AIRFRAME: Artificial Intelligence for Recognition of Fetal bRain AnoMaliEs at Second Trimester Fetal Brain Scan

Official title

Development of an Artificial Intelligence Algorithm to Recognize Abnormal Findings at Routine Fetal Brain Ultrasound. AIRFRAME (Artificial Intelligence for Recognition of Fetal bRain AnoMaliEs)

Overview

Obstetric ultrasound represents the standard of care for the screening of the fetal anomalies. However, its performance is dependent upon several parameters including type of anomaly, gestational age, maternal habitus and skills of the examiner. The use of Artificial Intelligence (AI) in medical diagnostics has been suggested not only to reduce the inter- and intra-operator variability, but also to compress the required time necessary to perform routine tasks, hence optimizing healthcare resources. Fetal brain abnormalities are among the most challenging fetal congenital anomalies in terms of ultrasound diagnosis, prenatal counseling and management. The access to new sources of technology, i.e. AI, has the potential to improve recognition, detection and localization of brain malformations. Therefore, we propose to develop an AI-based software, which would be capable to recognize the brain structures at antenatal ultrasound and discriminate between normal and abnormal fetal brain anatomy through fully automatic data processing.

Detailed description

The application of AI in obstetric ultrasound includes three aspects: structure identification, automatic and standardized measurements, and classification diagnosis. Since obstetric ultrasound is time-consuming, the use of AI could also reduce examination time and improve workflow.

Study design: this is a multicenter retrospective observational cohort study and subsequent prospective cohort study. The study design will be organized in two different phases.

The first phase, the feasibility retrospective study, has the objective to develop and train AI-Algorithm with normal and abnormal images retrospectively acquired during second trimester ultrasound scan from various international fetal medicine centers.

The second phase, a prospective clinical validation, has the objective to test the AI-Algorithm in the assessment of basic fetal brain anatomy in a real clinic setting with real patients from each of the participating fetal medicine centers.

Setting: Three (3) fetal medicine centers.

Participants: singleton pregnant population who underwent ultrasound examination between 19 - 22 weeks of gestation in the participating centers.

Primary endpoint: to validate a novel AI-based technology for the automated assessment of the basic anatomy of the fetal brain which could potentially be used to support second trimester screening scan.

Secondary endpoints:

To improve the performance of the standard second trimester screening of fetal brain anatomy ensuring its reliable sonographic assessment within a shorter time of execution.

To detect higher repeatability and reproducibility, allowing to implement the ultrasound screening also in terms of efficiency on a vast scale, optimizing healthcare resources In the first phase of the study, participating fetal medicine centers will search their electronic databases for images of singleton pregnant women who underwent ultrasound imaging at 19+0 - 22+6 weeks of gestation with any fetal brain anomaly. Normal images of the fetal brain at the same gestational age will be provided by the promoting centers - i.e., Fondazione Policlinico A. Gemelli, IRCCS and University of Parma. Clinical, ultrasound, prenatal and postnatal information of each case will be retrieved from patient's medical records and entered an electronic database collection file by the principal investigator from each participating center. The acquired images will be anonymized, saved as DICOM and shared through a dedicated cloud storage system which will be set up by the bioengineering team. Each center will be able to access the web system using a personal ID and password.

In the second phase of the study, the algorithm will be prospectively tested and validated in a real clinical setting with real patients from each of the participating fetal medicine centers. Inclusion and exclusion criteria, imaging protocol and data collection will be the same carried out during the retrospective phase.

Interventions

• Diagnostic test Development of AI algorithm for early detection of fetal brain anomalies in the second trimester screening scan
  Development of AI algorithm for early detection of fetal brain anomalies in the second trimester of pregnancy

Primary outcome measures

• AI algorithm [Time frame: 2 years]

Secondary outcome measures (1)

• Reproducibility [Time frame: 1 year]

Eligibility criteria

Inclusion criteria

• Women with singleton pregnancies undergoing ultrasound examination between 19+0 - 22+6 weeks of gestation

Exclusion criteria

• Women who did not have the second trimester screening scan at the settled gestational age.
• Women in which a good visualization of the transventricular, transthalamic and transcerebellar plane of the fetal head was not technically possible.
• Women who are not able to give the informed consent.

Criteria are shown verbatim from the registry (in English). Final eligibility is always assessed by the study center.

Healthy volunteers: Yes

Study design

Observational model

Cohort

Study locations

Italy · 1 center

• Fondazione Policlinico Universitario Agostino Gemelli — Rome

Publications

• Chen H, Wu L, Dou Q, Qin J, Li S, Cheng JZ, Ni D, Heng PA. Ultrasound Standard Plane Detection Using a Composite Neural Network Framework. IEEE Trans Cybern. 2017 Jun;47(6):1576-1586. doi: 10.1109/TCYB.2017.2685080. Epub 2017 Mar 30. PMID 28371793
• Yu Z, Tan EL, Ni D, Qin J, Chen S, Li S, Lei B, Wang T. A Deep Convolutional Neural Network-Based Framework for Automatic Fetal Facial Standard Plane Recognition. IEEE J Biomed Health Inform. 2018 May;22(3):874-885. doi: 10.1109/JBHI.2017.2705031. Epub 2017 May 17. PMID 28534800
• Yaqub M, Kelly B, Papageorghiou AT, Noble JA. A Deep Learning Solution for Automatic Fetal Neurosonographic Diagnostic Plane Verification Using Clinical Standard Constraints. Ultrasound Med Biol. 2017 Dec;43(12):2925-2933. doi: 10.1016/j.ultrasmedbio.2017.07.013. Epub 2017 Sep 28. PMID 28958729
• Ambroise Grandjean G, Hossu G, Bertholdt C, Noble P, Morel O, Grange G. Artificial intelligence assistance for fetal head biometry: Assessment of automated measurement software. Diagn Interv Imaging. 2018 Nov;99(11):709-716. doi: 10.1016/j.diii.2018.08.001. Epub 2018 Sep 1. PMID 30177447
• Rydberg C, Tunon K. Detection of fetal abnormalities by second-trimester ultrasound screening in a non-selected population. Acta Obstet Gynecol Scand. 2017 Feb;96(2):176-182. doi: 10.1111/aogs.13037. Epub 2016 Nov 22. PMID 27714775
• Hendricks KA, Simpson JS, Larsen RD. Neural tube defects along the Texas-Mexico border, 1993-1995. Am J Epidemiol. 1999 Jun 15;149(12):1119-27. doi: 10.1093/oxfordjournals.aje.a009766. PMID 10369506
• Klusmann A, Heinrich B, Stopler H, Gartner J, Mayatepek E, Von Kries R. A decreasing rate of neural tube defects following the recommendations for periconceptional folic acid supplementation. Acta Paediatr. 2005 Nov;94(11):1538-42. doi: 10.1080/08035250500340396. PMID 16303691
• De Wals P, Rusen ID, Lee NS, Morin P, Niyonsenga T. Trend in prevalence of neural tube defects in Quebec. Birth Defects Res A Clin Mol Teratol. 2003 Nov;67(11):919-23. doi: 10.1002/bdra.10124. PMID 14745929

Identifiers

NCT: NCT06675266 · Prod. ID 4813

Primary sources (government registries)

• ClinicalTrials.gov (US NIH)

View this study on ClinicalTrials.gov ↗