Mammography

According to a study by the Moscow Center for Diagnostics and Telemedicine, the use of at least one AI-assisted reading service significantly reduces the interpretation time for screening mammography in an outpatient setting. Without AI, the average interpretation time is 432 seconds. With AI, this time is reduced by 15.0%, or 1.08 minutes.

Current regulatory guidelines mandate a double-reading protocol for screening mammography. This means that the images of each examined patient must be independently interpreted and reported by two radiologists.

The AI software demonstrates a sensitivity of over 81%, which exceeds the sensitivity of double-reading by radiologists as reported in review publications (72.0–73.0%). Furthermore, the specificity of the AI algorithm, measured at 90.0% in this study, is on par with the specificity range reported for double-reading by two radiologists (88.0–98.0%).

Therefore, in a double-reading mammography scenario:

• First read by a physician + second read by a physician: The total interpretation time would be 432s + 432s = 864s.
• First read by AI + second read by a physician: The total time is 0s (as AI processing occurs prior to physician review) + (432s * 0.85) = 367s. This scenario yields a net time saving.

Chest X-Ray

The study indicates that when interpreting normal chest X-rays using AI, a 33% reduction in physician time is achieved. In screening radiography, normal cases constitute approximately 95% of the total volume.

Chest CT
According to the Moscow Center for Diagnostics and Telemedicine study, the average interpretation time for chest CT scans was reduced by 30% with the use of AI, decreasing from 23 minutes to 16 minutes.