PREDICTING INTRAUTERINE GROWTH RESTRICTION: A PILOT STUDY WITH FEED-FORWARD BACK PROPAGATION NETWORK

The prediction of intrauterine growth restriction (IUGR) represents a challenge for obstetricians throughout pregnancy, and the use of artificial intelligence could improve its screening. The aim of this pilot study was to prospectively design and test a Feed-Forward Backpropagation neural network (FFBPN) for the prediction of IUGR and its subtypes. Materials and methods: Between January and September 2023, we included 108 pregnant patients with singleton pregnancies who underwent conventional first trimester screening. Clinical and paraclinical data was used to construct a FFBPN, and its predictive performance was assessed using a sensitivity analysis. Results: Our results indicated that the FFBPN predicted the development of IUGR during pregnancy with a sensitivity ( Se) of 94.7%, specificity (Sp) of 97.7%, and a false positive rate of 2%. The value of the area under the curve for this neural network was 0.978. On the other hand, when used for the prediction of IUGR subtypes, the FFBPN achieved lower predictive performance. Also, the sensitivity analysis revealed that the FFBPN could better predict early IUGR, with a Se of 71.4%, Sp of 94%, and accuracy of 90.7%. Conclusions: The FFBPN could be an important tool for the prediction and stratification of various disorders, including IUGR.