Every year, thousands of women undergo a mammogram even though, in most cases, there are no warning signs. The aim of this exam is to find small changes before they cause symptoms, and that is where reading the image makes the difference. The Caja Costarricense de Seguro Social (CCSS) has now announced that it will test artificial intelligence to support this work, in a pilot project that, according to the sources consulted, will begin in the Tibás-Uruca-Merced health area.

This article does not repeat the announcement. It explains how this kind of tool works, what role each person plays in the process, and which questions remain open before drawing conclusions.

What will be tested and where

According to El Observador CR, the CCSS will use artificial intelligence software to analyze mammograms and identify possible breast cancer lesions during a pilot project that covers 2,400 women in the first three months. The tool is Lunit INSIGHT MMG, according to Delfino.cr, which also reports that the agreement is being signed with Roche as part of the ImpulSALUD strategy.

The source of these figures is the institution itself. Delfino.cr details that the project seeks to raise the monthly care capacity from 540 to 800 mammograms. That figure matches what El Observador CR attributes to the authorities. The first stage is aimed at women in the area who meet the criteria for having the exam. El Observador CR indicates that the aim is to invite women aged 40 to 75 who have not had a mammogram in the last two years. To locate them, the CCSS will use the Expediente Digital Único en Salud (EDUS) and the FOCO multichannel system.

La Nación (CR) reported, in its summary, that the pilot plan would include 15 health areas. The other two sources only describe the initial stage in Tibás-Uruca-Merced, so the expansion to other areas cannot be confirmed with the data available.

How assisted reading works

To understand the project, it helps to separate two moments. The first is the capture of the image. The second is interpretation, which in screening is performed by a radiologist. What this tool changes lies in the middle: the program analyzes the image before it reaches the specialist.

According to Delfino.cr, the system identifies alterations in breast tissue that merit a more detailed assessment and prioritizes studies with suspicious findings. In practical terms, the radiologist’s worklist can be ordered according to the level of attention each case requires. According to El Observador CR, the software returns the image with marked areas, which the physician described as “heat zones”. This marking works as a signal to look more carefully, not as an answer.

This type of support follows a known logic: an automated second look can help avoid overlooking small details, especially when there are many studies to review. But it can also generate marks that do not correspond to real lesions. For this reason, the value of these tools depends on how they are integrated into the workflow and on the subsequent evaluation of their results.

What AI does not do

The three sources agree on a central point: the diagnostic decision remains in human hands. El Observador CR reports the clarification from the director of the Cendeisss, Jorge Jiménez, who stated that artificial intelligence will not replace medical interpretation or determine on its own whether a patient has breast cancer. Delfino.cr attributes a similar idea to the CCSS medical manager, Marvin Palma Lostalo: it is a complementary resource that does not replace clinical judgment.

This matters because the final outcome is still a medical report. If an image is marked as suspicious, the radiologist decides whether an additional study is needed. According to the sources, the tool is not the decision point.

Why early detection is the core of the project

Early detection means finding a change in the breast before it causes symptoms or spreads. In general terms, when a finding is studied in time, there are more options for evaluation and follow-up. That is why screening programs are organized around invitations, age-based checks, and follow-up of unclear cases.

Delfino.cr stresses that the project is not limited to reading images. It also includes coordinating complementary studies, biopsies, oncology consultations and, if appropriate, the start of treatment. According to the source, the goal is to reduce waiting times between each stage. This is probably the most practical aspect of the pilot: not only reading faster, but shortening the path the patient takes from the moment something is detected until she receives an answer.

Delfino.cr also mentions that prioritization by clinical risk falls under Executive Decree No. 45889-S, on care waiting lists with reasonable waiting times. The source does not detail how this framework will be applied in practice.

What is still unclear

There are several points readers should keep in mind. First, there is no exact start date, according to El Observador CR. Second, the sources do not fully agree on scope: one speaks of 15 health areas and the others of an initial zone. Third, there is confusion between mammograms and women: the Delfino.cr summary mentions 2,400 mammograms, while the body of the article and El Observador CR refer to 2,400 women. Fourth, the age range varies depending on the source.

In addition, the pilot does not yet have results. According to Delfino.cr, the CCSS will evaluate the project using indicators such as the number of mammograms performed, the turnaround times for delivering results, and the time between a patient’s referral and her care. Until that data is published, it is not possible to know whether the tool improves detection, reduces waiting times, or changes the number of false positives.

When to see a health professional


This content is for informational purposes only and does not replace the advice of a health professional.

News sources

Keep reading on Santé Press