Novel AI Model may measure Fetal Intracranial Markers During First Trimester for detecting CNS defects: Study

The progress made in first-trimester scanning has greatly broadened the ability to identify fetal structural abnormalities during early pregnancy. However, detecting central nervous system (CNS) defects such as open spinal bifida (OSB), agenesis of the corpus callosum, and posterior fossa malformations—particularly Dandy-Walker malformation—has traditionally posed challenges in the first trimester. Recent study focuses on establishing reference ranges for fetal intracranial markers during the first trimester and developing a novel artificial intelligence (AI) model to measure these markers automatically. This research addresses a critical gap, as prior to this study, no standardized reference ranges existed for these markers in the Chinese population, making early detection of fetal central nervous system (CNS) anomalies problematic.

Methodology

To achieve these objectives, the researchers conducted a retrospective analysis of two-dimensional (2D) ultrasound images from 4,233 singleton normal fetuses, scanned between 11+0 and 13+6 weeks of gestation. The study analyzed 10 key intracranial markers such as the brainstem, intracranial translucency, and the cisterna magna using standard ultrasound planes, thereby creating the first specific reference ranges for these markers. Measurements were then compared between the AI model and manual procedures to evaluate intra-observer consistency and time efficiency.

AI Model Development

The AI model was designed using advanced convolutional neural networks to interpret and measure the selected fetal markers from ultrasound images quickly and accurately. This model was trained on images manually annotated by experienced sonographers and underwent rigorous testing to validate its performance. According to the findings, the AI model demonstrated strong alignment with manual measurements, evidenced by high intraclass correlation coefficients (ICCs > 0.75) and Pearson correlation coefficients (> 0.75). Notably, the mean absolute errors were minimal, only ranging from 0.124 to 0.178 mm, indicating excellent precision. The AI was able to measure these markers in an average of 0.49 seconds, making the process over 65 times faster than traditional manual measurements, while also achieving a 100% detection rate for abnormal cases.

Significance and Implications

The significance of this study lies not only in the establishment of normal reference ranges for fetal intracranial markers, which enhances early screening for CNS malformations, but also in the successful integration of AI into routine prenatal care. The proposed AI model is expected to streamline the workflow for sonographers, reduce measurement errors, and heighten overall efficiency in fetal assessments during early pregnancy. Ultimately, this advancement could facilitate earlier diagnosis and clinical intervention, improving fetal outcomes. The research hence provides a substantial contribution to the field of obstetric ultrasound and paves the way for future AI applications in prenatal diagnostics.

Key Points

- The study aims to establish reference ranges for fetal intracranial markers during the first trimester, filling a gap in standardization for the Chinese population, which is crucial for early detection of fetal central nervous system (CNS) anomalies.

- Researchers conducted a retrospective analysis of 2D ultrasound images from 4,233 singleton normal fetuses, focusing on key intracranial markers such as brainstem and intracranial translucency, to create the first specific reference ranges for these markers.

- A novel AI model was developed utilizing advanced convolutional neural networks for the rapid and accurate measurement of selected fetal markers from ultrasound images, trained on manually annotated images by experienced sonographers.

- The AI model achieved high intraclass and Pearson correlation coefficients (> 0.75) relative to manual measurements, with mean absolute errors ranging from 0.124 to 0.178 mm, demonstrating exceptional precision and accuracy in measurements.

- The AI model significantly reduces the time for measuring markers, averaging only 0.49 seconds, being more than 65 times faster than traditional manual methods, while also achieving a 100% detection rate for abnormal cases.

- The study’s contributions enhance early screening for CNS malformations and pave the way for integrating AI in routine prenatal care, improving workflow efficiency, decreasing measurement errors, and potentially leading to better fetal outcomes through earlier diagnosis and clinical intervention.

Reference –

Lingling Sun et al. (2024). A Novel Artificial Intelligence Model For Measuring Fetal Intracranial Markers During The First Trimester Based On Two-Dimensional Ultrasound Image.. *International Journal Of Gynaecology And Obstetrics: The Official Organ Of The International Federation Of Gynaecology And Obstetrics*. https://doi.org/10.1002/ijgo.15762