, and how we can improve on the orders that we had asked for and all the indications. Slowly, as you can imagine, that led to me gravitating towards radiology to the point where, after a couple of dedicated rotations, that was it. I knew that was home and that was where I needed to be.

Combine that with—this is about 2019 at the time—we hadn’t had the AI waves that we’ve had in the last five or six years. This was a time where there was quite a bit of advanced imaging in terms of modalities, but I was lucky to be in a place where we did have some local AI work implemented in our workflow at the time. As a medical student, even as a resident, that was exciting and thrilling to witness. That drew me in. That’s how I found myself eventually in radiology down this path at the intersection of imaging, advanced technology, and AI—really to where we are today after many years of rapid growth of AI solutions and how we think about and see AI. It seems to exponentially accelerate its pace, as we’ll see here in the coming months, too.

CHRIS ST. JOHN 00:03:30 From journeying through med student to where you are now, when did you become so excited about educating others?

ALI TEJANI 00:03:44 You don’t realize it in the moment when you’re teaching or learning and pursuing those interests, but in retrospect, thinking about it, I grew up in a household where I had educators as parents. Seeing them work in that field and then finding myself in a position where, initially as a young trainee and student, if I could learn a concept and teach it really well to someone else, I felt comfortable knowing that I had done my best to really master a topic.

Initially, it started off with that self-growth perspective and as a way to check myself and make sure I knew what I was talking about. But as I got older and found myself mentoring and teaching more, I found a really profound sense of fulfillment and satisfaction in seeing others overcome the barriers that I faced when trying to learn something. Since then, it’s become a drive to help make it easy for others to learn things without facing those same challenges or barriers, and to make it more efficient because I know that they will find that same satisfaction that I did and pass it on to the next generation in a cycle that continues.

When I was first encountering radiology, I wanted to learn more about AI. There wasn’t much out there aside from a few resources, especially in terms of formal curricula that could be used locally to educate those who wanted to learn more. There was a process of exploration and trial and error that took quite a bit of time and personal resources. At the end of the day, I started to realize there were some things that were really helpful and some things that weren’t so helpful, and we didn’t want others to have to go through that entire process to reinvent the wheel. So the AI education perspective became consolidating this information to present the highest-yield content to trainees. Following multiple iterations, we had curricula and tracks, and we’ve refined the AI education that’s available now. It’s not just an individual effort; it’s the resources and collaborations with multiple large organizations and numerous individuals for that goal.

CHRIS ST. JOHN 00:05:32 Your parents were educators, too, right? What was it like growing up with that approach?

ALI TEJANI 00:06:02 My mother was a Montessori teacher for many years, and I think that’s part of the inspiration. Growing up, I saw her selflessly dedicate her time to helping her students grow, even after hours beyond the workday. Seeing that drive and motivation was inspiring. Even though I ended up going into medicine and not necessarily directly into teaching, that same inspiration shaped the medical educator that I am today and hope to be in the coming years and decades.

CHRIS ST. JOHN 00:07:07 During your residency, you launched the Imaging Informatics and Business Intelligence track, right?

ALI TEJANI 00:07:16 A group of us did. I would hate to say that I did it alone because there was so much support from our program director, our vice chair of informatics, and all the others who contributed resources, content, and time. But yes, we set that up during residency when I was an intern coming into my radiology residency.

CHRIS ST. JOHN 00:07:31 Can you expand on that? What was the idea behind the track, what problem were you trying to solve, and who were some of those key players?

ALI TEJANI 00:07:42 Around that time, I was pursuing this self-inspired journey of learning about AI just to be someone who could practice radiology and still speak competently about artificial intelligence, which was starting to really pick up. When I went to my program director and the vice chair of informatics, I made this pitch and said, “Hey, I have found myself deeply immersed in this journey. There are a lot of insights here, and nothing would make me happier than to share those insights and see how it could benefit others and streamline the process of learning about this very important technology that will affect all of us.”

