Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Using A.I. to Detect Ovarian Cancer
220 Plays1 year ago
As women with PCOS or other hormone imbalances, we may be at a higher risk for ovarian and other reproductive cancers. Ovarian cancer is particularly deadly because it’s typically not symptomatic and diagnosed until its spread. In this episode I’m speaking with Dan Ure, He is a researcher who has helped develop an algorithm to detect ovarian cancer. He’ll tell us how it works, what to ask for at your doctor’s office and other ways that artificial intelligence is being used to support women's health.
For more information:
Aspira OvaWatch
Aspira Ova1Plus
IOTA App for Android
IOTA App for iPhone
The Life Detox is brought to you by Bubble & Bee Organic
You can download my free PCOS Thrive Guide with more than 40 pages of detailed information for how you can manage PCOS naturally.
Recommended
Transcript
Linking PCOS and Hormonal Imbalances to Ovarian Cancer Risk
00:00:05
Speaker
As women with PCOS or other hormone imbalances, we may be at a higher risk for ovarian and other reproductive cancers.
Challenges in Ovarian Cancer Detection
00:00:15
Speaker
Ovarian cancer is particularly deadly because it's typically not symptomatic and diagnosed until it's spread. Additionally, exploratory and diagnostic surgeries can be risky and end up inadvertently spreading the malignancy.
AI Innovations in Cancer Detection
00:00:30
Speaker
But new technology is about to change all of that.
00:00:34
Speaker
Today I'm speaking with Dan Urr. He is a researcher who has helped develop an algorithm to detect ovarian cancer. He'll tell us how it works, what to ask for at your doctor's office, and other ways that artificial intelligence is being used to potentially save lives.
Personal Anecdotes: Stephanie Greenwood and Dan Urr
00:00:53
Speaker
I'm Stephanie Greenwood, and this is The Life Detox.
00:01:08
Speaker
I've known Daniel since junior high. We had a lot of math and science classes together. We even went to homecoming one year. One of my favorite memories was going up to his house and playing cards with him and the guys, drinking 7-Up from oversized plastic cups, and just laughing endlessly.
00:01:26
Speaker
We've stayed in touch over the years, and he messaged me about a year ago to share with me his excitement that he'd been part of a study that was being published in a medical journal.
Accuracy and Impact of the New Algorithm
00:01:36
Speaker
I thought it would be a good episode for this podcast. He was part of a study that helped develop an algorithm to detect ovarian cancer, and it's 90% accurate.
00:01:48
Speaker
It's estimated that around 20,000 women in the U.S. are diagnosed with ovarian cancer every year. About 12,000 of those women will lose their lives. Ovarian cancer causes more deaths than any other gynecological cancer.
00:02:05
Speaker
If it's caught in its early stages, the 5-year survival rate is over 90%. But if it's stage 3 or higher, survival rates can be as low as 30%. Only 20% of all cases are found early, in stage 1 or 2. Symptoms of ovarian cancer can mimic other conditions and may just feel like bloating or constipation.
00:02:29
Speaker
Typically when symptoms are so severe that you visit a doctor for it, the cancer is in its later stages.
Screening Challenges and the Promise of AI
00:02:37
Speaker
There is no routine screening for ovarian cancer. Pap smears don't detect it, abdominal exams may, and if something is abnormal, you'll be sent for an ultrasound.
00:02:48
Speaker
First line of defense is typically an ultrasound done by your gynecologist. You know, those can be kind of messy or they don't have a clear picture or, you know, inconclusive. So, you know, usually when you come in with a ultrasound result, it can be labeled as benign or like very indeterminate. And I think in the
00:03:04
Speaker
the nine labels and or the indeterminate labels is where a lot of these like early stage ovarian cancers you know stage 124 are getting missed and once you hit the late stages like you said it's already usually spread or metastatic to perhaps other organs and then it's also it has a really high mortality rate. He says that there have been two blood markers that doctors have relied on for detecting cancer.
00:03:26
Speaker
two that have kind of been the workhorse of ovarian malignancy detection and they're called carbohydrate antigen 125 or cancer antigen 125 is for the sake of our conversation let's call it ca 125 so ca 125 expression is i think it's an inflammatory response a doctor can take a blood sample measure your ca 125 and have an idea and they're usually going to do that after the determination of some kind of system side of your
00:03:52
Speaker
reproductive system. They have cutoffs like for instance if you're pre-menopausal and anything above a 35 units per decoliter or something like that is at risk of malignancy and anything below that's considered normal but then you know if you're post-menopausal that value is higher. People started using other proteins in conjunction with CA125 so for instance one that's really indicative of potential cancers is called human epididymis protein 4
00:04:16
Speaker
or HE4 for short. And these two have been widely used in conjunction to determine whether or not it's a malignant cyst. But the blood markers have their drawbacks.
