For breast cancers discovered while the lesions are still small, the prognosis is extremely good. Early detection could truly be said to be the best treatment method for this disease. Most women are familiar with mammography, the examination that plays a central role in breast-cancer screening. This exam involves X-ray exposure to the breast while it is compressed using a dedicated mammography unit, to create an image known as a mammogram.
Example of a mammogram
The cloud of white specks visible in the magnified portion is clustered microcalcifications, which is an indicator associated with breast cancer.
Interpreting mammograms (reading and identifying mammographic findings in the images) is actually one of the most difficult tasks in radiographic diagnoses, and requires great concentration. Breast-cancer lesions appear as white densities in mammograms. They can be as small as several hundred microns to several millimeters in diameter. They have to be detected among the camouflaging cloud-like shadows of the mammary gland tissue, which also appears white in the mammograms.
Furthermore, breast cancer screening programs generate enormous number of mammograms every day. This creates an urgent need to lighten the workload of the medical professionals who interpret the mammograms. The computer-aided detection system developed by Konica Minolta is expected to meet this need.
By performing highly-accurate computer-aided detection of shadows indicating suspected breast-cancer lesions, and making the detection results available to the radiologist interpreting the mammogram at the time, this system can help raise the degree of certainty of the radiologist's diagnosis, by, for example, preventing lesions from being missed. By reducing the stressfulness of the task of interpreting mammograms, it can also make a major contribution to the early detection of breast cancer.
By applying advanced pattern-recognition and discrimination algorithms, Konica Minolta has developed a computer-aided detection (CAD) system that automatically detects shadows indicating suspected lesions in mammograms. By performing computerized analyses on the abundant information contained in mammograms, this system supplies added value that aids the radiologist's diagnosis.
The Konica Minolta CAD system was the subject of an evaluation study in which 6 qualified radiologists interpreted clinical mammograms pertaining to 168 cases. According to an academic article* on this study, it was confirmed, by a statistically-significant difference, that the radiologists' lesion-detection performance was higher when the CAD system was used, than when it was not used.
*Takako MORITA,“Kenshin manmogurafii no dokuei to CAD” (CAD and the interpretation of screening mammograms),Rinsho Gazo (Clinical Imagiology) Vol.24, No.4, pp. 408-415 (2010)
Comments made by radiologists using the Konica Minolta system include the following: “When I can have lesions as faint and fine as this pointed out to me, it takes a weight off my mind”, “This is helpful when I start to get tired after spending long hours interpreting mammograms” and “This makes me feel less anxious about missing lesions”.
To detect shadows indicating suspected lesions, the first requirement is accumulating an image database of breast-cancer cases. To pick out shadows characteristic of breast cancer in acquired images, algorithms are designed by applying advanced pattern-recognition techniques. Information is obtained from images of a vast number of cases covering all sorts of conditions. Some of these images feature breast-cancer lesions, while others feature similar-looking shadows that have proved not to be abnormal. This information is used to improve detection performance of the algorithms.
In the detection of typical breast-cancer indicators such as microcalcifications and mass shadows, pre-processing technology plays a key role. The Konica Minolta system selectively enhances the targeted lesions, using proprietary multi-resolution processing, filtering, contrast compensation and other means.