A New Preprocessing Filter for Digital Mammograms

Abstract

Abstract. This paper presents a computer-aided approach to enhancing suspicious lesions in digital mammograms. The developed algorithm improves on a well-known preprocessor filter named contrast-limited adaptive histogram equalization (CLAHE) to remove noise and intensity inhomogeneities. The proposed preprocessing filter, called fuzzy contrast-limited adaptive histogram equalization (FCLAHE), performs non-linear enhancement. The filter eliminates noise and intensity inhomogeneities in the background while retaining the natural gray level variations of mammographic images within suspicious lesions. We applied Catarious segmentation method (CSM) to compare the segmentation accuracy in two scenarios: when there is no preprocessing filter, and when the proposed preprocessing filter is applied to the original image. The proposed filter has been evaluated on 50 real mammographic images and the experimental results show an average increase of segmentation accuracy by 14.16% when the new filter is applied.

Mammography segmentation with maximum likelihood active contours

“…The above results support the conclusion of our previous work (Rahmati et al, 2010) which stated that FCLAHE works better for a PDF-based segmentation algorithm, such as MLACMLS. …”

“…The size of the tumors in our data set was not more than 256×256; however, if the tumor can potentially occupy an area larger than 256×256, then larger ROI must be used. Prior to segmentation, we filter the images with the FCLAHE filter, which requires the setting of an "openness" parameter β (Rahmati et al, 2010). The FCLAHE is applied only for the MLACMLS; however, the other competing methods benefit from applying the noise reduction filter proposed in their respective papers.…”

“…In this section, we review the applied preprocessing filter (the FCLAHE) for mammogram proposed in our previous work (Rahmati et al, 2010). As per the conclusions of Pisano et al on the efficacy of the CLAHE (Pisano et al, 2000), the original mammographic images are first processed using the CLAHE to eliminate noise and then the result is processed using a nonlinear filter, which incorporates the advantages of a nonlinear fuzzy function.…”

“…Prior to applying MLACMLS to segmenting a suspicious mass in low-contrast mammography images (such as the images from the DDSM database), we process the images with an enhanced version of the contrast-limited adaptive histogram equalization (CLAHE) (Pisano et al, 2000), which we referred to as the fuzzy CLAHE (FCLAHE) (Rahmati et al, 2010). FCLAHE enhances the original mammographic image while minimizing the effect of changing the statistical distribution of its intensity, for more details see section 2.…”

An Analysis of Image Enhancement Techniques for Dental X-ray Image Interpretation

“…One of the parameter to apply image enhancement is by manipulating the image contrast [10]. Contrast enhancement techniques such as histogram equalization (HE), adaptive histogram equalization (AHE), and contrast limit adaptive histogram equalization (CLAHE) have been widely used to improve the contrast in medical images [5][11]- [13]. For example, HE and AHE methods had been used to enhance fish bone impaction in the soft tissue of lateral neck radiograph [2].…”