Artificial intelligence (AI) is a hot topic everywhere, and mammography is no exception. While we all understand that AI can bring significant value to radiology practices, choosing the right one and implementing it effectively can be a real challenge.
With artificial intelligence being a fast evolving technology, there isn’t a clear playbook for its deployment. On top of that, with over 20 FDA-approved mammography artificial intelligence-based tools available on the market, finding the perfect fit for your radiology business can be particularly difficult.
We prepared a simple actionable 8-step guide to help with the selection process and ensure that your practice thrives with the perfect AI companion.
1. Identify the pain points of your radiology facility to be addressed by mammography artificial intelligence
2. Make a short selection of mammography artificial intelligence vendors
3. Compare different feature offerings of mammography AI software with your identified needs
4. Examine mammography AI’s clinical validation and its training dataset
5. Scrutinize regulatory clearance and data security of a mammography artificial intelligence solution
6. Assess the cost and return on investment (ROI) of a mammography AI in your specific situation
7. Assess through a pilot the user experience offered by each mammography artificial intelligence-based tool
8. Review AI’s integration process into your existing workflow
1. Identify the pain points of your radiology facility to be addressed by mammography artificial intelligence
To understand which mammography AI would serve you the best, identify the primary objectives of your radiology facility.
- Think about all specific pain points or improvement fields at your radiology practice to be covered by a mammography artificial intelligence tool (e.g. personnel shortage or burnout, accuracy issues)
- Perhaps most importantly, assess your existing resources to be used to implement a mammography artificial intelligence solution (including your budget, workforce, infrastructure, etc.)
2. Make a short selection of mammography artificial intelligence vendors
Now that you have defined the gaps to be covered by AI, it is crucial to select the right mammography artificial intelligence vendor. With numerous algorithms available, it’s vital to ensure that the vendor you choose is reliable and is a cultural fit for your business.
Nina E. Kottler, MD, MS, Associate Chief Medical Officer in Clinical AI at Radiology Partners recommends to consider evaluating the following factors when comparing potential AI vendors:
- Cultural alignment: Does the vendor share your values and vision for the future of AI?
- Funding: Does the mammography artificial intelligence vendor have sufficient resources to support innovation and maintain their operations?
- Regulatory expertise: Can the vendor efficiently navigate the FDA processes and get their products cleared swiftly?
- Support and training: Does the vendor provide comprehensive support for the technical implementation of their product? Can they help you educate your radiologists about their mammography AI tool?
3. Compare different feature offerings of mammography AI software with your identified needs
Remember that not all breast cancer screening AI solutions provide the same number of features. Opt for a mammography AI that matches the needs of your practice. For example, check out if the radiology artificial intelligence-based solution:
- Takes priors into consideration,
- Helps with worklist organization,
- Provides image quality assessment,
- Assesses breast density,
- Can be used both for mammography and tomosynthesis.
4. Examine mammography AI’s clinical validation and its training dataset
The best AI-based algorithm is the one that has clinically proven its accuracy and efficiency. Meticulously examine all clinical studies and research projects that evaluated mammography artificial intelligence performance. Here are some points to look at when appraising a study:
- Even though most organizations use the same number of readers, you should still consider how many readers are involved in mammography AI algorithm clinical study,
- Quantity of centers participating (single-center or multicenter clinical study),
- Type of clinical study employed for validation. Prospective studies, incorporating both case and non-case data, hold greater importance than retrospective studies.
Keep in mind that although FDA clinical validation studies use the same methodology, their results can not be directly comparable due to the difference in data sets. This is where the importance of the data set assessment comes out.
Evaluate the mammography AI training dataset in terms of:
- Quantity of images employed to develop the AI-based algorithm,
- The prevalence of breast cancer in the training and clinical validation dataset (should be similar to that of the natural distribution),
- If the images were collected from a wide spectrum of population demographics (coming from various geographic regions, and different racial and ethnic groups),
- If the algorithm was tested on a different set of images than the one used for its training and validation,
If the screenings were biopsy-proven cancer cases.