00:04:27
Speaker
So it's pretty specific to a lot of cancers, but if you just have a cyst or something, it can also be elevated in that particular patient. Endometriosis and polycystic librarian syndrome, all of these can increase your CA125 value and not really be indicative of malignancy because you just have some inflammatory response to a cyst. So CA125 and both HE4 will be elevated in patients that have malignancy, but if you just have a benign cyst, HE4 isn't as
00:04:55
Speaker
highly increased. The things that we run into in particular to say 125 is either false positives or false negatives and so if you get a false positive your doctor's sending you to a highly risk surgery for an evaluation on whether or not your cyst is benign or malignant and you know there's a chance that it can
00:05:12
Speaker
plants and spread as you know they're doing the surgery and they're also you know kind of biological impacts. The doctor thinks that it's at high risk of malignancy. You may have some of your you know of hysterectomy and oophorectomy to remove some of those components and depending on your age that may be something that's really traumatic. A lot of people have been investigating additional biomarkers beyond HEC 125 and HE4.
Reducing False Positives with AI
00:05:36
Speaker
The algorithm he developed looks at seven different markers.
00:05:40
Speaker
What our work is, is an algorithm based off of biomarkers. And so biomarkers, in this case, can be proteins or inflammatory response measures or hormones. The company that I work for has done some literature search and a lot of experimentation on thousands of patients that they've come from surgically confirmed cases of ovarian malignancy.
00:06:02
Speaker
And so we've come up with a list of seven different biomarkers and they all have three different level expression and different combinations. So, you know, you do seven factorial and all of a sudden it's a really complicated look and trying to figure out whether or not this is a malignant or benign mass and that really is going to affect the surgical treatment of the patient. So that's where I come in.
00:06:22
Speaker
is we've got a bunch of different proteins and we've got different levels of expression. We've got age, we've got whether or not this patient is pre-menopausal or post-menopausal. That's where we can develop an algorithm to look at that and remove some of the human error and thought process associated with it. The algorithm that I've most recently finished and the company has started selling is seven different proteins along with age and menopausal status that go into this deep neural network.
00:06:48
Speaker
So when you say a neural network, is that, that means like an AI technology. That's right. That's an AI technology where you provided inputs and you kind of tell it what you want it to learn and how you want it to learn, you know, like a zero or a one and one for a malignant and a zero for a benign. And then it generates a score somewhere in between looking at those protein values. The network is trained on some 3,500 patients and those are surgically confirmed malignant patients and it
00:07:15
Speaker
goes back and forth and finds out where it got it wrong and then corrects it, goes back and forth thousands and thousands of times until it comes up with a network or an algorithm that can detect.
00:07:26
Speaker
from all of these proteins, whether or not the mass that you have is considered likely to be benign or likely to be malignant. That can aid your gynecologist or your gynecological oncologist in some of your treatment options. They'll take that in conjunction with your ultrasound and then run it through this algorithm. It's just a blood test. This is pre-surgery. Then
00:07:47
Speaker
gives the algorithm returns a probability or just a score from one to ten on how likely it is that you've got a Boolean. It's kind of interesting because you know a lot of algorithms are multiply the value for one times the biomarker value by ten and the other by five and then it gives you a number and but this is kind of a black box approach a lot of times you don't understand what the network is doing it's just going back and forth and learning and then that has some multiplication factors that you don't necessarily understand
00:08:17
Speaker
But leads to really good prediction. Amazing. And it's about 90% somewhere in there accurate.
Future of AI in Routine Cancer Testing
00:08:24
Speaker
Yeah, we'll catch about 90% of malignancies, and then the biggest improvement that's coming from the algorithm that I designed is in its specificity, which means it's decreasing the number of false positives that you're going to get. So previous tests that are on the market have a 90% true positive rate, or it's going to get the malignancy correct, but you're also coming back with maybe 30% or 40% false positive rates.
00:08:51
Speaker
And so, you know, you're sending a lot of people to surgery unnecessarily based off of this algorithm. It's always better to go and have it explored if you think you're at high risk, if your doctor thinks you're at high risk. But there is, again, as we kind of discussed earlier, unnecessary or unwarranted or unwanted complications from exploratory surgery like that.
00:09:08
Speaker
Do you think that this will become something that is like an annual detection, kind of like a mammogram or cervical check where you have your ovarian cancer blood test? Yeah, that's the hope actually is that you can, if you've got either an indeterminate or likely benign mask from your ultrasound and your doctors, your OBGYN is looking at your results.
