2.1.

Subjects

This study included 48 patients who were diagnosed with breast cancer from September 2017 to November 2018. Patients who received neoadjuvant chemotherapy ($n=8$) or hormone therapy ($n=3$) were excluded, leaving 37 patients. All were Asian women with an age range of 25–88 years, median age of 65 years, and median body mass index of 22.8 (range: 16.4–32.4). Nine patients were premenopausal and 28 were postmenopausal. In this study, only data for normal-side breasts were evaluated. The study protocol was approved by the Ethical Review Committee of the Hamamatsu University School of Medicine, and written informed consent was obtained from all patients.

2.2.

TD-DOS and US Measurements

The optical properties of breast tissue were measured using a six-wavelength TD-DOS system (TRS-21-6W, Hamamatsu Photonics K.K., Hamamatsu, Japan). The system was developed based on a three-wavelength TD-DOS system (TRS-20SH, Hamamatsu Photonics K.K.) used in our previous studies.^9–12

The TRS-21-6W system uses the time-correlated single-photon counting (TCSPC) method to acquire the time response profile of pulsed laser light penetrating the target tissue. The system includes two light source units, each containing three laser diodes (762, 802, and 838 nm for the first unit; 908, 936, and 976 nm for the second unit) that emit a light pulse with a full width at half maximum of ~100 to ~200 ps at a repetition rate of 5 MHz (custom-designed, Hamamatsu Photonics K.K.). The photodetector units consist of two photomultiplier tubes (GaAs and InGaAs, Hamamatsu Photonics K.K.), constant fraction discriminators, time-to-amplitude converters, analog-to-digital converters, and histogram memories. The six pulsed lasers are integrated into one source fiber with a diameter of 1 mm and a numerical aperture of 0.29 (Sumita Optical Glass, Inc., Saitama, Japan). The irradiated light from the source fiber is collected by a detection fiber bundle with a diameter of 3 mm and a numerical aperture of 0.29 (Sumita Optical Glass, Inc.). The specifications and analysis method of the improved system are described in detail in Ref. 8. The system uses the TCSPC method to acquire the temporal response profiles of tissue from optical pulse inputs, which enable quantitative analyses of light absorption and scattering in tissue based on photon diffusion theory.²⁶

The water and lipid content and the concentrations of ${\mathrm{O}}_2\mathrm{Hb}$ and HHb are calculated from the absorption coefficients for the six wavelengths. The concentration of tHb is the sum of the ${\mathrm{O}}_2\mathrm{Hb}$ and HHb concentrations. As described in our previous study,⁸ the water and lipid content are the volume fractions of water and lipids measured by the TRS-21-6W system. Notably, these are relative values compared with those of pure solutions and not the actual composition ratios of the tissue.²⁷ The measurement accuracy of the absorption and scattering coefficients and the adequacy of the water and lipid content were evaluated in Ref. 8.

Breast measurements were carried out in the reflectance geometry with a source–detector separation of 3 cm. The measurements were performed in supine position and performed simultaneously with the TRS-21-6W system and an US system (EUB-7500; Hitachi Medical Corporation, Tokyo, Japan) using a custom spectroscopic probe in which the resulting optical path is set orthogonal to the US image. The probe was attached to a linear probe (EUP-L65, Hitachi Medical Corporation), as shown in Fig. 1(a), and lightly pressed on the breast to measure the absorption and reduced scattering coefficients at six wavelengths while simultaneously obtaining US images. The light from the source fiber is scattered and absorbed in the medium before it reaches the detector. The photon path between the source and the detector fiber is considered to be banana-shaped [Fig. 1(b)].²⁸

Fig. 1

(a) Optical probe attached to an ultrasound (US) probe. (b) Schematic illustration of the diffusion of light propagating in the breast and positioning of the optical fibers and US probe.

2.3.

Mammography

MMG results obtained at our current or former hospital on the nearest date to the optical measurement were categorized into four types based on BI-RADS 5th ed.: (a) the breasts are almost entirely fatty; (b) there are scattered areas of fibroglandular density; (c) the breasts are heterogeneously dense, which may obscure small masses; and (d) the breasts are extremely dense.

