Latest / "Isn't That Interesting?!": Breast Cancer Survivor Stories + Remarkable Research / Episode 15: Dr. Debbie Anaby
Transcript
- 0:00The human eye can only see so much, but add artificial
- 0:04intelligence to the mix and the view changes.
- 0:07An Israeli researcher did just that using an MRI for an earlier
- 0:11detection for breast cancer, especially in high risk women.
- 0:15That caught the eye of earlier.org and is the subject
- 0:20of this episode of Isn't That Interesting?
- 0:23Here's Martha Kaley, founder and executive.
- 0:25Director. Ofearlier.org.
- 0:28I want to welcome everyone to this week's episode of Isn't
- 0:31That Interesting with Doctor Debbie Anaby, who is head of the
- 0:35Breast Cancer Imaging Lab Diagnostic Imaging department at
- 0:40Sheba Medical Center in Israel. She has expertise in breast
- 0:45imaging, specializing in medical image processing and AI based
- 0:50tools for early detection of breast cancer and personalized
- 0:54risk stratification, particularly with high risk
- 0:57populations. So in 2021, she received a grant
- 1:03from earlier.org titled towards earlier than early diagnosis of
- 1:08breast cancer in the BRCA carrier population.
- 1:13And I just found this, that her whole approach here was so
- 1:18great. And, and I have to tell you when
- 1:21it started out that you started with a different approach and I
- 1:26thought that that was so good that you recognized that you if
- 1:31you shifted it towards other technology with it, I thought,
- 1:35you know, that this is what these grants are about is to
- 1:38allow researchers to have some flexibility.
- 1:42So, so if if you wouldn't mind just addressing the grant and
- 1:46then I'll have you get into the the weeds about it, OK.
- 1:53Maybe I'll give you just a very quick background about Chipa
- 1:56Medical Center. First of all, so it's the
- 2:00largest hospital in Israel and we have a very special breast
- 2:05center in which we scan more than 5000 breast Mris per year,
- 2:10which is a big number. We also scan a large population
- 2:14of high risk women including the RCA carriers who are recommended
- 2:19usually to perform yearly breast Mris.
- 2:22So we have this whole rich database to work on, a very
- 2:27large MRI database and also a special one including these very
- 2:32high, high risk women that have these consecutive Mris that we
- 2:36can look at. So throughout my work, I
- 2:41realized that a very high percentage of BRCI carriers who
- 2:46were diagnosed with breast cancer had already an
- 2:50abnormality in a prior MRI. But this abnormality was not
- 2:56usually was not even suspected at that time radiologically.
- 3:00And that's because usually their appearance is very much like
- 3:06normal breast tissue. So it's very difficult to
- 3:10identify by the eye, but in a retrospective visualization that
- 3:17we we did here, we saw that in about 60% of those diagnosed
- 3:23women in a prior MRI, we already see some abnormality in that
- 3:28same area where the tumor later. Developed right?
- 3:31Right. And I was just amazed by the
- 3:35fact that there is a lesion that is within the resolution of an
- 3:41MRI, but it's not diagnosed at that time point.
- 3:46Right. Actually, those women already
- 3:49have a tumor but are sent home healthy.
- 3:54Yes. Right.
- 3:55And they and they are diagnosed usually a year after in the next
- 3:59MRI. Don't you just find that
- 4:02fascinating? I mean it's just but, but that's
- 4:05why you brought brought in AI with it, right?
- 4:10Right. So I understood that it's not
- 4:12related to the expertise or only to the expertise of the
- 4:17radiologist. It's just really impossible in
- 4:21most of the times to say something about these
- 4:25abnormalities by the eye. It's just so similar to normal
- 4:29breast tissue or benign, small, benign lesions that it's really
- 4:33impossible. But I thought that they are
- 4:38already there. They appear in our image.
- 4:40There must be something that we can do to say something about
- 4:45it. Not even if it's not OK, I'm
- 4:47sure this is cancerous, but maybe just help the radiologist
- 4:51to suspect that area and do something about it.
- 4:58So, and this is, and this is the grant that I received from you
- 5:03was actually the first support that I had for the development
- 5:08of the AI model. Oh my goodness.
- 5:11Well, yeah, Yeah, right. I saw.
- 5:13Thank you for. That.
- 5:16Our medical Advisory Board is superb and the researchers who I
- 5:23have been interviewing, it really has made a difference for
- 5:27everyone of you. I, I mean, it's just been so
- 5:32rewarding and validating for our work.
