Prospective comparative study assessing role of ultrasound versus thermography in breast cancer detection

Introduction

Diagnosis of breast cancer during the early stages is directly associated with convenient treatment, which would also decrease mortality and morbidity of it [1,2]. Mammography is considered the gold standard screening method for breast cancer detection; it can show tumors long before they appear clinically, or are big enough to be felt by patients [3]. However, mammography has some limitations as well, especially in patients with dense breasts and in young patients. Ultrasound and clinical breast exams are used as adjunctive tools in conjunction with mammography for screening within a diagnostic setting [4]. Some other modalities which are available for breast disease diagnosis are based on light (optical), sound (ultrasound), heat (thermogram), magnetism, microwave, nuclear modes, electrical impedance, and computer modeling (inverse simulation), or a fusion of different methods.

Thermal imaging or infrared imaging (IR) is a non-invasive, non-contact system of recording body temperature by measuring infrared radiation emitted on the body’s surface [4]. The use of IR is based on the principle that the metabolic activity and vascular circulation in precancerous tissue and surrounding areas are often higher than normal breast tissue [5]. From the late 1950s when Lawson first reported on thermography, numerous projects were accomplished with it in breast cancer detection [6]. Although it has not been accepted as a screening method yet, it was awarded the US Food and Drug Administration’s (FDA) approval as a viable tool in breast cancer diagnosis. Previous studies [7,8] accompanying with some newer studies [9–11] used thermography as a screening tool and compared thermography with mammography. In recent years, the infrared camera in addition to image processing methods improved, which may lead to better detection of breast lesions. We planned to compare IR results with ultrasound and to evaluate the usefulness of IR in breast imaging regarding sensitivity, specificity and accuracy.

Material and methods

There were 78 women for whom breast ultrasonic imaging was requested by the clinician; they were randomly selected for thermography on the same day from June to August 2015. This prospective study was explained for them orally and via a written pamphlet. Written informed consent was taken for all included patients, with any identity-revealing information preserved. This study was conducted according to the principles of the Declaration of Helsinki and the participating researchers declare no conflicts of interest. Exclusion criteria consisted of patients with known cancer, previous cancer treatment, and patients undergoing chemotherapy or radiotherapy. The technician who did the thermal exams was a certified medical thermographer and interpretations were blind to the ultrasound results.

These patients had different sonographic results, including normal, benign, or suspicious. Before starting the thermography test, patients were asked to undress from the waist up to balance their body temperature with the environment. The patient was positioned in front of the thermal camera, with 5 captured images from 5 angles (frontal, 45 left/right, 90 left/right). Infrared camera parameters are given in Table 1. In order to compare images, we tried to control the temperature of the room and maintain 24–26^∘ C . The ambient temperature was controlled between 25–27^∘ C so as to have no effect on thermal conditions.

Potential sources of additional heat generated by windows with the sun shining in, or additional heat generated by computer equipment and other devices were eliminated from the imaging room to reduce thermal artifact. The distance was fixed between the camera and patients in terms of her anatomy.

All patients who volunteered to carry out this test were examined by sonography (GE Voluson E6 Expert machine, Fairfield, CT, USA) as well which was done by a single breast specialized radiologist. Sonography and thermography of each patient were done on the same day.

There is some variation in a thermography report, so we labeled and evaluated results according to the classification in Table 2 [12]. Every image has a ‘TH’ number, which shows the patient’s status; abnormal signs would be:

Asymmetric and hyperthermic vascular patterns.

(Figs 1 and 2).

Results

The average age of 78 patients in the study was 41.0 ± 10.4 years, from 18 to 67 years. Among subjects, 14 patients had the sonographic breast imaging and reporting data system (BIRADS) of 4 or 5, which were considered as suspicious cases and then underwent core needle biopsy under ultrasonic guidance. BIRADS I was considered normal, while BIRADS II and III were regarded as benign cases and followed for one year until the summer of 2016. For thermography, TH1 and TH2 were labeled as normal/benign cases, while TH3, TH4, and TH5 were suspicious. Since evaluations were performed for both patient breasts separately, each one was considered a distinct case for analytical purposes. Table 3 presents the descriptive statistics of the evaluated variables in the sample population.

Of the 14 breasts whose lesions were biopsied, 7 cases were found to have malignant pathology. As mentioned earlier, all other cases were followed for at least one year and none showed any increase in the size of the lesions or progressed to malignancy, so were classified as normal/benign to be included in the analyses.

Tables 4 and 5 present the correlation between BIRADS and TH categories according to the pathology results of the patients’ lesions, with the final classification of cases in Normal/Benign and Malignant lesions with diagnostic characteristics of the two methods. Based on these findings, ultrasound BIRADS classification has a sensitivity of 100%, specificity of 95.3%, PPV of 50.0%, NPV of 100%, and an overall accuracy of 95.5% in diagnosing malignant breast lesions. As for the TH classification, the sensitivity was calculated to be 85.7%, specificity 78.5%, PPV 15.8%, NPV 99.1%, and overall accuracy was 78.8%.

