AI Diagnoses Pig Pregnancy in 10 Seconds... Nongjincheong Develops World's First Technology

Pregnancy Confirmation with 95% Accuracy by Recording Just 10 Seconds of Ultrasound Video
Nongjincheong: "Improves Farm Work Efficiency and Reduces Production Costs Such as Feed Expenses"

The path has been opened for non-experts to easily perform 'pregnancy diagnosis of sows,' a task previously carried out by skilled professionals, through artificial intelligence (AI).

On the 7th, the Rural Development Administration announced that it has developed the world's first AI-based pig pregnancy diagnosis technology.

Generally, the pregnancy status of pigs is confirmed by the manager observing the sow's behavior from 21 days after artificial insemination and interpreting uterine ultrasound images after 25 days when the amniotic sac (baby pouch) is clearly visible. The interpretation of ultrasound images greatly depends on the manager's skill level, affecting the timing and accuracy of pregnancy diagnosis, and non-experts can only determine pregnancy status after 28 days.

Ultrasound imaging for pregnancy diagnosis of a sow.

Accordingly, the research team collected over 200,000 high-resolution (5MHz) uterine ultrasound images and performed AI training to create an AI model applying a computational method (algorithm) capable of confirming pregnancy status.

Additionally, they collected 230,000 low-resolution (3MHz) uterine ultrasound images mainly used in pig farming sites, applied low-resolution image enhancement technology, and trained the AI model to develop a pig pregnancy diagnosis program suitable for farm distribution.

An official from the Rural Development Administration explained, "If you take an ultrasound image of the sow's abdomen for more than 10 seconds using ultrasound equipment and then run the program, the AI will diagnose and inform you of the pregnancy status," adding, "It showed over 95% accuracy based on 22 to 25 days after artificial insemination."

Moreover, earlier pregnancy diagnosis extends the health management period for pregnant sows, enabling productivity improvement. In particular, non-pregnant sows can reduce non-productive days through re-artificial insemination, leading to feed cost savings. Non-productive days refer to the total period during which the sow is neither pregnant nor nursing piglets.

The Rural Development Administration has filed three patents related to AI-based pig pregnancy diagnosis technology and plans to transfer the technology to interested companies. They also plan to enhance the AI model to achieve over 95% diagnostic accuracy at 18 to 21 days of pregnancy, which is before the sow's estrous cycle.

Lim Gi-sun, Director of the National Institute of Animal Science at the Rural Development Administration, said, "In addition to pregnancy diagnosis, we are promoting the development of various AI technologies such as sow body condition management and piglet risk detection," and added, "We will continue to expand smart farm technologies that can increase productivity and reduce breeding costs to contribute to strengthening the competitiveness of pig farms."