Leaders in Breast Health: Dr. Nisha Sharma, Leeds Teaching Hospitals, NHS Trust, UKSep 28, 2021
Dr. Nisha Sharma is Director of the Breast Screening Program in Leeds/Wakefield, Leeds Teaching Hospitals, NHS Trust, UK. She is a member of several national breast screening committees driving innovation, and a professional clinical advisor driving quality improvement in breast screening programs. She is an early adopter of A.I. to amplify radiologists’ performance and improve health system efficiencies.
Dr. Rola Shaheen, Chief and Medical Director of Diagnostic Imaging, Peterborough Regional Health Center in Canada, speaks with Dr. Sharma on the implications of A.I. in breast cancer screening, particularly as it relates to health system efficiencies and radiologist performance at times of workforce shortages.
As you know, breast imaging continues to face resource shortages across the world, especially with COVID-19 variants emerging globally. Workforce shortages as well as pressures on operational and system efficiencies are taking a toll on the delivery of breast screening services and can impact the quality of patient care. Can you describe how workforce pressures have impacted screening services in Leeds and the UK generally?
In the UK, we run a three-yearly breast screening program; we’re the only country in the world that runs a three-yearly screening program. But we have significant workforce shortages and we’ve had that for a while.
This is continually getting worse – by 2025, a third of our consultant radiologists will be retiring. That’s in the next four years. And this is also mirrored in the radiographic workforce, so we have a lot of radiographers that will also be retiring.
But I think my concern is that during the COVID-19 pandemic, we stopped breast screening within the UK towards the end of March and we resumed at the end of June, so we had the cessation of about three to four months of breast screening which leads to deskilling of our mammographers. Then, when we did restart screening, it was quite stressful for them because we had to lengthen the appointment time due to the personal protective equipment. Our appointment times initially were six minutes but now became 15 minutes. So that significantly reduces your screening capacity. We also noticed that there was an impact on the quality of the mammograms because the radiographers were redeployed to do chest x-rays during that time, because that’s where the requirement was within the organization.
The workforce crisis is real and we need to look at innovative ways that we can support our current workforce and encourage innovative ways of creating skills so that assistant practitioners can continue to do mammography when we have a shortage of radiographers.
This is really a very sad situation that we’re all in the same boat. This redeployment nightmare has been at the back of the mind of every radiology department. You’re right; our technologists have been moving around doing other services that they’re probably not comfortable with. And as you said, innovation is probably the way around this.
I’m going to take us back to some workflow issues. Could you talk a little bit about breast density and how it affects clinical image quality and your diagnostic confidence when reporting a screening mammogram?
We recognize that breast density is important as an independent risk factor for breast cancer and we know that women with dense breasts are at increased risk. We also know that it can hide cancers, so it becomes challenging as a radiologist when interpreting the mammogram.
In the UK, we don’t record breast density as part of our national breast screening program because we don’t have a system to be able to do so and because visual assessment is very subjective. If I call a mammogram dense, I might have a colleague that says it’s not dense, and that’s not fair on your ladies, we need to know whether the breast is dense or not because it has a different implication.
So I think density is important. And we’re beginning to recognize that within the UK, but what we need is an automated way of assessing breast density that is simple and effective to help us as readers. I think also what that does is when you do get a dense breast, you know that you need to interrogate that mammogram for a bit longer because you need to make sure you’re not missing any subtle lesions. Therefore the quality then becomes really important because you need to have a good quality mammogram to diagnose a cancer; that’s the gold standard for screening, that’s our bread and butter. And if your image quality isn’t there, you’ll struggle to diagnose the cancers. This becomes particularly important in women with dense breasts.
In the US, there was a strong movement to record breast density. It became legislative law, which was great. We haven’t quite got there in the UK, though we are working towards that.
Even here in Ontario we have started to include the four categories of breast density because we realize the importance of the impact of this, on reporting and, as you said, having an objective measure is very, very important.
You touched a little bit on a system that may objectively get you that kind of information. What are the biggest gains in operational and system efficiencies that A.I. can help to achieve in the NHS especially relating to mammography quality issues?
