Configuring optical character recognition (OCR) rules in IMPAX

OCR (Optical Character Recognition) is the recognition of printed or written characters by a computer. If a modality generates images into the system but not enough information about a study is sent, OCR templates read information directly off the burned demographics. An OCR template is the definition of the size and shape of the characters, together with a layout that indicates the position of the characters on the image.

For older modalities, which are not capable of sending their images via DICOM, IMPAX can use OCR (Optical Character Recognition) to read the burned in demographics from images acquired with capture devices such as the Agfa MG3000. This requires a significant amount of work initially, but it works well.

To configure OCR rules in IMPAX

Acquire a set of images with samples of each character to be recognized (called a training set) and sample images showing the standard locations for each piece of information.
Log in as the ocr_train user with its password.
The OCR Training menu opens.
Use the acquired images to generate a template which gives IMPAX the shape of each character to be used for recognition (uppercase only and multiple examples in a row if analogue capture is used instead of digital capture).
Use the standard location images to generate a layout to define where information appears on the image.

For example, patient ID always in the top left corner, accession number below that, procedure code in the top right corner, and so on.
Save the layout in the mvf/etc/ocr folder.
Assign an OCR template to a particular modality in the Source Manager of the Administration Tools.

OCR Configuration also affects the modality operators:

The operators must use uppercase only. Number and punctuation are acceptable.
The operators must always follow the standard layout.
The operators must be aware that if the text is obscured by the image for all images in a batch, the OCR likely fails for the batch. In such problem scenarios, the operators may want to start each study with one clean image just for OCR.