5. Scrutinize regulatory clearance and data security of a mammography artificial intelligence solution
Regulatory clearance of a breast cancer detection AI software testifies directly to the quality of the solution, and is mandatory for its use. Ensure that mammography artificial intelligence solution complies with regulatory requirements, such as FDA approval in the United States or CE marking in the European Union.
Another aspect to consider is ensuring patient data privacy and security throughout the AI workflow. For this, make sure that the mammogram reading AI adheres to healthcare data protection regulations for patient data, such as the Health Insurance Portability and Accountability Act (HIPAA) in the United States and the General Data Protection Regulation (GDPR) in the European Union.
6. Assess the cost and return on investment (ROI) of a mammography AI in your specific situation
To truly evaluate the return on an AI investment, ensure that it alleviates the pain points you’ve discovered at your organization. Let’s explore the advantages that the adoption of AI in mammography can bring to your radiology practice. The AI-based solution may address the following issues:
While there will be a cost of AI implementation and maintenance, you should be able to experience benefits that increase your financial standing resulting from AI adoption.
One more factor to consider is pricing models. Different breast imaging AI vendors propose different pricing options (i.e. CapEx, Opex, per click, per gantry). After careful evaluation, you will find the one that suits you best.
7. Assess through a pilot the user experience offered by each mammography artificial intelligence-based tool
When you integrate AI into your radiology department, your staff is likely to be the most concerned about the changes, and it is completely understandable. Whereas you recognize the long-term value AI for mammography will bring, not all breast imaging professionals may be ready to embrace new technology, potentially leading to significant modification of their work habits. This is why user experience is a cornerstone when choosing an AI in mammography.
That’s why it’s best to choose the least intrusive technology that can smoothly integrate with a physician’s mammogram reading routine. Here are some thoughts to consider:
Examine if an AI for mammography displays results quickly using a notification system, within the reading interface, or if it provides the analysis results on a separate screen.
Is it possible to filter the shown findings and hide the least important ones? One major complaint about old CAD systems is that too many marks are shown. With AI, that has been addressed, and specifically with MammoScreen®, it displays only marks that are considered important for review.
How much time does it take to display the results? Is the score provided by AI software easily understandable, etc.? Explore actual case studies of AI in work provided by physicians who have tested the software.
8. Review AI’s integration process into your existing workflow
Incorporation of an AI-based solution into existing mammography systems needs to be thoroughly prepared. For seamless integration, ensure that it is compatible with current systems like PACS (Picture Archiving and Communication System), RIS (Radiology Information System), and your viewer and reporting software.
Evaluate how the vendor of the artificial intelligence for mammography handles data transfers and storage, and if they ensure patient data privacy. AI should improve workflow and reduce manual tasks without disruptions. Lastly, think about providing adequate training and support for staff to maximize the benefits that the AI for mammography offers.
Conclusion
The right AI-based breast imaging solution may enhance your radiology practice. While finding the most suitable tool is challenging, involving all stakeholders (radiologists, technologists, administrators, and IT staff) can help streamline the process.
Together define your pain points, choose a reliable AI vendor, analyze mammography AI features, and assess if its accuracy and efficiency received sufficient clinical proof. Don’t forget to ensure its regulatory clearance and data protection. Pay attention to the training data set of the AI for radiologists. Evaluate the ROI, examine the user experience, and ensure seamless integration with your existing radiology systems.
Ready to make a step forward? Discover MammoScreen.
MammoScreen is the first and the only mammography AI software that takes priors into consideration, available both for 2D mammography and tomosynthesis. We provide seamless integration into your workflow at a flexible fee rate.
Let’s discuss how MammoScreen, our AI-powered solution for breast cancer screening can make a difference for your team and help improve patient outcomes. We’re here to chat and answer any of your questions.