Program Information
Initial Investigation of a Vendor-Neutral Approach to Objective Patient Chest Image Quality Assessment
J Wells*, E Samei , Duke University Medical Center, Durham, NC
Presentations
WE-G-201-7 (Wednesday, August 2, 2017) 4:30 PM - 6:00 PM Room: 201
Purpose: This work aims to provide a vendor-neutral quality control tool for simple yet comprehensive assessment of patient image quality in the area of digital chest radiography.
Methods: The framework of an existing vendor-specific chest image quality analysis program [Med. Phys. 39, 7019-7031 (2012)] served as the basis for the current work. To fairly assess images from multiple vendors, DICOM Pixel Data (7FE0,0010) were converted to P-Values and adjusted to 12 bit depth. This incorporated the effects of window/level and LUT display functions while simultaneously producing pixel values linearly proportional to displayed luminance. Images were then rescaled and resampled in a manner consistent with the default function of medical display devices at our facility. To test algorithm performance, 79 AP chest radiographs from four mobile radiography units (four models from three vendors) were obtained with IRB approval. Image processing settings on each system had been initially set to deliver satisfactory image quality according to the preferences of a senior cardiothoracic radiologist. The algorithm measured ten image quality characteristics including graylevel, noise, and detail analyses of the lung, mediastinum, and subdiaphragm regions.
Results: Out of ten total image quality characteristics assessed, only one unit produced images with all average scores inside a previously defined “consistency” range; six of ten scores on the same unit fell within the “quality” range. The three other units each returned six of ten scores within the “consistency” range; two to three of these scores fell within the “quality” range on each unit.
Conclusion: The vendor-neutral nature of the proposed tool facilitated an objective comparison of image quality across vendors and systems. Results showed that, in spite of the subjective adequacy of sample chest radiographs from four mobile radiography units, objective image quality analysis of actual patient images revealed large discrepancies in average performance between individual systems.
Contact Email: