Abstract
Background: Hyperplasia of mammary glands (HMG) is a frequent disease, with increased cancer risk for women aged 20-55 years. The aim of this study was to explore a non-invasive method to identify which patients with breast complaints need additional mammography for HMG diagnosis. Patients and Methods: Skin digital infrared thermal imaging (DITI) in 74 patients with HMG and 63 controls was carried out. Results: In the controls, the temperature of points close to the breasts and ovaries decreased with age. In women older than 39 years, HMG patients showed persistently high temperatures but in the lower extremities there were no differences. With a threshold for thoracic skin point KI21 of 33.2°C, sensitivity and specificity in distinguishing controls from HMGs were 96% and 52% (p=0.0001) respectively, as validated in a test set, similar to recent DITI results for breast cancer detection. Conclusion: Infrared temperature imaging of specific skin points is a rapid, non-invasive method to identify patients requiring mammography to confirm HMG.
Breast cancer is the most frequent malignancy among women in the Western world and the incidence has sharply increased over the past four decades (1, 2). In contrast, the incidence in China, more specifically Shanghai, has traditionally been very low (5/100,000), but since 1995 there has been a sharp (eight fold) increase in the incidence from 5 (1985-1995) to 40/100,000 (in 2005) (3-5).
Many women between 20-55 years of age with complaints such as uni- or bilateral breast hypersensitivity or pain, worsening around or during menstruation, are understandably concerned and may seek medical attention. In most cases, clinical examination does not show anything specific. However, in a certain percentage of patients, physical evaluation reveals that one or both breasts have single or multiple small nodular changes, the borders of which are not well defined, have no cohesion with the surrounding tissue and are therefore easily movable. If in addition, molybden target X-ray mammography shows no signs of cancer but high-density mammographic parenchymal patterns, as measured by the proportion of breast area composed of epithelial and stromal tissue are detected (6), the combination of these clinical-palpatory and radiological findings is called hyperplasia of the mammary glands (7) (see Table I).
As Figure 1 shows, HMG is not the same as mammary dysplasia, a histopathological condition which develops through a gradual process from normal breast tissue to hyperplasia, atypical hyperplasia, to carcinoma in situ (CIS), and invasive carcinoma (8). Mammary dysplasia is a precancerous lesion and in some dysplasias, invasive cancer can be found in the follow-up. The risk of breast cancer in HMG is however, much lower, although is still higher when compared to healthy women (7). This means that performing mammography (which can be an unpleasant procedure) is overtreatment in those women with HMG who may not actually develop breast cancer. Therefore, it could be of great benefit to develop a method to identify those patients with breast complaints which are not HMG, as such women do not require mammography. Such a method should be rapid, easily applicable, inexpensive, sensitive and if possible, non-invasive and harmless.
Digital infrared thermal imaging (DITI) was introduced into medicine in the late 1950s. Early studies suggested that the method could be useful for breast cancer detection, amongst others. But the early instrumentation was not sensitive enough to detect the subtle changes in temperature needed for accurate detection and monitoring of breast disease. However, in recent years the sensitivity and resolution of infrared instruments has greatly improved. One recent study on breast biopsies which were recommended based on prior mammogram or ultrasound indicated that DITI is a valuable adjunct to mammography and ultrasound, especially in women with dense breast parenchyma (9).
In agreement with these findings, we recently found that the infrared radiation temperature of certain skin points is significantly different in patients with HMG when compared with healthy women of the same age group (20-55 years). In order to increase the sensitivity of DITI for identifying HMG, it remains to be assessed whether these findings are age dependent (10). If so, DITI could be used in specific age groups as a simple outpatient clinic system, identifying those patients with clinical HMG signs who do and those do not require additional mammography. We have therefore analyzed 74 patients with clinically radiologically confirmed HMGand 63 healthy control women of the same age. The 2008 World Health Organization (WHO) system was used to select the different skin points for measurement (11).
