Breast cancer is the leading cause of death, linked to primary cancer. Screening of Thermogram images, the most robust method for early diagnosis of breast cancer is widely recommended with the introduction of several Computer Aided Diagnosis (CAD) techniques. The main difficulties of the thermography asymmetrical temperature distribution leads to abnormality for even disease. The authors have presented one of the fastest pattern recognition techniques that have been more efficient in the classification of tumors as benign or malignant – Fast Support Vector Machine (FSVM). This method has been developed and implemented in statistical learning theory over the past decade and they gave promising classification results for efficient tumor diagnosis. The main objective of the proposed work is to help in diseases diagnosis by Thermogram analyses applying a three phase approach. In the first phase of work, Thermogram images apply preprocessing and is segmented by separating left and right portions of the breast regions. After segmentation process (second phase), some textural features are extracted using Discrete Curvelet Transform (DCT): temperature range, mean temperature, standard-deviation, and the quantization of higher tone in an eight level pasteurization. This last feature considers the entire image temperature and measures the percentage of area occupied by pixels with the higher temperatures of the image. In the final phase of the work, a supervised learning method based on Fast Support Vector Machine (FSVM) is used for the extracted feature classification. The features are extracted from a set of 50 images confirmed by physician diagnosis. The proposed method achieved the average results of accuracy 98.5%, sensitivity 96%, and specificity 96.5%.