00:09:29
Speaker
and said, you know, I don't know if the surgery is necessarily the right option. The mass is small or kind of indeterminate. Let's do this blood test and then it'll give you a score. And if it comes back as a low risk, then what they might want to do is do a watchful rate watchful and weight strategy, which is, you know, they'll come back every six months or annually, depending on what your OBGYN thinks is most appropriate.
00:09:50
Speaker
see how it changes. So if the protein values change and you're, you know, closing in on a higher risk or if it remains low risk and results itself as sometimes, you know, cysts do, then, you know, you may not even need to go to surgery. So it's one of the things that we're really excited about is kind of stop sending people to surgery that don't need to go and with its associated complications.
00:10:10
Speaker
So if you are asymptomatic or if you haven't had an ultrasound done, because usually don't go into ultrasound unless there's some kind of problem, can you get this test? Can you have it ordered? Yeah, you can have the test ordered. If you are asymptomatic and you have a test, it can come back either positive or negative and be a good indicator on to whether or not you're at risk.
00:10:34
Speaker
And like you said, some of the early stage malignancies are pretty hard to catch in either ultrasound or proteins. But, you know, if you've got an algorithm that at least you have some confidence in a pretty good confidence, 90% confidence really, that you're catching something that's a malignant.
00:10:50
Speaker
We're going to take a short break here, but when we come back, Dan will tell us what to ask for in the doctor's office if you'd like to take the test, which tests are covered by insurance and which are not, an iPhone app that could save your life, and the exciting new algorithm he's developing now. Stay tuned.
00:11:13
Speaker
If you're enjoying this podcast, please consider checking out bubbleandb.com. We have more than 150 different products to choose from like soaps, insect repellent, facial care, and the world's largest selection of USDA certified organic deodorants. These are all products I've formulated myself to avoid xenoestrogens and not interfere with our hormone health.
00:11:33
Speaker
and we make everything in our own facility so we know exactly what's going in our products. So visit bubbleandbe.com and support this podcast and my little family. That's bubbleandbe.com. Okay, back to the show.
Differentiating Endometriosis from Ovarian Malignancies
00:11:54
Speaker
I'm interested to hear what other projects you're working on right now. So some of the spinoffs that we got from our patient collection, the biggest population group inside the benign cases were endometriosis patients. So the symptomology of endometriosis highly mimicking the symptomology of an ovarian malignancy.
00:12:14
Speaker
And so trying to do a very similar approach to what we did with the malignant cases, trying to develop an algorithm that's explicit to endometriosis, but also rules out ovarian malignancy. And so coming back to the two biomarkers that we talked about previously in ovarian malignancy in endometriosis and including PCOS, the CA125 value is all increased, but in
00:12:39
Speaker
Only in malignancy is that HE4 increased. You can detect endometriosis and again it would be a similar screening test where you go in and have that checked. Your doctor will ultimately need if you come back at high risk to do some surgical confirmation or a laparoscopy or something like that.
00:12:58
Speaker
In development is an endometriosis test. Typically, your doctor's already ruled out malignancy, but the biggest problem with endometriosis is the time to detection. On average, it takes somewhere about eight years for patients to get an appropriate diagnosis of endometriosis. The doctor's trying to rule out so many other things and there's so many other symptomology that's really similar. Bowel problems as well as PCOS is very similar to endometriosis and symptomology. That's one of the things that we're working on.
00:13:27
Speaker
Wow. Well, is there anything else that you want us to know about this technology and the projects that you're working on?
Enhancing Detection with AI and Ultrasound
00:13:34
Speaker
Yeah. So one of my favorite things that we're kind of starting to do right now is a conjunction of looking at the ultrasound images that your physician takes prior to getting the blood test and using visual recognizing tools from
00:13:49
Speaker
artificial intelligence to also spot and improve the ultrasound reads as well. So it's like taking a bunch of images in from, uh, ultrasounds and knowing which ones are labeled appropriately. And so there's all kinds of complications. So you really need a bunch of images that you can collect to, to develop some kind of, uh, visual algorithm like that. And then combine all of that information with.
00:14:15
Speaker
you know, your age and your menopausal status, maybe even your BMI as well, and, you know, kind of further improve the algorithm.
00:14:22
Speaker
based off of additional features. And so like kind of trying to paint a complete picture rather than segmented. So I got a couple of links for you that I'll send over, you know, close the conversation. And, you know, one of them is an organization that is pretty prevalent outside of the United States, but we don't use it very frequently. And I think that, you know, there's a lot of unexplored territory there. It's called the International Ovarian Tumor Analysis.