Patients and Methods
Diagnostic criteria, patients and healthy controls. The 2002 diagnostic criteria for HMG (7) were used. Seventy-four patients with HMG were diagnosed at the Longhua Hospital, Shanghai, China. During the same period, healthy women in the same age groups as the patient group were selected from the employee file of the Shanghai University of Traditional Chinese Medicine. Participation was strictly voluntary (the study resources did not allow us to provide a volunteer for all 74 patients). Each of the 63 volunteers was given a thorough physical examination by a medical specialist in breast disease and additional standard breast examinations were performed as indicated by the 2002 criteria (7). None of the volunteers had any signs of breast, endocrine of hormonal abnormalities, neither at examination nor in her history. The diagnosis in all cases was confirmed by B-mode ultrasonic examination or molybdenum target X-ray.
Determination method. The ThermaCAM™ P30 infrared thermal imaging system manufactured by the FLIR System Company in Sweden was selected to determine room temperature of 22(±2)°C, and relative humidity of 40(±10)%, in a dark room with two 36-watt fluorescent lamps. There was no obvious air flow, no loud noises and no electromagnetism in the room. Before determination, the participants were requested to enter the laboratory, untie their bra, and sit for 30 min to adapt to the surroundings. After that they were asked to sit up erectly, and expose the body parts to be assessed. The examiner adjusted the lens to capture the selected locations from a fixed position 1 m away. Figure 2 shows the examples of the measurement results. In the control group, there was no difference between the control points on the left and right trunk (p=0.56) nor the left and right feet (p=0.20). As there was no correlation between the points on the trunk and the feet, points on the trunk and feet were always used for control purposes.
WHO 2008 skin points. The infrared radiation temperature image on 10 skin points was recorded. The location of points is based on the location of points stipulated by the 2008 Standard Skin Points according to the World Health Organization (11). Figure 3 shows where these points are located on the body surface.
Statistical methods. The statistical management was performed with the SPSS15 software package (SPSS inc, Chigago, Il, USA). Normality of the distributions was tested and no differences were found. The infrared radiation temperature comparison on body surface of points was made by Student's t- and Mann-Whitney tests (whenever applicable). Correlation analysis used the Pearson and Spearman tests. We found practically the same results in the parametric and non-parametric test, and present the latter. Scatter plots and box plots were used for graphical analysis of the data. Receiver operating curve (ROC) analysis was used to discretize the continuous variables for stepwise multivariate regression analysis, using the diagnostic groups (healthy controls, HMG patients) as the classification variable. Random numbers were used to create a set for learning and testing the results. Moreover, the results were cross validated, in which each case was classified by the functions derived from all cases other than that case (leaving one out method). Sensitivities, specificities, positive and negative predictive values and overall predictive values were calculated. Multivariate binary logistic analysis was performed to find the most optimal combined set of points discriminating HMG patients and healthy controls.
Results
The age ranged from 20-55 years (mean and standard deviation: 36.3±8.8 years) for the 74 HMG patients and from 20-50 years (mean and S.D.: 34.5±8.9) for the 63 healthy controls. The difference in age was not significant (p=0.63).
In the total group of 74+63=137 women, the temperature of many points on the trunk decreased with age. However, separate analysis in healthy and HMG women showed that there is a difference between the two groups (Table II). Whereas in the controls the temperature of nearly all points decreased with age, in HMG patients the temperature of several points persisted at a high level. The persistent high temperature of trunk points mainly concerned HMG patients over the age of 39 years (p<0.0001, Figures 4 and 5) and was especially evident in points close to the breasts (WHO skin points KI21 and LR14), and in the midline of the lower abdomen (point CV4). On the lower extremities, the skin temperatures were different from the findings on the trunk as the temperature was found to decrease with age both in the healthy controls and HMG, and was therefore basically the same in the two groups studied.
As the results are clearly most significant for those over the age of 39, these patients were further investigated. In order to obtain more reliable results, the analyses were carried out of the learning set and validated in the tests set. The temperature at points KI21, LR14 and CV4 gave the most significant discriminating features. However, KI21 and LR14 were strongly correlated and binary logistic regression analysis showed that KI21 and CV4 (p=0.0001 and 0.04 respectively) provide the optimal set of features discriminating between the HMG patients and healthy controls. With a threshold of 33.2 °C for KI21, the optimal sensitivity (96%), specificity (52%), positive predictive value (PPV) 68% and negative predictive value (NPV) 93% were obtained (area under the curve 0.82, standard error 0.06, p=0.0001). Using the ‘leaving one out’ method, cross validation analysis again indicated that KI21 and CV4 were the only two skin points with additional independent discriminating value. The classification results of the test set and cross validation are nearly identical, with 77% and 75% correctly classified cases, respectively (Table III).
Discussion
The results show that compared with healthy controls, patients with HMG have a pathologically persistent high temperature on body surface points closely located to the breasts (KI21, LR14), particularly in women 39 years of age and older. The fact that the results are very similar in the set for learning and testing, and in the cross-validation analysis, indicates that the number of patients is adequate and the results robust and reliable. Nevertheless, validation in future independent patients remains important.
Interestingly, a recent article found that DITI identified breast cancer with 97% sensitivity, 44% specificity, and 82% negative predictive value (9). In the current study to detect breast hyperplasia, the values are the same or better (97%, 52% and 93%, respectively, see Table IV). This indicates that the accuracy of DITI in detecting different types of breast lesions is similar. The high NPV of DITI for detection of HMG makes the technique suitable to identify patients with breast complaints who need additional mammography to confirm the diagnosis of HMG.
The findings may reflect up-regulation of the growth activity of glandular cells of the breast whereas the abnormally persistent high temperature at skin point CV4 possibly indicates persistent overactivity of the ovaries with high oestrogen secretion. This may result in an oestrogen-dependent hyperplasia of the mammary glands, as continuous growth stimulation of the breast cells has been hypothesized as being a cause of breast cancer. Indeed, long-term hormonal replacement therapy is related to a significant increase in breast cancer (12). The question is whether persistent high activity at CV4 is correlated with persistently high blood oestrogen levels. Unfortunately, we do not have endocrinological data of our patients, but such a study would be scientifically interesting and important. Another possible explanation is that the persistent activity may be related to the patient being overweight, which has recently been classified by the World Health Organization and the International Agency for Research on Cancer as an important risk factor for breast cancer (13). The increase of overweight individuals in the Western world, and also in China, is alarming. It would therefore be important to study whether a relationship exists between the body mass index and skin point temperature. Preferably, the proliferation activity of the breast glandular cells should also be analyzed in such future studies, as this by definition is an important aetiological factor in breast carcinogenesis. Proliferation depends in part on the vascularization of the tumour, which explains that the degree of vascularisation in certain studies is prognostically important (14). Indeed, measurement of the vascularisation in breast cancer by automated digital image processing is strongly prognostic (15, 16), but requires biopsy material. Dynamic nuclear magnetic resonance imaging is possibly an alternative non-invasive method to measure vascularization, but unfortunately is labour intensive and expensive.
Conclusion
Female patients with HMG 39 years of age and older show a pathological persistence of high temperature on body surface points related to the breasts and ovaries. A similar but non significant trend was observed for those under 39 years old. Therefore, in women 39-55 years of age and possibly also younger women with breast complaints, infrared temperature measurement of specific skin points may be a non-invasive method to decide which patients need additional mammography to confirm the diagnosis of HMG.
Acknowledgements
This study was partially supported by the National Basic Research Program of China (2005CB523306 2009CB522901) and project 2008-08 from the Stichting Bevordering Diagnosatische Morfometrie, Middelburg, the Netherlands.
- Received June 8, 2009.
- Revision received September 14, 2009.
- Accepted September 22, 2009.
- Copyright© 2009 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved