Untutored discrimination training on paired cell images influences visual learning in cytopathology.

BACKGROUND

Cytologists must learn how to discriminate cells that might be visually very similar but have different neoplastic potential. The mechanism by which trainees learn this task is poorly researched and is the focus of the current investigation. Cognitive science offers a theoretical platform from which to design meaningful experiments that could lead to novel training strategies.

METHODS

The interpretation of a cell image is a category-discrimination task, and the process by which discrimination improves with practice is called perceptual learning. The study authors operationalized this concept by training 150 naive observers on paired cell images without providing explicit tuition, employing cervical cytology as a model system. Six strategies were tested, which differed according to the diagnostic category and level of interpretive difficulty of each image. Participants were tested before and after training to determine the extent to which visual learning had occurred.

RESULTS

Diagnostic accuracy improved for participants who were trained on normal/abnormal image pairs in which at least one member of the pair was "easy" to interpret (P < .05). Training was not effective when image pairs were drawn from the same diagnostic category or when both members of the pair were "difficult" to interpret (P > .05).

CONCLUSIONS

Training on paired cell images without explicit tuition can be an efficient and effective means of visual learning in cytopathology, but only if care is taken to avoid image pairs from category boundaries. Training on same-category image pairs is ineffective. This study is a step toward the development of perceptual learning modules for cytopathology.