IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 1, NO. 3, SEPTEMBER 1997 161 Improvement of Sensitivity of Breast Cancer Diagnosis with Adaptive Neighborhood Contrast Enhancement of Mammograms Rangaraj M. Rangayyan, Senior Member, IEEE, Liang Shen, Yiping Shen, J. E. Leo Desautels, Heather Bryant, Timothy J. Terry, Natalka Horeczko, and M. Sarah Rose Abstract— Mammograms are difficult to interpret, especially of cancers at their early stages. In this paper, we analyze the effectiveness of our adaptive neighborhood contrast enhancement (ANCE) technique in increasing the sensitivity of breast cancer diagnosis. Seventy-eight screen-film mammograms of 21 difficult cases (14 benign and seven malignant), 222 screen-film mam- mograms of 28 interval cancer patients and six benign control cases were digitized with a high-resolution of about 4096 2048 10-bit pixels and then processed with the ANCE method. Unprocessed and processed digitized mammograms as well as the original films were presented to six experienced radiologists for a receiver operating characteristic (ROC) evaluation for the difficult case set and to three reference radiologists for the interval cancer set. The results show that the radiologists’ performance with the ANCE-processed images is the best among the three sets of images (original, digitized, and enhanced) in terms of area under the ROC curve and that diagnostic sensitivity is improved by the ANCE algorithm. All of the 19 interval cancer cases not detected with the original films of earlier mammographic ex- aminations were diagnosed as malignant with the corresponding ANCE-processed versions, while only one of the six benign cases initially labeled correctly with the original mammograms was interpreted as malignant after enhancement. McNemar’s tests of symmetry indicated that the diagnostic confidence for the interval cancer cases was improved by the ANCE procedure with a high level of statistical significance ( -values of 0.0001–0.005) and with no significant effect on the diagnosis of the benign control cases ( -values of 0.08–0.1). This study demonstrates the potential for improvement of diagnostic performance in early detection of breast cancer with digital image enhancement. Index Terms— Adaptive neighborhood contrast enhancement (ANCE), early detection of breast cancer, image processing, interval breast cancer, mammography, receiver operating char- acteristic (ROC) analysis. I. INTRODUCTION M AMMOGRAPHY is currently the best radiological technique available for early detection of nonpalpable Manuscript received December 3, 1996; revised July 11, 1997. This work was supported by the Alberta Breast Cancer Foundation, the Alberta Heritage Foundation for Medical Research, and the Natural Sciences and Engineering Research Council of Canada. R. M. Rangayyan, L. Shen, and Y. Shen are with the Department of Electrical and Computer Engineering, The University of Calgary, Calgary, Alta., Canada T2N 1N4. J. E. L. Desautels, H. Bryant, T. J. Terry, and N. Horeczko are with Screen Test, Alberta Program for the Early Detection of Breast Cancer, Calgary, Alta., Canada T2P 3G9. M. S. Rose is with the Department of Community Health Sciences, The University of Calgary, Calgary, Alta., Canada T2N 4N1. Publisher Item Identifier S 1089-7771(97)07956-9. breast cancer, one of the most common cancers among women [1], [2]. In order to reduce mortality, early detection of breast cancer is important since therapeutic actions are more likely to be successful in the early stages. However, the detection of early malignancies by mammography can be difficult in clinical practice because of their small size and subtle contrast compared with normal breast structures. The situation is even more challenging when radiologists routinely interpret large numbers of mammograms in screening programs where most of the cases are normal. Digital-image processing techniques have been applied to digitized mammograms during the past 10 to 20 years for vari- ous purposes: image quality improvement, mammographic fea- ture enhancement, malignant sign identification/analysis, and digital image compression to facilitate efficient archival and communication (teleradiology). Although a clinically proven computer-aided diagnosis (CAD) system for breast cancer is not yet available, research activity in computer-based pro- cessing and analysis of digitized mammograms has been increasing in recent years. Accurate diagnosis often depends upon the quality of mammograms; detection of small, low- contrast objects within the breast image is extremely important. Structures of radiological concern that are typically visi- ble in mammograms include microcalcifications, masses, and architectural distortion of breast parenchyma [3]–[7]. Unfor- tunately, contrast between malignant and normal tissues is often subtle. Hence, enhancement of contrast in mammograms, especially for dense breasts, may be useful. In 1974, Dronkers and Zwaag [8] proposed a photographic contrast-enhancement technique for mammograms using re- versal film; their results suggest that mammograms taken under nonideal conditions could be enhanced to make them more useful. In the early eighties, a photographic unsharp-masking technique for mammographic images was proposed by Mc- Sweeney et al. [9]. Ram [10] suggested that images considered unsatisfactory for medical analysis may be rendered usable through various digital-enhancement techniques. Considering the promising results of these initial investigations, various mammogram-enhancement algorithms have been developed and described since then; they can be sorted into three cat- egories: algorithms based on conventional image-processing methods [11]–[16], adaptive algorithms based on principles of human visual perception [17]–[23], and multiresolution- enhancement algorithms [24]–[30]. In order to evaluate the 1089–7771/97$10.00  1997 IEEE