Fetal Ultrasound Image Segmentation and Biometric Measurements

Develop and evaluate an artificial-intelligence workflow for fetal ultrasound image segmentation and biometric measurements, using a clinically curated dataset and a prespecified reference standard. The project will prioritize clinically meaningful validation rather than model accuracy alone: internal holdout testing, class-wise performance, calibration where applicable, subgroup performance, error analysis and transparent reporting will be incorporated into the final study design. The final model architecture and validation strategy will be selected by the microscopIA engineering and data-science team after dataset audit and label-quality assessment.

This project is designed to produce decision-relevant evidence for clinical practice, health-system planning or research. Participation combines applied training, verifiable scientific work and contribution to a publication-quality manuscript.

Development and validation of an AI model using a clinical reference standard, quality control, internal validation, class-wise metrics, error analysis and subgroup assessment.

Clinically curated imaging or biomedical datasets with prespecified reference-standard labels.

Participants will work primarily on clinically supervised data curation and quality control. • Complete project-specific annotation and adjudication training. • Label or verify images/signals according to the approved data dictionary and reference standard. • Conduct duplicate quality-control review and resolve annotation discrepancies with the assigned clinical expert. • Support literature review, dataset documentation, error taxonomy and clinically meaningful subgroup definitions. • Contribute to interpretation, manuscript drafting, revision and scientific dissemination. Programming experience is not required for participant positions. Model development and statistical processing will be conducted by the microscopIA data-science and engineering team.

microscopIA is designed so participants can progress at their own pace within project deadlines. Work is primarily asynchronous and there is no fixed weekly schedule. An average commitment of 5–7 hours per week is anticipated and can be distributed flexibly. Mandatory meetings are kept to the minimum necessary. Under a self-governance model, each team may choose to meet more often when that improves coordination. The key expectation is to complete agreed tasks and deadlines to a high standard. The model is designed to remain compatible with medical school, rotations, clinical duties, call schedules and other professional responsibilities.

Clinical data curation Quality control Interpretation of AI metrics Scientific writing

microscopIA will provide annotation manuals, calibration cases, reference-standard definitions, data-governance guidance, quality-control workflows and project-specific clinical teaching. Assigned clinicians will adjudicate uncertain labels; microscopIA data scientists and engineers will lead model development and statistical validation.

Expected outputs include a quality-controlled research dataset, documented annotation protocol, trained and validated AI model, error and subgroup analysis, reproducible performance tables/figures, and a manuscript intended for a PubMed-indexed journal.

Manuscript intended for submission to a PubMed-indexed journal; preprint whenever appropriate.

Active microscopIA membership is required. Working language: English.

Medical students, junior doctors and early-career professionals interested in Artificial Intelligence, Obstetrics, Maternal–Fetal Medicine, with attention to detail, independence and the ability to deliver high-quality work in a flexible environment.

No programming certificate required; project-specific annotation calibration is mandatory.

microscopIA will cover approved article-processing or publication fees for the anticipated manuscript. Participants will not be required to personally pay approved publication charges. Coverage is subject to the project reaching manuscript quality, compliance with the approved protocol, and submission through the microscopIA research program.

The project is designed to produce a substantive research record and a manuscript intended for submission to a PubMed-indexed journal. Authorship is contribution-based and follows ICMJE criteria: substantial contribution to the work, drafting or critical revision, final approval, and accountability. A preprint will be prepared whenever appropriate. Poster presentation opportunities may be assigned based on project performance and conference fit. Certificates, reference letters and other academic documents remain subject to the applicable microscopIA Membership and Project Terms.

Each participant may declare a maximum of two institutional affiliations. Format: Department, Organization (Hospital, University, Company), City, Country.

Jorge Racedo at microscopIA; External Clinicians. Senior clinician or subject-matter expert allocated by microscopIA.