I think image quality assessment is what’s really important. In the UK, we do currently have a quality assessment program, but it’s often done downstream: a lady has a mammogram but the technologist will review the mammogram further down the line after it’s already been reported by the radiologist, so the real-time feedback isn’t there for the technologist. Also, because they only review a percentage of the mammograms they perform, you don’t have an overview of the strengths and weaknesses of each individual technologist. It therefore becomes really hard to tailor their education to help them improve in their areas of weakness, congratulate them on their areas of strength, and create partnerships where two technologists can come together and you can help improve the quality of the image.
I think artificial intelligence is a system efficiency, because the A.I. can take away the image quality assessment from the radiographers, so rather than spending the time assessing their own mammograms, they’re looking at the information that the artificial intelligence has provided them, telling them what their strengths and weaknesses are and providing them with evidence. They will benefit from that in real time, and as a result we as radiologists will benefit because the quality of the imaging that we get will be far greater than what we’re used to since everyone’s now informed and educated and supported in their training.
Therefore, when we’ve got a workforce crisis, if we’ve got our radiographers spending less time looking at the image quality assessment, they can then spend more time doing the actual mammograms. As radiologists, if we start getting good quality images, we will spend less time seeing that the lady needs to come back for technical recalls. This then improves the patient experience as well. And that’s where I think the greatest gains can occur.
You alluded to this, but maybe you can give us a bit more description in terms of what part of your reporting A.I. can help you with and how important it is that the radiologist still has the final say, and why.
I look at A.I. as being low risk or high risk. I think if you’ve got an algorithm that’s going to help improve quality, be informative, and provide education and training, I think that’s low risk. I think A.I. that’s going to influence your diagnostic interpretation of a mammogram that could have a real clinical impact on a patient is something I would call high risk. That’s something that you need to be really careful of before you deploy.
I think artificial intelligence is there to assist us to do our job better, but not to replace us. We have years of experience. Gut instinct plays an important role; we gather information that we don’t realize we’re gathering that makes us decide that this mammogram is normal or abnormal. Artificial intelligence, I think, will play an important role, but it won’t be able to function without us.
So we need to have the final say, we need to be in control, and we need to take accountability because at the end of the day, we will miss things and our patients will complain and someone has to be accountable. At the end of the day, it has to be the clinical team that’s accountable. It cannot be the A.I. algorithm, because the A.I. algorithm is part of the team.
A.I. is important. And we do need to support it, but it’s there to assist us so that we can do our job well and do it more efficiently.
I totally agree. In fact, I think the word out there now is that A.I. is not going to replace radiologists, but it’s going to replace radiologists who are not using A.I. The idea is to get used to the fact that we need to utilize these tools that are available for us to make our job better and save more patients.
Now, a quick point about diversity. It looks like the Leeds program has a diverse population of screen-eligible women. How do you observe the impact of the pandemic on the participation rate? Do you feel like this diversity in terms of ethnicity, socioeconomic status, age, etc. has contributed to the participation rate and can A.I. help at all with that?
I think the pandemic has definitely had an impact because people are nervous. What you find is that older women feel particularly vulnerable. They’ve been told they’re vulnerable, and age plays an important role. Now the message has changed from “stay at home, don’t come to the hospital,” to “come to the hospital, you’ll be safe.” It’s taking time for women to accept that.
And ethnicity plays an important role as well.
I think the important thing about artificial intelligence is that it doesn’t discriminate. If you have an algorithm that’s going to be applied to a mammogram, it will happen for any woman who attends for screening. It’s very inclusive.
The key thing is getting the women to attend the screening program. That’s one of the things with artificial intelligence is we’ve got to educate the public. We’ve got to create a transformation in terms of the cultural way of thinking and understanding that A.I. is good. We have an important role to play: we have to embrace artificial intelligence and we have to tell our women that it’s a good thing that the organization is embracing A.I. But the good thing about A.I. is that once you’re there, it doesn’t discriminate.
Let’s Stay Connected, Subscribe for Updates
Join our email list to stay up to date on the latest advancements in breast health technology.