00:14:46
Speaker
or iota they have a set of rules that they established in 2008 or so that have been pretty well standardized so they have like features that part that they find in ultrasound you know like the size of the diameter of the cyst or you know whether there's some acoustic shadows and they also look at the blood flow going to these particular tumors and so like if you can take that information and improve on a ultrasonographer what your gynecologist can read or your obituary can read from
00:15:15
Speaker
a pretty standard ultrasound and just train an algorithm on thousands and thousands of different ultrasound samples, you're going to be able to come up with a more complete picture. It's some pretty exciting stuff. Moving forward, we're hoping that you can catch malignancy earlier and then reduce a lot of unnecessary surgery and then also go to a watchful strategy. Then also at the same time, if you do have a really high risk of malignancy,
00:15:42
Speaker
get you the appropriate surgical referral. So instead of having your OBGYN perform the surgery, it might be better to have gynecological oncologist perform it.
00:15:52
Speaker
Amazing. That's incredible that you can use AI to analyze an ultrasound and, and detect things. And I'm sure it's just going to get more accurate and more available as time goes on. And as women that are my age enter, you know, postmenopausal eras and the next, you know, 10, 20 years and our cancer risks go up that who knows what will happen in the next 20 years with this technology.
The OvalWatch Algorithm and Partnerships
00:16:19
Speaker
So one of the things from the iota group that I can send you is there's like an Android app or an iPhone app that you can download and you insert the features of your ultrasound and it'll give you some kind of like score as to how likely you are to have a percentage on malignancy risk.
00:16:38
Speaker
Wow. Wow. That's amazing. Okay. I'll make sure to get that link in there. Do you want to link to the company that you work for? Sure. I'll send you the company I work for. It's called Aspira Women's Health and the algorithm that I designed is called OvalWatch and it's primarily designed for that weightful watch strategy to make sure that we're not sending people to surgery unnecessarily. They also have a couple of
00:17:02
Speaker
project products that are ten years old or so that are also algorithm based that are on par with one that i designed one of the things that you're gonna run into as well as that unfortunately a lot of these tests as their first released are not gonna be covered by insurance because they don't have clinical utility study the test i have is working on getting its insurance clear but
00:17:22
Speaker
there are some other tests that already exist that are a little less specific in terms of reducing false positives, but very good to make sure that the mass you have is getting appropriately treated as malignant. Just kind of like arming yourself with the information when you go to your OBGYN and they're suspecting you've got some kind of cyst and you're concerned about malignant C, then you can go armed with the appropriate questions. Can I get a CA125? Can I get a CA125 and an HE4? Can I get
00:17:49
Speaker
get an ova watch test to see that my masses are still likely benign. In terms of PCOS, even though you've got multiple cysts, like 10% of cysts are typically malignant. But if they're popping up all the time inside of you, then yeah, you're going to have a really high risk of
00:18:06
Speaker
ovarian malignancy. So a lot of peace of mind as well as a good chance of catching it early. That's kind of the aim here. If you catch it early enough, you can reduce a lot of the mortality that's associated with malignant ovarian cancer. The name of the algorithm-based blood test that Dan helped develop is called Overwatch by Aspire Women's Health. I'll link to all of these in the show notes. Since the time of this recording, some larger insurance providers have started covering it.
00:18:35
Speaker
The older test that he mentioned that is typically covered by insurance is called OVA One Plus. Both tests, if not covered by insurance, are typically around $195 and can only be ordered through a doctor's office. I've also linked to the IOTA Android and iPhone apps where you can enter parameters from an abdominal ultrasound and get an ovarian cancer risk score.
00:19:01
Speaker
It was so fun and interesting to catch up with my old friend Dan about this life-saving technology. It gives me so much hope for the future as these tools will only become more accurate and more available.
PTSD and Aggressive Ovarian Cancer Risk
00:19:15
Speaker
But where does the theme of my show come in here? What role does trauma play in all of this? A 2019 study from the Moffitt Cancer Center and Harvard University found that women who experienced six or more symptoms associated with PTSD had a significantly higher risk of developing the high-grade serous histotype of ovarian cancer, which is the most common and aggressive form of the disease.
00:19:43
Speaker
So if you're being abused, please use this information as loving motivation to safely get out. Take care of yourself, your health and your mind. You can thrive.
Podcast Disclaimer and Future Episode Preview
00:20:06
Speaker
The Life Detox is produced by me, Stephanie Greenwood, and brought to you by Bubble and Be Organic. The views and opinions expressed are the speakers' own and do not necessarily represent those of myself or my company. Material and information presented here is for general information purposes only and is not medical advice. Being a guest on this show does not imply endorsement of Greenplay LLC or any of its projects. Stay well, friends.
00:20:36
Speaker
Next week on The Life Detox. It's a big one. I'm talking to award-winning author Donna Jackson Nakazawa. We have so much we'll be talking about with trauma and how it affects our health. And we'll discuss her new book, Girls on the Brink. How young women today are being affected by social media and what we can do to help. I hope you'll join us.