By James Laskaris

Breast cancer is the second-most common cancer among women in the United States. Last year approximately 268,000 new cases of breast cancer were diagnosed in the U.S., and over 40,000 women died of the disease. When breast cancer is found early, the five-year survival rate is 96%. If not identified until more advance stages (e.g., stage 4), the five-year survival rate drops to 16%. The key to survival is early detection.

One area of increased focus is improving technology to increase the accuracy of diagnosing cancer in women with dense breasts. Manufacturers have turned to developing artificial intelligence (AI)-based software for consistency, improved image quality, and detection of lesions in dense tissue. With the added promise of improving workflow, AI is projected to grow to a $2 billion market by 2023.

Screening is the key to early detection. In 2002, there were 10 million mammography screening procedures performed in the U.S. Now the number has grown to approximately 37 million per year, performed in more than 10,000 certified screening centers in the U.S. This rapid increase is primarily due to a combination of increased payment levels for digital mammography, legislative initiatives, and guidelines that lower the recommended age of screening for certain patients.

The downside of mammography is that it can produce both false positives and negatives. Of the 37 million screening procedures performed each year, 10% of patients will require a follow-up diagnostic screening, ultrasound, or biopsy to rule out cancer. This can add to the cost of the procedure as well as place undue stress on the patient. On the other side, about 20% of mammograms produce false negatives. A missed lesion can lead to a delay in treatment, which can impact the patient prognosis. Patients most at risk are younger women with dense breast tissue.

Using technology to assist physicians in diagnosing breast cancer is not new. CAD systems first entered the market in 1998 (Hologic’s R2 ImageChecker) in a drive to help improve the diagnosis rate for mammography. The technology is based on using image recognition of preprogrammed patterns that indicate possible diseases such as cancer. This had its limitations, resulting in unnecessary follow-up procedures or missed cancers. Artificial intelligence (AI) first appeared in the 1950s in academic settings. In the late 1990s it entered the healthcare and data mining markets. AI has the ability learn new patterns of cancers on its own (unsupervised learning) from what it has experienced over time. Because of this, the effectiveness of AI will continue to improve with use over time.

The vendor market is currently in its infancy, with the first systems FDA-approved for use in the U.S. in 2018. Currently there are four vendors that have developed systems. iCAD’s and ScreenPoint Medical’s product lines are currently FDA-approved and are available in the U.S. European-based Therapixel (Mammoscreen) and Kheiron Medical Technologies (MIA) have CE markings (indicating conformity with European Economic Area standards) but are not available in the U.S.

Though this is still an emerging market, iCAD is currently the market leader in the field. iCAD offers two systems: the PowerLook, which is targeted to breast tomosynthesis, and the SecondLook for 2D mammography. These systems are offered as stand-alone workstations under iCAD’s letterhead or as an add-on to Hologic, GE, and Siemens mammography systems. ScreenPoint Medical’s Transpara system is FDA-approved, but we have not seen this system sold in the U.S.

There is a wide range of pricing for base mammography technology. Basic Digital systems range from $190,000 for 2D technology to over $400,000 for 3D technology. Tomosynthesis systems start at $500,000. Artificial intelligence is designed to be used as an add-on workstation platform. A complete system includes a workstation and software license. Pricing is similar to CAD technology, which can range from $60,000 to $100,000 depending on the number of software licenses. Service for the platforms start at $8,000 per year, and software support is approximately $3,000 per license.

Reimbursement is key for the adoption of any technology. CMS first approved reimbursement for CAD in 2002 as an add-on code. This increased payment an average of $15 per study. In 2018, computer-aided detection (CAD) was wrapped in with both digital screening and diagnostic codes under the 5-digit CPT codes. This is a clear indication that CAD has become a standard of care over the years. Unless vendors lobby CMS for a unique code, AI will be wrapped into screening and diagnostic codes similar to the CAD payment structure.

Current Payment Structure

77065 - Diagnostic mammography, including computer-aided detection (CAD) when performed; unilateral — $96 (CMS average payment level, Technical Component)

77066 - Diagnostic mammography, including computer-aided detection (CAD) when performed; bilateral — $122 (CMS average payment level, Technical Component)

77067 - Screening mammography, bilateral (2-view study of each breast), including computer aided detection (CAD) when performed — $101 (CMS average payment level, Technical Component)

77063 - Screening digital breast tomosynthesis, bilateral (List separately in addition to code for primary procedure) — $25 (CMS average payment level, Technical Component)

Studies indicate that AI provides equivalent or better results than CAD for unaided interpretation by a radiologist while not adding to the reading time. This should help with the patient throughput, especially for busy imaging centers. One study reflected that sensitivity increased with AI from 83% to 86%, while specificity increased from 77% to 79%.

Another study compared AI with the performance of 101 radiologists. Using the Area Under the Curve (AUC) as a guide, the AI system had an AUC higher than 61.4% of the radiologists.

One primary goal of AI is to reduce false positive per image (FPPI). A study published in the Journal of Digital Imaging in April 2019 showed that using the AI-based CAD, as compared to CAD, resulted in a 69% decrease in FPPI. This should translate to significant savings to the healthcare industry.

The two prime directives of value-based healthcare are improving outcomes and controlling costs. The total cost for follow-up diagnostic mammogram studies ranges from $150 to $300 (global fees under Medicare payment rules). Ultrasound ranges from $120 to $150 (global fees under Medicare payment rules). These represent a $4 billion per year market.

When cancer is suspected, the only definitive way of ruling it out is by performing a biopsy and examining the tissue under a microscope. The cost of minimally-invasive breast biopsy ranges from $1,500 to $2,500. This translates to an estimated breast biopsy market of over $3 billion per year. Considering 30% of all biopsies turn out negative, even a 2%–3% improvement in ruling out false positives offers the promise of better outcomes and millions in cost avoidance. Overall, it’s a win/win for the patient and the payer. And AI is still learning.

About the author: James Laskaris is the senior clinical strategist at MD Buyline.