Advances in PSMA-Targeted PET Imaging for Prostate Cancer - Episode 10

The Potential of Using Artificial Intelligence in PSMA-PET Results Interpretation

, , ,

Thought leaders discuss the possibility of using artificial intelligence to help interpret PSMA-PET results.

Transcript:
David Albala, MD:
I want to look into the crystal ball and look toward the future. One of the key things that’s starting to [be utilized] is artificial intelligence. How do you think artificial intelligence will be integrated in this type of technology? Do you foresee something in the future soon? Is it way down the road? How do you see it helping us, as clinicians and radiologists, using this artificial intelligence to help us with this imaging? Gary, I’ll start with you, and then I’ll whip around the panel.

Gary Ulaner, MD, PhD, FACNM: I will start by saying I am not an expert in artificial intelligence, but when I started in radiology, several people said to me, “Why bother? By the time you graduate, artificial intelligence will replace all radiologists.” Fortunately for me, that hasn’t happened yet. I don’t think artificial intelligence is going to be replacing radiologists. I’ve heard it said that radiologists who use artificial intelligence will replace radiologists who don’t use artificial intelligence. Currently, we use artificial intelligence systems that screen for things such as pulmonary emboli, rib fractures, other incidents in the brain, which you can overlook when you’re scrolling through 2000 images. I see artificial intelligence performing a greater role in radiology. I don’t foresee artificial intelligence replacing the radiologist anytime soon. Then, for PSMA PET [prostate-specific membrane antigen positron emission tomography] specifically, the role of artificial intelligence has been to pick up lesions that might miss our consciousness. Small pulmonary emboli are easy to miss when you’re scrolling through images. With PSMA PET, the sensitivity doesn’t seem to be the problem. You see the red dot or the black dot, or whatever color scheme you put your PET scan on; it’s then that the problem is the interpretation of the dot. I would say artificial intelligence has proven to be more valuable in other fields of radiology than it has yet been proven to be valuable in PSMA PET.

Ashley Ross, MD, PhD: Gary’s comments were excellent. People like to say “artificial intelligence” for everything. I would say “computer assistance”—none of us think computer assistance is not going to be a big part of PET reading because we had more storage space, more fast computing time, [and the] ability to analyze this large data. You can think that you don’t really need artificial intelligence per se, but you can have the computer go through, recognize the lesions, [and] put their SUVs [standardized uptake values]. If you start doing serial imaging of progression or per-treatment response in the future, knowing which lesions responded to what, [you could] look at the microbiology [to figure out] what are you going to biopsy, [and] when. Or to just make the scans happen faster. That’s the next step. It’s just like our cars. I still drive a manual transmission; I haven’t gotten on board at all with anything automatic. However, I love having the blind spot warning [icon light alert me]if someone’s in my blind spot. I don’t drive fast anymore, but it makes me drive safer. So, we have a lot of that computer assistance, just like Gary was saying. “Artificial intelligence,” I don’t know what that means. But I do like his statement. I think you’re going to see radiologists who [leave the profession] and radiologists who use computer assistance. [That’s] going to be the how the future goes.

Transcript edited for clarity.