The Role of Artificial Intelligence in Non-Invasive Prenatal Testing and Genetic Abnormality Prediction

Written by: Julia Richie

The early weeks of pregnancy can be a period of excitement, hope and vulnerability. For

soon-to-be parents, these feelings may be accompanied by a flurry of questions and uncertainties.

Among some of the most potent sources of uncertainty are often questions of genetic health: Is

the baby healthy? Could there be any risk of chromosomal conditions? These questions, once

impossible to answer, are now illuminated by an array of detection and diagnostic tools, now

perhaps accompanied by an unexpected partner: artificial intelligence.

Non-invasive prenatal testing (NIPT) has already transformed how we screen for chromosomal

conditions in early pregnancy and has become a widely adopted tool for aneuploidies (abnormal

number of chromosome copies) such as trisomy 21 (Down Syndrome), trisomy 18 (Edwards

Syndrome) and trisomy 13. These methods can offer soon-to-be parents a glimpse into the

genetic profile of their unborn child and can provide clinicians with valuable risk assessment

information, sometimes pointing towards further testing. Although NIPT does not specifically

diagnose genetic disorders, it analyzes cell-free fetal DNA (cffDNA) circulating in the mother's

blood to detect pregnancies that may be at a higher risk for chromosomal abnormalities.

However, like all technologies, NIPT does not come without its limitations. The accuracy of

results is vulnerable to factors such as sequence noise, low fetal fraction, and biological

variability, which can obscure subtle abnormalities or lead or uncertain results. In some cases,

this may result in false positives or false negatives, especially when the fetal DNA concentration

is too low or the signal-to-noise ratio is suboptimal.

But what happens when we integrate the power of AI (artificial intelligence) into this process?

AI, primarily through machine learning and deep learning, brings a new and revolutionary

dimension to prenatal testing, filled with increasing accuracy, precision, and speed to detect

patterns invisible to the human eye. Rather than relying solely on pre-set thresholds or rule-based

decision trees, AI systems are trained on vast genomic and clinical datasets. These models can

learn to interpret subtle trends, assess multiple risk factors simultaneously, and refine predictions

based on real-world variability.

AI may also enable practitioners to provide personalized recommendations to patients, for

example, by taking external factors such as maternal age, gestational week, previous pregnancy

history, and fetal fractional variability into account—something that standard NIPT screening

methods are not equipped to do.

The integration of AI into NIPT processes may also add complementary layers to its scope of

detection. A 2023 study by Troisi et al. involves combining AI with maternal serum

metabolomics. This method analyzes hundreds of small-molecule metabolites in maternal blood

and uses AI to identify biochemical patterns associated with both chromosomal and structural

abnormalities. In their study of over 2500 samples, the AI-powered model achieved an accuracy

of 99.4%, a specificity of 99.9%, and a sensitivity of 78%, proving capable of detecting defects

missed by both NIPT and routine ultrasound. A broader review in 2024 by Boddupally and

Thuraka confirms that these innovations are not isolated. Across dozens of studies, AI models

have been applied to fetal DNA analysis, fetal fraction prediction, and chromosomal mosaicism

detection with extremely promising results. This suggests that AI may serve as a bridge across a

significant diagnostic gap through screening for a broader range of fetal abnormalities in the

same non-invasive and low-cost manner as NIPT.

Still, this emerging field faces challenges. Many models are trained on narrow datasets that lack

ethnic and socioeconomic diversity, limiting their generalizability. There are also ethical

concerns around overreliance on automated predictions in highly sensitive clinical decisions.

Further, while AI can suggest a likely diagnosis, it cannot replace human empathy and

context-based judgment, especially in prenatal genetic counseling, where decisions often carry

emotional and moral weight.

The implications for prenatal diagnosis are vast and profound. Through the combinatory use of

AI with NIPT, clinicians can offer expecting parents earlier and more comprehensive information

about their unborn child’s genetic well-being, empowering patients to make increasingly

informed decisions for proactive care planning. The fusion of AI and NIPT-based chromosomal

and genetic detection may serve not only as an upgrade but as a transformative and impactful

tool to empower expectant parents with the knowledge to make informed decisions for their care.

Bibliography:

Jacopo Troisi, Martina Lombardi, Giovanni Scala, Pierpaolo Cavallo, Rennae S. Tayler, Steven

J.K. Symes, Sean M. Richards, David C. Adair, Alessio Fasano, Lesley M. McCowan &

Maurizio Guida. (2023). A screening test proposal for congenital defects based on maternal

serum metabolomics profile. American Journal of Obstetrics and Gynecology.

https://doi.org/10.1016/j.ajog.2022.08.050.

Boddupally K, Rani Thuraka E. (2024). Artificial intelligence for prenatal chromosome analysis.

Clin Chim Acta. https://doi.org/10.1016/j.cca.2023.117669.