Reducing noise of mammographic LCD Displays

DESCRIPTION (provided by applicant): This SBIR Phase I proposal addresses the problem of image quality degradation due to noise of Active Matrix Liquid Crystal Displays (AMLCD), which are increasingly used as primary diagnostic softcopy display in clinical medicine, especially for mammography. The primary goal of this proposal is to explore the properties of LCD noise and to develop a reliable and easy to implement method to evaluate and reduce it. LCD spatial noise is defined as small but detectable and stationary inter- and intra- LCD pixel luminance differences. Quantization noise results from insufficient quantization (bit-depth) of the display. We will use a high quality CCD (charge coupled device) camera for physical evaluation. The spatial noise of the central portion of a 5-Mpixel EIZO LCD will be evaluated. Spatial noise properties will be analyzed and estimated from the camera images via signal modeling and image processing. An error diffusion based algorithm to compensate for the LCD spatial noise as well as the quantization noise will be used to process images before they are displayed. We will use the Visual Image Quality Metric (VIQM) model developed at Siemens Corporate Research with channelized model observers to predict the effects of noise compensation on human detection performance. A human observer study will also be conducted to verify the predictions of the model and evaluate the impact of noise reduction on actual diagnostic performance. The proposed research focuses on the reduction of mammographic LCD display noise. Displays with reduced noise could lead to easier or earlier detection of disease in Full Field Digital Mammograms (FFDM). If the detection of breast cancer can be accomplished at an earlier stage, there will be a higher chance for successful therapy. An error-diffusion-based display noise reduction scheme will be designed and implemented. A human visual model developed at Siemens will be used to quantitatively evaluate the performance of the approach. A pilot human observer study using sections of FFDMs will be conducted to evaluate the impact of noise reduction scheme on diagnostic performance.