The objective of this paper is to evaluate the potential of utilizing a structured methodology for breast lesions screening, based on infrared imaging temperature measurements of a healthy control group to establish expected normality ranges, and of breast cancer patients, previously diagnosed through biopsies of the affected regions. An analysis of the systematic error of the infrared camera skin temperature measurements was conducted in several different regions of the body, by direct comparison to high precision thermistor temperature measurements, showing that infrared camera temperatures are consistently around 2°C above the thermistor temperatures. Therefore, a method of conjugated gradients is proposed to eliminate the infrared camera direct temperature measurement imprecision, by calculating the temperature difference between two points to cancel out the error. The method takes into account the human body approximate bilateral symmetry, and compares measured dimensionless temperature difference values (Δθ¯) between two symmetric regions of the patient’s breast, that takes into account the breast region, the surrounding ambient and the individual core temperatures, and doing so, the results interpretation for different individuals become simple and non subjective. The range of normal whole breast average dimensionless temperature differences for 101 healthy individuals was determined, and admitting that the breasts temperatures exhibit a unimodal normal distribution, the healthy normal range for each region was considered to be the dimensionless temperature difference plus/minus twice the standard deviation of the measurements, Δθ¯‾+2σΔθ¯‾, in order to represent 95% of the population. Forty-seven patients with previously diagnosed breast cancer through biopsies were examined with the method, which was capable of detecting breast abnormalities in 45 cases (96%). Therefore, the conjugated gradients method was considered effective in breast lesions screening through infrared imaging in order to recommend a biopsy, even with the use of a low optical resolution camera (160×120 pixels) and a thermal resolution of 0.1°C, whose results were compared to the results of a higher resolution camera (320×240 pixels). The main conclusion is that the results demonstrate that the method has potential for utilization as a noninvasive screening exam for individuals with breast complaints, indicating whether the patient should be submitted to a biopsy or not.
Read full abstract