A lot of it is having the right mentors. This resonated with both of them, and they were very selfless with their time and resources in clearing any barriers to setting this up. They were able to guide us through formalizing this and making sure that it wasn’t just an experience that lasted one or two times, but something longitudinal, reproducible, and robust that would last sustainably over many years.

When we set that up, we rolled out a pilot year that started with content built directly into the radiology residency curriculum once a month, and then due to demand and popularity, twice a month. Thinking about how much you already have to learn in radiology, finding time to include that content can be tough, so it says a lot about training demand that we were asked to present more and had to find places in the curriculum for it.

We also presented advanced content to a select group of trainees who expressed interest through an advanced track that met after hours once or twice a month. There we had small, focused discussions. For example, when ChatGPT became really big news in 2022, that month we had a focused small group on how large language models affect radiology. Many of the use cases that we now see in journals, the vendor field, and papers were brought up by trainees on the front lines who saw the value and the future where large language models would fundamentally affect the way you practice radiology. We looked into generative AI and more advanced topics like bias, clinical deployment, and AI monitoring. That set a beautiful stage for our radiology trainees to graduate not just with core clinical knowledge to be excellent clinicians, but also to be leaders, stakeholders, and stewards for the proper, effective, ethical use of AI in practice.

CHRIS ST. JOHN 00:10:53 You’re talking about your mentors, and you’re talking about younger trainees. There’s a lot of cultural conversation right now about different generations and learning styles. Having worked with younger trainees, how do you think their learning preferences differ from past generations of radiologists?

ALI TEJANI 00:11:25 It’s a really interesting question. In my role as an AI advocate, I realize I’m in a bit of an odd position where I’m almost a bridge between generations. You have groups who have been practicing radiology for decades—longer than I’ve been alive—who really established our field. Then you have this other group of energetic, excited trainees at the front lines who want to change the field and adapt it to new trends.

I see myself in a position where it’s important to understand how we set up our practice to educate older generations of radiologists on new technology while addressing the desires of younger generations—not just educating them, but helping them appropriately integrate technology that will eventually become the status quo for them.

I often see more of an early-adopter mentality in the younger generation, where change management becomes a big conversation for older generations because introducing this technology changes how we practice, even if just in a single workflow piece. But there is no black and white in radiology; there’s a gray area. I have been inspired by many early adopters and innovators among older generations, and there are others in younger generations who are content to go with the flow.

If you’re going to be an advocate for AI and AI education, it’s important to understand these different perspectives so you can tailor information effectively. One educational strategy or integration strategy for one group of radiologists may not be as effective for another.

CHRIS ST. JOHN 00:14:05 Do you see younger generations gravitating toward other forms of content for their education, such as digital content, podcasts, social media, and videos, over traditional lectures and journals?

ALI TEJANI 00:14:41 The short answer is yes. There is a generational shift away from traditional didactics. It would be very hard to have a monolithic curriculum that only consists of attending a lecture, taking notes, and taking a test. Many individuals now opt for virtual options to watch lectures at double speed because it’s more efficient, opening up time to learn through more interactive, personalized ways tailored to how they learn.

The general trend is toward consumable digital content—small, targeted, interactive segments rather than two-hour lectures in one sitting. People want content that breaks up the monotony of sitting and listening to someone talk.

In the National Imaging Informatics Course—co-sponsored by RSNA and SIIM—we have flipped classroom sessions, workshops, interactive sessions, and traditional didactics. That hybrid method caters to a diverse learner population without excluding any group. If you update your virtual platforms and learning management systems to feature hybrid content, that’s the way to go. Podcasts are great, too. If your eyes are strained at the end of the day as a radiologist, popping in headphones to listen to an interesting, educational conversation is ideal. People have even created podcasts for board exam studying.

CHRIS ST. JOHN 00:17:09 Can you tell us some of the trends you’re seeing in the National Imaging Informatics Course with a bit more granularity?

ALI TEJANI 00:17:29 There is increasing demand for AI and imaging informatics content even earlier in the PGY years of residency, which ensures a good future for our field because learners want to absorb that information early and build on it.

As medical trainees and residents have busy schedules, the course doesn’t force you to be there in person. There are asynchronous ways to watch videos, catch up, and go through three or four questions that emphasize key take-home points rather than 50 questions at the end of a lecture.

The flipped classroom method has been really popular: watching a condensed video the night before and then having a rich discussion with the speaker the next day. Interactive workshops—where experts walk us through creating AI models, developing informatics projects, navigating Google Colab notebooks, or working through prompt engineering—have also been very well received. Embracing that hybrid approach is key.

CHRIS ST. JOHN 00:19:15 In residency, how did you structure case sessions to move away from the traditional high-stress “hot seat” environment and create a better learning culture?

ALI TEJANI 00:20:17 When I was on the chief residents’ program, we discussed how to update the curriculum to resonate with trainees. By far the most popular way to convey information was case sessions where trainees applied what they learned in lectures.

Traditionally, generations of radiologists talk about “hot seat” conferences where you feel intense pressure. Current generations don’t want that, and I don’t think that’s the way to go. That might be a hot take that disagrees with several generations of radiologists, but while there was value in case sessions, having a collaborative, safe space to make mistakes is essential.

It’s okay to be wrong—that’s why we’re in training. But that gets lost when you feel you have to impress everyone under high stress. If you already knew all the answers, you wouldn’t need to be in residency. In our residency, we shifted to sessions where it was okay to be wrong, which made it comfortable and fun to take cases. When attendings normalize making mistakes, we learn from them in training so we don’t make them at the workstation.

CHRIS ST. JOHN 00:22:36 I love challenging the status quo. Creating a safe learning space leads to better peer learning and ultimately yields better results for patients.

ALI TEJANI 00:23:37 It takes away the fear and intimidation associated with older formats and allows for relaxed concentration to identify knowledge gaps. Instead of feeling embarrassed, the trainee focuses on what to study to become a better radiologist. As educators progress, they often teach the way they were taught, especially with clinical and administrative pressures. It takes an intentional effort and the right people in the right positions to institute change.

CHRIS ST. JOHN 00:24:35 You’ve published and written on how social media can be used in radiology education. For someone like me who is skeptical of social media, what good have you seen come from it in medical education?

ALI TEJANI 00:25:00 There are well-documented negative sides to social media, but there are definitely positive ways to leverage it for education. Some of the most effective educational initiatives have been social media-driven projects presenting focused visual content for visual learners. On platforms like X, Instagram, and BlueSky, educators develop graphics that address concepts that are hard to grasp just by listening to a lecture.

Followers gravitate toward those resources because it’s a quick read while eating breakfast, commuting, or unwinding at the end of the day when you don’t have the energy to read a textbook chapter. Meeting users where they are on these platforms allows them to personalize learning to their needs. It takes a learning curve to use these tools effectively, but tapping into these platforms is an important landscape for educators.

It’s a continuous learning process. Using the right hashtags helps reach specific audiences, and when sharing content, tagging the authors and creators expands your scope and engages their audience. Beyond the algorithm, it’s a great way to meet people, collaborate, network, and broaden your horizons.

CHRIS ST. JOHN 00:29:50 When trying to decipher truth from fiction in online spaces, what advice do you have for trainees looking at higher-level educational content?

ALI TEJANI 00:30:04 The risk is taking an unvetted post at face value. Trainees should look for trusted experts and individuals who are verified, sponsored by recognized organizations, or known from presenting at major conferences. Start there and take unverified content with a grain of salt rather than treating it as a ground truth reference standard.

CHRIS ST. JOHN 00:31:26 Do you see paths forward for social media and web-based learning methods to start fulfilling CME requirements?

ALI TEJANI 00:31:40 CME is evolving toward more interactive, problem-based formats rather than traditional multiple-choice questions. Groups are already posting short clips of CME content on social media as previews to invite attendance. CME won’t stay static, and platforms that offer engaging, efficient learning will see the highest adoption.

CHRIS ST. JOHN 00:33:20 From a high-level perspective, what do radiologists right now really need to know about AI?

ALI TEJANI 00:30:30 If a radiology training program is not providing AI content to trainees, it is underpreparing them for the future. Regardless of the practice setting, AI will impact daily practice as a radiologist.

Radiologists don’t need to know how to code or the complex mathematics behind AI, but they must know how to be effective, informed, critical users—understanding when a tool is sensitive, what its pitfalls are, and how to avoid automation bias. It’s like ultrasound: you understand the physics, but you become proficient when you gain hands-on experience operating the probe and adjusting settings. Radiologists need to know how to be functional, ethical users so AI enhances rather than disrupts their workflow.

CHRIS ST. JOHN 00:37:20 What about the timing of introducing these AI tools, and how do we balance the risk of de-skilling?

ALI TEJANI 00:37:40 That is the exact discussion happening right now among educators. I want to acknowledge one of my mentors, Dr. Tessa Cook from Penn, who is a selfless educator and has been instrumental in these conversations.

If you introduce AI too early in medical training, you risk trainees becoming dependent on AI as part of their search pattern. If they go into a practice setting with fewer resources and less AI integration, they might feel like a fish out of water. On the flip side, if you wait too long to introduce AI, there won’t be enough time in the curriculum to ensure they become effective users.

If you introduce AI early, you must be very intentional. We hold clear, upfront conversations with trainees: when opening a case, don’t run straight to the AI results for intracranial hemorrhage. Go through the images yourself first, search for the findings, and then confirm with the AI output. By having orientation sessions and reviewing model performance together, we teach trainees to be exceptional radiologists who know how to use AI well without being dependent on it.

CHRIS ST. JOHN 00:40:56 For the residents and fellows listening, what advice would you give about staying curious, staying engaged, and getting more involved in education?

ALI TEJANI 00:41:11 Stay hungry and stay curious. Challenge yourself to learn new content, and don’t be afraid to reach out to people in the AI space—we want to help the next generation. Take advantage of resources from major organizations like RSNA, SIIM, and ACR that are free to trainees. If a resource isn’t offered at your institution, advocate for it.

If you are a medical student interviewing for residencies, ask programs how they educate the next generation with AI. Programs need to meet trainees where they are.

To tie this back to education: let’s not revert to traditional didactics when teaching AI. Put AI tools into the hands of trainees in controlled environments, like sandboxes or playgrounds, where they can walk through pitfalls without impacting patient care. When they sit at the workstation, they will know exactly what to do with an AI result.

My calls to action: if you’re a trainee, use these resources and reach out. If you’re an educator, push the mold and advocate for the next generation.

CHRIS ST. JOHN 00:44:36 Where can folks find you online if they’re interested in connecting?

ALI TEJANI 00:44:41 You can find me on LinkedIn, Instagram, BlueSky, or via email. I’m always happy to engage with trainees and colleagues.

CHRIS ST. JOHN 00:44:56 Ali, thank you so much for joining us today on Rethink Imaging. It’s been a total pleasure talking to you.

ALI TEJANI 00:45:03 Likewise. Thank you so much for the invitation and for bringing attention to such an important topic.

Questions from This Episode

What do practicing radiologists actually need to know about AI?
Not how to code or how the math works. Tejani compares it to modality physics: useful background, but the working skill is knowing how each tool behaves, where it fails, and how to guard against automation bias when the model is wrong. He frames it as being an informed, critical, ethical user, the same way hands-on ultrasound time matters more than lecture slides about the probe.
Does introducing AI early in training risk de-skilling residents?
It can, if AI becomes part of a trainee’s search pattern before their own reading skills are solid, then they change jobs to a practice with fewer tools and are, in his words, a fish out of water. His answer is early exposure with intentional framing: read the case yourself, find or rule out the finding, then check the AI result, backed by orientations on every deployed model and sandboxes where mistakes carry no patient risk.
How do younger radiologists prefer to learn?
The lecture-notes-test loop is fading. Trainees watch recorded lectures at double speed to free time for interactive, personalized formats: short digital segments, flipped classrooms, case discussions, podcasts, and focused social media graphics from trusted experts. Tejani’s advice is a hybrid curriculum that keeps some traditional didactics, because the learner population is diverse and no single format covers everyone.