Comparison of breast infrared imaging results by three independent investigators

The purpose of this study was to determine whether three investigators would produce similar results when reading/scoring the same breast infrared images. They used their standard methods of analysis (subjective to semi-objective analysis). Two of the investigators submitted breast infrared images from 71 screened patients. The images were stored in a database and displayed for scoring by the three investigators (two who submitted the data and one additional reader). The left and right breasts were analyzed separately. The investigators submitted their scores to the database without knowledge of the scores of the other two investigators. Overall, concordance of results among all three investigators was 76% (all three investigators' readings agreed on 107 of 141 breasts). Comparison of paired results of the three investigators resulted in 79 to 94% agreement for the six comparisons (three investigators and two breasts) with an overall agreement of 88% (371 of 424 paired comparisons). This preliminary comparison of three investigators' blinded results of breast infrared image readings from a screening population, demonstrates that breast infrared images taken at different centers with different techniques can be interpreted by different investigators with very similar results. As the database grows there will be an increase in the number and percentage of patients with abnormal infrared images (high risk and breast cancer patients). This will allow a better and more thorough analysis of the results to refine and standardize the reading technique, and further to allow the assessment of previously untested algorithms with this unique database.