A computer-based training system for breast fine needle aspiration cytology

Area of Science:

• Medical Informatics
• Pathology
• Artificial Intelligence in Medicine

Background:

• Fine-needle aspiration (FNA) cytology is a key diagnostic tool for breast disease.
• Subjectivity in FNA interpretation can lead to diagnostic errors.
• There is a need for standardized training and decision support in breast cytology.

Purpose of the Study:

• To develop and evaluate CytoInform, a diagnostic decision support system (DDSS) for breast FNAs.
• To implement a computer-based training (CBT) system for cytopathologists using a Bayesian belief network (BBN).
• To reduce diagnostic subjectivity and improve training in breast cytology.

Main Methods:

• Development of CytoInform, a DDSS based on a BBN for breast FNA diagnosis.
• Implementation of a CBT system guiding trainees through diagnostic features with visual clues and reference images.
• Generation of evidence vectors and updating diagnostic probabilities via the BBN.
• Comparison of trainee assessments with expert data from a teaching file.

Main Results:

• In trials with two pathologists, CytoInform proved effective in conveying diagnostic evidence and protocols.
• Each pathologist misinterpreted only one case.
• Trainees achieved high accuracy in assessing diagnostic clues (86% experienced, 88% inexperienced).
• The system ensured trainees followed expert-consistent diagnostic pathways.

Conclusions:

• CytoInform effectively reduces diagnostic subjectivity in breast FNA analysis.
• The integrated CBT system offers a novel and effective approach to cytopathology training.
• The system enhances diagnostic accuracy and promotes consistent decision-making aligned with expert protocols.