Artificial intelligence (AI) in mammography is a promising solution that optimizes radiologists’ efficiency and helps improve patient care. Nevertheless, there are still a few barriers to its adoption that prevent this technology from being widely used. Still, we believe there are solutions to the factors slowing down AI adoption in mammography.

Data security

Health providers operate huge data volumes, including patients’ personal data and health information. Unfortunately, data breaches are not uncommon and it has never been more important to ensure the highest level of data protection at medical facilities. AI vendors often use cloud-based servers to store the data. As opposed to the traditional on-site physical servers, this type of data storage may be perceived to be at a higher risk for cyber-attacks.

However, the reality isn’t so. More often, data breach reports are about on-site network vulnerabilities rather than cloud-based ones: high levels of data security can be ensured with a cloud-based server. When choosing an AI provider, it’s important to make sure that it conforms to the standards for information security. For instance, ISO 27001 is an international information security framework enabling organizations to efficiently and safely manage assets.

Reimbursement policies

Screening mammography has been covered by health insurance for a long time. In 2018, all separate codes or payments for Computer-Assisted Detection tools (CAD) in the US were removed, and a new current procedural terminology bundled code (CPT) was adopted. The CPT codes comprise performance of screening mammography and interpretation including utilization of CAD.

To be widespread, AI technologies may need a strong stimulus, such as the types of reimbursement CAD received for a period of time before being accepted into standard practice. This is where the New Technology Add-On Payment (NTAP) program can help. Created in 2001, NTAP is part of CMS’s Inpatient Prospective Payment System aimed at encouraging the implementation of innovative technologies. NTAPs are granted to new technologies that demonstrate substantial clinical value. An AI capable of detecting strokes on CT scans has recently received NTAP status. If replicated with other software, the NTAP program can help remove barriers to the broad use of AI algorithms.

Business model

It’s generally thought that implementation of innovative technologies is always costly and requires heavy investments. In some cases, this may be true, but there are exceptions. For example, if a healthcare provider opts for capital expenditure and acquires AI for mammography on a perceptual license basis, this might require heavy investments.

At the same time, if an AI tool is distributed with a software as a service (SaaS) model, no upfront investments are needed and the payments are based on the volume the software is used. This makes AI affordable even for facilities with a low budget.

AI is a black box

AI is a new technology and can be complicated to understand. Thus, many physicians have serious trust issues with AI. Think about it: how can one rely on the results provided by a machine (making life-changing decisions such as whether to recall a patient or not) when one does not understand how the machine came to the conclusion?

Good news is that not all AI technologies are black boxes. One of the examples of explainable AI solutions is MammoScreen®, a breast cancer screening decision aid. It analyzes screening mammograms and gives every case an overall score. To understand the score, physicians can track down the AI results to the corresponding breast(s) and lesion(s).

Keep up the pace with the ever-changing field of AI in mammography. Contact us for a free MammoScreen® demo to learn how this technology can work in your practice.