Pattern Recognition and Anomaly Detection in Fetal Morphology Using Deep Learning and Statistical Learning (PARADISE)

Congenital anomalies (CA) are the most encountered cause of fetal death, infant mortality and morbidity.7.9 million infants are born with CA yearly. Early detection of CA facilitates life-saving treatments and stops the progression of disabilities. CA can be diagnosed prenatally through Morphology Scan (MS). Discrepancies between pre and postnatal diagnosis of CA reach 29%. A correct interpretation of MS allows a detailed discussion regarding the prognosis with parents. The central feature of PARADISE is the development of a specialized intelligent system that embeds a committee of Deep Learning and Statistical Learning methods, which work together in a competitive/collaborative way to increase the performance of MS examinations by signaling CA. Using preclinical testing and clinical validation, the main goal will be the direct implementation into clinical practice. This multi-disciplinary project offers a unique integration of approaches, competences, breakthroughs in key applications in human, psychological, technological, and economical interest such as the 'smarter' healthcare system, opening new fields of research. PARADISE creates an environment that contributes significantly to the healthcare system, medical and pharma industries, scientific community, economy and ultimately to each individual. Its outcome will increase impact on the management of CA by enabling the establishment of detailed plans before birth, which will decrease morbidity and mortality in infants.

Study Overview

• Status: Recruiting
• Conditions: Congenital Abnormalities
• Intervention / Treatment: Other: Ultrasound

Detailed Description

Probe guidance: The IS guides the sonographer's probe for better acquisition of the fetal biometric plane - Basic scanning to be performed by non-expert(> 90% accuracy (AC)) Fetal biometric plane finder: The fetal planes are automatically detected, measured and stored - Insurance that all anatomical parts are checked (100% AC) Anomaly detection: unusual findings are signaled - Assistance in decision making (>90% AC)

• Study Type: Observational
• Enrollment (Estimated): 4000

Contacts and Locations

• Study Contact: Name: Dominic G Iliescu, Assoc. Prof.; Phone Number: +40 723888773; Email: dominic.iliescu@yahoo.com
• Study Contact Backup: Name: Smaranda Belciug, Assoc. Prof.; Phone Number: +40 729127574; Email: sbelciug@inf.ucv.ro

Study Locations

• Romania
  • Dolj
    • Craiova, Dolj, Romania, 200643
      • Recruiting
      • University Emergency County Hospital
      • Contact: Dominic G Iliescu, Assoc.Prof.; Phone Number: +40 723888773; Email: dominic.iliescu@yahoo.com

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years to 50 years (Adult)
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

The pregnant women that have their second trimester morphology scan scheduled. They are included in the study consecutively.

Description

Inclusion Criteria:

• Second trimester pregnant women

Exclusion Criteria:

-

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort

Intervention / Treatment

Second trimester; Second trimester fetal morphology Collect patient data, anonymize and label it. Written informed consent or verbal recorded consent (if the participant lacks the ability to write or sign) will be obtained before performing the MS. In the unlikely event that some of the participants will withdraw their consent after the ultrasound has been performed, the data collected will not be used in the project. Data for publications and dataset will be previously made anonymous following standard practices. The participants will sign a GDPR form.

Other: Ultrasound; Collect patient data, anonymize and label it. Written informed consent or verbal recorded consent (if the participant lacks the ability to write or sign) will be obtained before performing the MS. In the unlikely event that some of the participants will withdraw their consent after the ultrasound has been performed, the data collected will not be used in the project. Data for publications and dataset will be previously made anonymous following standard practices. The participants will sign a GDPR form. The DL/SL algorithms will work in a competitive/collaborative way. Following the 'no-free-lunch' theorem, we shall use the competitive phase to establish the most suitable DL/SL technique for the identification and anomaly detection of each organ, and the collaborative phase to make all the algorithms work together in providing a 'second' opinion.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Signal congenital anomalies	Number of congetinal anomalies found in a fetus at the second trimester morphology scan	32 months

Collaborators and Investigators

• Sponsor: University of Craiova
• Investigators: Principal Investigator: Smaranda Belciug, Assoc. Prof., University of Craiova

Publications and helpful links

General Publications

• Belciug S. Learning deep neural networks' architectures using differential evolution. Case study: Medical imaging processing. Comput Biol Med. 2022 Jul;146:105623. doi: 10.1016/j.compbiomed.2022.105623. Epub 2022 May 17.
• Belciug S, Ivanescu RC, Popa SD, Iliescu DG. Doctor/Data Scientist/Artificial Intelligence Communication Model. Case Study. Procedia Comput Sci. 2022;214:18-25. doi: 10.1016/j.procs.2022.11.143. Epub 2022 Dec 8.

Helpful Links

• Project website

Study record dates

Study Major Dates

• Study Start (Actual): May 4, 2022
• Primary Completion (Estimated): December 31, 2024
• Study Completion (Estimated): December 31, 2024

Study Registration Dates

• First Submitted: February 1, 2023
• First Submitted That Met QC Criteria: February 11, 2023
• First Posted (Actual): February 22, 2023

Study Record Updates

• Last Update Posted (Actual): September 13, 2023
• Last Update Submitted That Met QC Criteria: September 11, 2023
• Last Verified: September 1, 2023

More Information

Terms related to this study

• Keywords: ultrasound; deep learning; fetal morphology; view planes; learning curves
• Additional Relevant MeSH Terms: Congenital Abnormalities
• Other Study ID Numbers: PARADISE

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES

IPD Plan Description

Element 1: Data Type Primary and secondary data from RGB 2D ultrasound images (4000 images)

The data can be used by researchers to further improve the diagnostic of congenital anomalies.

Element 2: Related Tools, Software and/or Code:

There will be no specific tools for accessing data.

Element 4: Data Preservation, Access, and Associated Timelines

A. Repository where scientific data and metadata will be archived:

www.zenodo.org

When and how long the scientific data will be made available:

December 2024-December 2034

Element 5: Access, Distribution, or Reuse Considerations Informed consent, anonymized data, privacy constraints and applicable ethical norms, national laws, privacy policies

Element 6: Oversight of Data Management and Sharing:

Renato Constantin Ivanescu- anonymizing the data and collecting it Dominic Iliescu, Rodica Nagy, Anca Ofiteru, Cristina Comanescu - gathering data Smaranda Belciug - overall supervision role for data management

• IPD Sharing Time Frame: The data will become available in December 2024 and will be available until December 2034
• IPD Sharing Access Criteria: No criteria
• IPD Sharing Supporting Information Type: STUDY_PROTOCOL; SAP; CSR

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No