Receiver operating characteristics (ROC) curve analysis was also performed to calculate the area under the curve (AUC) with each method. Accordingly, the AUC was calculated to be 0.821 (p = 0.004) for the TH classification and 0.977 (p < 0.001) for the sonographic BIRADS classification (Fig. 3).

Discussion

Current breast screening imaging methods have limitations [13]. One of the best complementary methods is sonography, which is safe, accessible, and cost-effective, with some limitations such as being operator-dependent and machine-type dependent. In addition to taking images, a probe must be in contact with the patients’ body, which is often uncomfortable for patients. Breast sonography is also time consuming and radiologist-dependent – resulting in patients having to wait for the appointment [14]. Other complementary breast imaging modalities would be valuable.

Thermography does not present information of the structural changes of the breast, but does provide functional information on thermal changes of tissue. It is believed that increased angiogenesis and metabolic rate in tumors cause increased temperature of the area and functional changes [15], which may occur before morphological features change.

Early clinical trials in the 1960s and 1970s indicated thermography’s ability to detect breast cancer, but in 1976, Moskowitz et al. [16] announced that it did not show good accuracy of early breast cancer and might not be suitable for breast cancer screening; this led to a decrease in medical centers’ interest in this technique for diagnosis.

Regarding a review by Deborah A. Kennedy none of the imaging modalities is perfect in breast cancer screening; however, they proposed the combination of modalities for increasing both sensitivity and specificity. They also suggested that with technological advances of the thermography, additional research should be done to evaluate the potential of this technology as an effective adjunctive tool for breast cancer screening [17].

In one multi-centric clinical trial in 2003 with 769 patients and 875 biopsies of suspicious mammographic lesions [18], the sensitivity of thermography was 97%, with a specificity of 14%, and a PPV of 25%.

Aora et al. [15] in 2008 highlighted the diagnoses of 58 out of 60 malignancies by thermography, with 97% sensitivity, 44% specificity, and 82% NPV.

In 2010, Wishart et al. [19] reported that thermography was effective in women under 50 with high sensitivity (78%) and specificity (75%); this is not true in older women, likely because of reduced breast vascularity in this age group.

Yao X announced that the infrared test had greater sensitivity and specificity in early breast cancer in 2014 [20].

An important issue is that previous authors indicated that thermography could predict breast cancer 8–10 years before mammography (20), while later studies concluded that infrared test abnormalities in breast tissue could be independent breast cancer risk factors [21].

In a recent study by Omranipour et al. with 132 women who were candidates for biopsy, thermography, and mammography – each done before biopsy with a sensitivity of 47.1% for thermography and 80.5% for mammography; however, they included TH 4–5 as suspicious, while adding TH3 sensitivity reached 81.6% [10].

In a study on the efficacy of thermography, all false negative cases were due to microcalcifications, showing that infrared imaging is not as good as mammography in finding cancers with microcalcifications [18]. Our single false negative case in this study was a 9 mm mass with invasive ductal carcinoma pathology, which was given TH 2.

In a Prospective study which was done at Kasturba Hospital, Manipal, India, 65 known breast cancer were evaluated with thermography. Malignancy was accurately detected in 60 patients (92.31%), while mammography missed three patients with breast cancer which were diagnosed with thermography [22].

A newer study of Colleen H.Neal published in 2017 included 38 patients with abnormal thermography who were referred for evaluation by sonography or mammography. They reported 100% negative predictive value for thermography and they did not find any cancer among asymptomatic women. All the patients diagnosed with cancer who were 5% of the cases, had co-existing suspicious clinical finding [23].

In contrast to most other studies, we compared the results of thermography with sonography – and accordingly, all diagnostic value indices of thermography were lower than sonography – and thus not as good in diagnosis of breast cancer; moreover thermography cannot localize the lesion. In clinical settings whenever possible, in addition to mammography, we need other modalities for breast evaluation, sonography would be a better choice, but in special conditions, such as limitations of a sonography machine or an expert operator, thermography would be an acceptable choice with a sensitivity of 85.7%, specificity 78.5%, and accuracy 78.8%, as well as being taken without any skin contact, which is more desirable for patients and does not need any ionized rays to make images: it is also a noninvasive and safe method. Our study has a few limitations such as the sample size was not large and not all patients had a biopsy. We followed-up with patients with sonographic BIRADS of 1, 2, and 3.

Conclusion

Although thermography is a non-invasive, safe, and easy-accepted method by patients and despite recent technical improvement of it, its diagnostic abilities including sensitivity, specificity, accuracy all are still less than those of sonography, so we suggest using it as an adjunct in breast cancer diagnosis in selective situations.