Artificial Intelligence in Newborn Health – Project SENDD

IT company from Croatia TIS, in cooperation with neuropediatricians, has developed an AI system that can diagnose the neurodevelopmental disorder in infants based on video recordings.
This sophisticated tool for assessing the brain function of newborns is a revolutionary step, given that at such an early age it is difficult to diagnose this type of disorder.

Reading Time: 3 minutes

ai babies

Illustration: Milica Mijajlovic

What is SENDD? 

The name SENDD stands for System for Early Neurological Deviation Detection, a digital system for the early detection of potential neurodevelopmental deviations in infants. Thanks to this technology, early intervention is possible that would alleviate or completely stop the further development of the disease. 

TIS Group, which has been present on the regional tech scene for more than 30 years, has joined forces with neuropediatricians from the Sabol Children’s Hospital to come up with a unique solution that would contribute to the early intervention of high-risk neurodevelopmental deviations. 

IT experts worked on the development of a software solution for assessing the quality of infant spontaneous movements, the so-called General Movements Assessment, which is based on artificial intelligence technology, i.e. on computer vision for detecting infant poses aged 1.5-3.5 months. 

As you probably know from the ChatGPT story, what is crucial to the quality of the results that AI displays is the material with which the system is “fed.” In the case of SENDD, it is a medically proven method and classification of general movements assessment according to MD, Ph.D. M. Hadders-Algra, which over the past 30 years has repeatedly proven reliability in assessing outcomes, especially in high-risk children. 


Thirty years ago, scientists discovered that the quality of fidgety or spontaneous movements could be analyzed and linked to the development of deviations in mobility later in life. The more diverse and complex the flutter, the better, and if it is less variable, even uniform, stereotypical, it means that we can expect deviations in motor development.

Project team’s official statement 

Not only that, but neuropediatricians of the Polyclinic dr. Sabol, the first private children’s polyclinics in Croatia, are continuously involved in the project, by analyzing videos and classifying movements into one of four categories. They are a kind of tutor to this AI system who they guide, based on the collected videos, to predict what type of infant movement indicates a particular disorder and how certain it is. 

Therefore, the entire team working on this project started in 2020 is from Croatia, and is part of the activities of TIS’s internal AI/ML team, which was formed as a result of the desire to engage in the field of artificial intelligence through the development of globally recognized, unique solutions. 

How does SENDD work? 

Before answering the question about the way SENDD works, it is necessary to explain why this particular age dominated by fidgeting is key to the analysis. 


During infancy, the child acquires more and more skills, e.g. he begins to reach towards objects and grabs them, turns from the back to the peritoneal position, sits independently and that is when we first assess whether he deviates for the acquisition of a certain skill by age, and then whether he deviates from the quality of the acquired skills for age.

Project team’s official statement 

SENDD is currently in the test phase and is open to the public, so all interested parents can send a video showing the movements of a baby aged 1.5-3.5 months and receive a free general movements assessment. 

Since nowadays a good part of the population has smartphones with a high-resolution cameras, they can record video material usable for analysis by themselves, without going to the clinic, waiting for an appointment or the newborn experiencing any inconvenience and pain. 

However, this system cannot analyze just any video, so it’s necessary to comply with the following conditions: 

  1. The angle of the shot should be from a bird’s eye view; 
  2. The shot should be made under good lighting, without strong contrasts; 
  3. The child should be placed on the back, on a flat, neutral surface; 
  4. It is desirable that the child doesn’t wear clothes with sleeves and trousers; 
  5. The quality of the recording should be good, preferably high definition (HD); 
  6. The baby should be in an optimal mood, well-rested and fed; 
  7. There should be no distractions around the baby that would affect her movements; 
  8. The recording should last about ten minutes; 
  9. It is desirable to send more unique shots during the child’s first 2-3 months. 

The SENDD app can be downloaded for free and is available for Android and iOS devices. 

Positive reactions from parents confirming the reliability of this method are already coming. One of the comments of parents who tried analysis confirmed on LinkedIn that the SENDD diagnosis is correct, and that they received the same diagnosis from their doctor. 

What are the advantages of this method? 

One might wonder, especially if s/he is skeptical about new technologies and artificial intelligence, is this method more advanced and reliable than standard ones, such as magnetic resonance imaging, brain ultrasound and traditional neurological examination. 

As Tomislav Strgar, Head of the AI department at TIS, explained at the conference “AI2FUTURE2021”, it is a reliable, scientifically proven, harmless and non-invasive method, which is also faster and more cost-effective compared to the standard ones. 

“For us, this is not an ordinary project, it is not a business-as-usual project. We believe SENDD is a huge contribution to medicine, especially infant health care. I personally think that every parent around the world can identify with the benefits of this project,” Strgar said.   

As added by neuropediatrician Goran Krakar, his motivation for participating in this project was largely due to the fact that when he was a newborn he was told that he belonged to a high-risk group, but at the time there were currently no reliable methods that could accurately determine whether it was brain damage and what could the possible outcomes be. 

“General movements assessment was called a methodological breakthrough at the beginning of the millennium. Although it is simple, safe and cost-effective, it is not used in ordinary medical practice. The doctor’s assessment can be subjective, as it is based on observation; however, the moment AI is introduced as a way to objectively look at the continuum from normal to abnormal spontaneous movements, which allows you to re-validate the assessment, you get an objective result, and that’s where AI is tremendously helpful to us,” explained Dr. Krakar. 

A journalist by day and a podcaster by night. She's not writing to impress but to be understood.