Solid tumors are generally associated with reduced oxygen saturation (due to hypoxia) [14
] and with increased hemoglobin concentration (due to unregulated angiogenesis) [12
]. NIR diffuse optical imaging and spectroscopy has been used for noninvasive characterization of tumor-induced tissue parameter changes [10
]. Previous researchers have demonstrated that breast tumors (both breast cancer and fibroadenoma) typically show relatively lower oxygen saturation and higher hemoglobin concentration in comparison with those of reference breast tissues [9
]. The pilot clinical trial reported in the present manuscript demonstrated consistent results in comparison with findings reported by previous researchers.
Despite the observed differences between tumor and normal tissue in terms of hemoglobin concentration and oxygen saturation, clinical application of NIR technology into the arena of breast cancer detection has not yet been realized. One of the major obstacles is the difficulty to derive generalized criteria for characterizing tissue differences that would ultimately allow one to distinguish 'benign' findings versus 'malignant' tumors within the breast. The difficulty comes from the significant tissue heterogeneities and interpatient variations that may exceed the characteristic difference between benign and malignant tumors [22
]. To overcome this difficulty, some investigators studied tissue dynamic characteristics in response to physiologic, chemical, and mechanical stimuli [27
]. Other investigators derived relative diagnostic criteria by calculating differential oxygen and hemoglobin parameters between the tumor and the reference tissue in the contralateral, cancer-free breast [20
To minimize the influence of tissue heterogeneity and the interpatient variation, we developed a dynamic breast imaging schema. By defining relative tissue parameters and studying pressure-induced changes in these parameters, we hope to derive more effective detection criteria that are less sensitive to tissue heterogeneities and interpatient variations. Although some tissue parameters defined in this paper are not yet fully utilized due to the instrumentation limitations, the dynamic imaging concept discussed in the present paper paves the way for a low-cost, portable, reproducible imaging method for breast cancer detection.
The results of the current prospective pilot clinical trial confirmed the clinical potential of the currently evaluated dynamic NIR imaging schema. Statistically significant differences were observed between 'tumor' and its surrounding normal reference tissue for averaged [StO2] and [Hbt] levels. Generally speaking, 'tumor' shows higher [Hbt] (probably due to angiogenesis) and lower [StO2] (probably due to hypoxia) than the normal surrounding breast parenchyma. Likewise, among all 'tumors', differences between 'benign' tumors and 'malignant' tumors were demonstrated in terms of differential hemoglobin contrast, indicating that 'benign' lesions display higher heterogeneity in terms of hemoglobin concentration than do 'malignant' lesions. No difference was observed, however, between 'benign' lesions and 'malignant' lesions in terms of other tissue parameters, such as averaged [StO2] and [StO2] contrast.
From the results of the current prospective pilot clinical trial, it is evident that several technical limitations exist. Such limitations of the currently evaluated dynamic NIR imaging schema will need to ultimately be overcome in order to create a more clinically relevant imaging schema.
The first technical issue of the current prospective pilot clinical trial relates to the use of ramped compression (that is, gradual increase or decrease of the compression load). Such a ramped compression profile may not be the most ideal mechanism for providing the dynamic stimulus. During ramped compression, hemodynamic changes are intrinsically coupled with tissue viscoelastic deformation that is induced by the action of ramped compression. This makes the resultant quantitative analysis of such events quite difficult. For this particular reason, and as previously mentioned in Materials and methods, our statistical analyses only considered two datapoints of each dynamic loading cycle (that is, the baseline measurement and the compression peak measurement). Additionally, the 5-second ramped compression load used in this protocol was not sufficient to result in significant tissue oxygenation changes, as would be observed during tissue ischemia secondary to vascular occlusion. In our protocol, therefore, the [StO2] fluctuations measured during compression were not the result of tissue ischemia, but were instead the result of additional tissue heterogeneity induced by relative changes in the position of a given suspicious breast lesion and its adjacent normal reference tissue.
We propose two potential solutions to resolve these issues with ramped compression. First, we can separate tissue deformation from tissue physiologic changes by simultaneous structural and functional imaging using an integrated NIR and ultrasound imaging system. Second, we can replace the ramped compression profile with a stepped compression profile (that is, sudden increase or decrease of the compression load). Such a stepped compression profile would uncouple the tissue mechanical reaction (that is, transient tissue deformation) from the physiologic reaction (that is, tissue ischemia). As the result of using an integrated imaging system and a stepped compression profile, we expect to see characteristic differences in other tissue parameters that will ultimately help to better differentiate 'malignant' from 'benign' lesions.
The second technical issue of the current prospective pilot clinical trial relates to the instrumentation of the NIR imaging system utilized. In general, NIR imaging systems fall into three technical platforms, based on differences in the time dependence of the excitation source intensity and the detection mechanism [10
]. These technical platforms include continuous-wave devices, time-domain devices, and frequency-domain devices [10
In the current prospective pilot clinical trial, we utilized a continuous wave NIR system (that is, the P-Scan imager). The intrinsic limitation of such a continuous-wave system is that the scattering coefficient cannot be explicitly resolved, and therefore differential path length factors or reduced scattering coefficients have to be assumed [43
]. In our current prospective pilot clinical trial, constant reduced scattering coefficients were used (
= 4/cm). The mismatch between the assumed scattering coefficient and the actual scattering coefficient can introduce measurement bias in [St
] and [Hbt]. In our current prospective pilot clinical trial, a relatively higher [St
] and a relatively lower [Hbt] were observed as compared with [St
] and [Hbt] measurements recorded by other NIR imaging systems, such as a frequency domain device [40
]. Despite the above measurement bias, the use of a continuous-wave device may still preserve its clinical applicability secondary to its simplistic design and its low cost. Furthermore, in our continuous-wave dynamic NIR imaging schema, we derived only relative, differential tissue parameters based on [St
] and [Hbt] measurements, regardless of their resultant bias in absolute values. In this regard, the influence of the scattering coefficient mismatch would be further reduced.
A final technical issue regarding the instrumentation of the NIR imaging system utilized in the current prospective pilot clinical trial is related to the limited image resolution. This limited image resolution is the result of an insufficient number of source-detector pixel positions for optical measurements and the poor coregistration between sequential NIR and ultrasound imaging. The intrinsic problem of our limited imaging resolution is related to high scattering and exponential attenuation of light in biological tissues. To improve upon the limited imaging resolution of our current NIR imaging system instrumentation, we plan to implement a step-motorized scanning mechanism to increase the number of optical measurement positions. Additionally, we plan to utilize an integrated NIR and ultrasound imaging system for simultaneous image acquisition during each dynamic loading cycle in order to monitor tumor deformation and changes in tumor depth. We believe the implementation of such an integrated system will allow us to more precisely characterize specific differences between benign and malignant breast lesions.