Diagnosis of breast cancer with infrared spectroscopy from serum samples

The detection of breast cancer has a special value in the diagnosis of cancer diseases. It is the most frequent type of cancer among women's. We have developed a simple and rapid method for the detection of breast cancer with IR-spectroscopy. The method needs only 1 μl of a serum sample. The serum sample is dried on a suitable sample carrier such as a Si-plate. After drying the IR-spectrum is measured. Every disease leaves a typical fingerprint in the IR-spectrum of serum. This typical fingerprint can be used to identify different patient groups. The identification system can be trained by classification methods. We used two independent classification methods, cluster analysis and artificial neural networks (ANN). The study was carried out with 196 patients. With cluster analysis (a method of unsupervised learning) we achieved a sensitivity of 98% and a specificity of 95%. With ANN (a method of supervised learning) sensitivity of 92% and specificity of 100% was being determined. To sure that we do not have any interference with other diseases the breast cancer patients tested against 11 other diseases separately. Altogether, 3119 people took part in the study. The criterion was how many patients were assigned to the right group. 91% of all patients were assigned to the right group. Breast cancer was assigned to 79% to the correct group. These results suggest that IR-spectroscopy in combination with intelligent mathematical evaluation tools such as ANN or cluster analysis is a good tool for the diagnosis of breast cancer.