- 5:34So thank you for saying that and I'm so glad that we could do
- 5:37that for you. That was very generous and
- 5:41important and it allowed me to do this very preliminary
- 5:46research about this idea. And we gathered a limited group
- 5:53of BRC carriers with Mr. IS here at Chiba Medical Center with
- 5:59consecutive Mr. is actually. And we found those that were
- 6:02diagnosed and those that were healthy.
- 6:06We looked a year back. We annotated manually those
- 6:10abnormalities that we retrospectively visualized and
- 6:15we developed a model that can classify these lesions,
- 6:21cancerous or non cancerous. We are successful in about 65%
- 6:27of those cancerous lesions. That's great.
- 6:31Which is, yeah, it's, we're working on, on, on increasing
- 6:36this accuracy. But I think that, you know, to
- 6:39say that more than 60% of those women that were diagnosed could
- 6:43have been maybe diagnosed long before their formal diagnosis.
- 6:49Right. It would make it would make a a
- 6:52huge difference, right? I mean getting into treatment so
- 6:56which earlier than early, right? Right, It's earlier than what we
- 7:01know today that the early that we have today.
- 7:05And you know, many of these women are young.
- 7:07They're, they have, I don't know, sometimes they get
- 7:11pregnant, sometimes they do, you know, changes in life that we
- 7:17need. It's better to know if there is
- 7:20something suspicious. OK, let's wait 3-4 months.
- 7:24Let's check again. Right.
- 7:29Does it with the the women who perhaps I mean you mentioned
- 7:34they are young and maybe get pregnant.
- 7:37Is there any concern with doing the Mris with their if they're
- 7:42pregnant? Would that be a consideration?
- 7:45Yes, because we we don't we don't do the normal MRI scan
- 7:51during pregnancy. We cannot inject contrast
- 7:55enhancement. And that's actually the most
- 7:58important part of the of the breast MRI.
- 8:02That's the diagnostic imaging tool that we use.
- 8:06So for pregnant women we cannot use that.
- 8:10There are some other methods that are spoken about, but
- 8:15usually they don't. They don't do an MRI.
- 8:17They can do other exams, just sound mammography maybe, but but
- 8:21it's not, it's not the same. Sensitivity of MRIMRI is the
- 8:26most sensitive modality, right? Now, right, Yeah.
- 8:31Well, so as I was reading about your work, one of those things
- 8:35that I found interesting about thinking about I think that you
- 8:39said in there about the the particularly sensitive to BRCA
- 8:48abnormalities, How, how, why do you think it that this is
- 8:52particularly sensitive to that? I'm not sure.
- 8:56I'm not sure about that because we haven't checked other high
- 8:59risk populations yet. We are doing it now.
- 9:03We don't have results yet. I did begin with BRCA because I
- 9:08wanted to be kind of homogenic, Take a homogeneous population.
- 9:15They usually have a special type of tumor.
- 9:19I thought it would be more robust for me, but I did found
- 9:24that it was difficult to find enough cases to train our
- 9:28algorithms. So at a later point I decided to
- 9:32take a larger population including other high risk women
- 9:38as a first research just to develop a more robust model.
- 9:44Later. I think we can divide this
- 9:47population into sub populations and try to see if we can, you
- 9:52know, if there are any differences or, or if it's, you
- 9:58know, more sensitive for one type of population or, or, or
- 10:01some types of some types of tumors or I don't know.
- 10:05I mean we could, could check that at a, at a later time
- 10:09point, I think. But we we're now developing a
- 10:13more, you know, robust generalized model.
- 10:18Right. So what was the the demographic
- 10:21you said most mostly younger women.
- 10:24Was that what you You. So I don't remember the average
- 10:31age it was. It was around the 5050 year old
- 10:35women from from my from my what I know from you know from our
- 10:39database. So BRCA women are recommended to
- 10:44be scanned yearly with MRI from the age of 2530 until 7075,
- 10:54right? Many of them are very young, you
- 10:57know, before the age of 40, before menopause.
- 11:00Right, right. So.
- 11:03So now your plan with with this data that you have accumulated.
- 11:09So are you looking to do a broader study?
- 11:17How, what? What's your goal on this?
- 11:19Definitely. So a few things academically, I
- 11:24am working on data agreements with other institutions outside
- 11:28of Israel. That's great to get to, to use
- 11:32more data which is more heterogeneous, so from different
- 11:36populations, even if it's BRCA, but from different part of the
- 11:39world and from different MRI machines because that's, that's
- 11:44crucial for me to have like different quality images,
- 11:49different protocols and MRI, you know, they differ between
- 11:52machines. Even within SHEBA, we have data
- 11:56from 2 MRI machines and it's so different.
- 11:59The model cannot cope with the 2 two images, although they're the
- 12:03same the same type of images. I never thought.
- 12:07I never thought about that. Yeah, yeah.
- 12:09So it's very important for generalization.
- 12:14So that's one thing. The other thing is that we
- 12:17actually established a company to develop a system for early
- 12:23detection for AI of breast MRI in general and early detection
- 12:28of breast cancer just two months ago.
- 12:30It's very new, but I, I understood that that's the only
- 12:35way to push this forward to become, you know, a real
- 12:40decision support tool for a radiologist, which is, you know,
- 12:46that that's my dream. Well, you know, the consistency,
- 12:50I think just just framing that is, is a major, a major project
- 12:57to do because I that's, that's really critical for being able
- 13:02to have good results, don't you think?
- 13:07Definitely, definitely. I, I, I know from, from other
- 13:13studies that we've done and other studies that others
- 13:16perform that you can develop a model that works excellent on
- 13:22your data. And then once you bring other
- 13:25data, it's, it doesn't, doesn't work as well.
- 13:28And it's demands work. It demands, you know, to
- 13:34generalize it so that, yeah, we have to train it with a lot, a
- 13:40lot of data, heterogeneous data for that.
- 13:43So with your data now, are you applying for larger grants and
- 13:48and where is where would that be coming from for you?
- 13:52So we actually we were awarded 2 more grants since then, larger
- 13:58grants, one from the Israeli government, but and the other
- 14:02from ICRF, the Israeli Cancer Research Fund.
- 14:06And now we have so like two more sub projects going on here with
- 14:13this and with the company we were also seeking for funding
- 14:18now. So that's like much larger
- 14:21that's. Great.
- 14:22So, so this grant that we gave you allowed you to branch
- 14:28forward into that, is that correct?
- 14:30Definitely, definitely. Well.
- 14:33That's the very, you know, the basis of everything that I'm
- 14:38doing now in this field. I bet you went way right.
- 14:44Yeah. Well, that's great.
- 14:47I am. So I'm so excited for you and
- 14:50I'm so excited for us that we were able to give you a boost to
- 14:55move forward with this. I think this is very
- 14:58instrumental and, and as AI progresses, I, it's going to be
- 15:03an interesting to watch all of the process going along.
- 15:07So, so I would love it if you would keep us updated on, on
- 15:12your successes because I know you're going to have a lot of
- 15:16them. So would you like to say
- 15:18anything else or are you OK? What do you feel comfortable
- 15:22with? I can add that I I don't know if
- 15:26you know, but I was awarded another grant from you not long
- 15:30ago. I.
- 15:31Remember that? Yeah.
- 15:33So it's also. In the same area of early
- 15:37detection, it's another like characteristic that we're trying
- 15:41to extract from the MRI images. So it's also another like angle
- 15:49for this project. So that's very, you know, very
- 15:52generous and very helpful to. Be well, the interesting thing
- 15:58I've seen, I seem to use that word all the time.
- 16:01Interesting. But what I've been able to see
- 16:05in these interviews is how many different angles we are
- 16:10approaching breast cancer from all the different like with AI
- 16:15now, but with all of the various things, liquid biopsies and
- 16:20looking at proteins. And so I think it's really in a
- 16:25very important time that we've been able to, to jump in on
- 16:30this. So thank you for your your
- 16:34contribution to that too. And thank you for your support,
- 16:38OK. Well, thank you so much and have
- 16:41a good day and I thank you for the interview.
- 16:43Thank you, Martha. Great talking to.
- 16:46You bye, bye. This episode of Isn't That
- 16:49Interesting was made possible thanks to volunteers and donors
- 16:52just like you. It was hosted by Martha Kaley,
- 16:55founderofearlier.org and edited by Liam Mcburney.
- 16:59I'm Valerie Amsterdam, reminding you that for the last 30 years,
- 17:03this is North Carolina based organization has helped fund
- 17:06groundbreaking research all over the world, from molecular
- 17:10imaging to a sensitive blood test for breast cancer.
- 17:13Plain and simple, earlier detection saves lives.
- 17:16And you can help host a fundraising event, make a
- 17:19donation or volunteer your time. You can find us not only on the
- 17:23web, but on Facebook and Instagram too.
- 17:26Let's change the picture of breast cancer together.