<title>Computed radiographic examinations of subtle bone pathology: implications for liquid crystal displays in radiology</title>

Abstract

Examinations of the skeletal system have historically used 'detail' x-ray film-screen combinations having film-lightbox image quality that is challenging to replicate in a soft-copy environment. A review of the basic diagnostic imaging tasks for this sub-specialty is presented, including resolution and gray scale requirements for visualization of subtle fractures, bone mineral loss, implant loosening, and soft tissue disease. Measurements are made to specify image quality metrics in terms of Michelson contrast, spatial resolution, and dynamic range. Image contrast requirements are presented using Just Noticeable Differences (JND) values based upon the National Electrical Manufacturers Association (NEMA) and Digital Imaging Communication in Medicine (DICOM) Grayscale Softcopy Standard ofAnnex B, Part 14. (1). Image data is used to calculate modulation transfer function (MTF) for a type of liquid crystal active matrix display (monochrome) being considered for clinical use.No differences in visualization of image contrast or detail of subtle bone pathology was observed between film or cathode ray tube (CRT) viewing for any of the images, provided that the CRT display included window-level adjustment, zooming, and image re-location. This added activity on the part of the radiologist results in greater fatigue with CRT viewing.Measurements of image pixel data revealed that 'subtle' pathology wth orthopedic examinations had contrast values in the range of 1 to 10%. Actual measurements of displayed film and CRT conspicuity (using the metric of JND values) at the regions of pathology show no overall average change. However, about one-half of the images actually had a loss of JND values (mean loss of4O%) when viewed on CRT rather than on filmThe use of computed radiographic imaging plates having 1 00 micron resolution allows for the visualization of trabecular bone detail (mean diameter of 1 00 microns) on both film and with display optimized CRT. This implies that an equivalency of image quality detail for a liquid crystal display for orthopedic viewing would be '-1OO microns, i.e. at least 5 line pairs per mm. Our measurements of pixel size (100 microns) and fill factor (0.36) for the liquid crystal display we tested gave satisfactory results as determined by calculations ofMTF (7 line pairs per millimeter). However, the fill factor of this display substantially reduces overall luminance, and potentially contrast.