2.4.

Measurement and Evaluation Procedures

Measurements were performed at several locations on the normal-side breast. Our evaluation was first conducted on data that had a two-layer structure of lipids and the chest wall but no mammary gland in the subclavian area or in breast area away from a nipple, as confirmed by simultaneous US images. Data were obtained for 38 locations from 25 patients, with a maximum of four locations for each patient.

To evaluate normal breast data, data for 133 locations were included. Measurements were performed on normal breast tissue at a maximum of six locations for each patient. We excluded data measured directly on the nipple.

Figure 2 shows US images of normal breast tissue and of tissue in the subclavian area with a two-layer structure of lipids and the chest wall. The skin-to-mammary gland distance and the skin-to-chest wall distance were measured at the center of US images. To measure the skin-to-chest-wall distance, the chest wall was defined as the surface of the major pectoral muscle or anterior serratus muscle.

Fig. 2

(a) US image of normal breast tissue and (b) the subclavian area with a two-layer structure of lipids and the chest wall. The images show the skin (S), lipid layer (L), mammary gland (M), and chest wall (C).

A previous study⁹ showed that the tHb value increased as the skin-to-chest wall distance decreased, indicating that the chest wall has a strong influence on tHb measurements. To study the influence of the chest wall on water and lipid content, we examined the relationship between water and lipid content and the skin-to-chest wall distance for the two-layer structure of a lipid layer and chest wall using the sigmoid curve-fitting method. We also compared the data with tHb measurements for the same structure.

The following regression equation was selected for the sigmoid curve based on previous studies:^29,30

Next, we plotted the tHb concentration, water content, and lipid content of normal breast tissue against the skin-to-chest wall distance and skin-to-mammary gland distance. The tHb concentration result was used as a reference. We also applied the sigmoid curve-fitting method to the tHb concentration, water content, and lipid content.

Two expert radiologists assigned BI-RADS MMG density categories for all 37 patients. When images were classified into different categories by each radiologist, the final classification was determined after discussion.

We compared the tHb concentration, water content, and lipid content in normal breast tissue for each MMG category. To assess the potential to distinguish dense from non-dense breasts based on optical data, the tHb concentration, water content, and lipid content for all 133 data points were shown separately with skin-to-chest wall distance in dense and non-dense breasts. Boxplots were used to display the water and lipid content for locations at which the skin-to-chest wall distance was large enough to avoid the effect of the chest wall and the skin-to-mammary gland distance was small enough to represent the mammary gland. The optical data were then analyzed using receiver-operating-characteristic (ROC) curve analysis to determine their potential to distinguish dense from non-dense breasts.

2.5.

Statistical Analysis

Statistical analyses were performed using Microsoft Excel 2016 (Microsoft Corporation, Redmond, Washington), StatFlex version 6.0 (Artech Co., Ltd, Osaka, Japan), and SigmaPlot version 14.0 (Systat Software, Inc., San Jose, California). A sigmoid curve (1) was applied for curve fitting of the tHb, water content, and lipid content in the two-layer structure and in normal breast tissue.

The tHb concentration and water and lipid content in normal breast tissue in each MMG category were compared using nonparametric Mann–Whitney U tests with Bonferroni correction. The skin-to-mammary gland distances of dense and non-dense breasts were compared using nonparametric Mann–Whitney U tests. The water and lipid content of dense and non-dense breasts were compared using nonparametric Mann–Whitney U tests. Differences with $p<0.05$ were considered significant.

To identify an optimal threshold for discriminating dense from non-dense breasts, ROC curve analysis of optical data was performed by incrementally increasing the cutoff values. The area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were analyzed.

3.1.

tHb Concentration, Water Content, and Lipid Content in the Two-Layer Structure of Lipids and the Chest Wall

The tHb concentration, water content, and lipid content in the two-layer structure comprising a lipid layer and the chest wall are shown in Fig. 3. The tHb concentration increased as the skin-to-chest wall distance decreased from $\sim 20\mathrm{mm}$. The water content increased and the lipid content decreased as the distance decreased from $\sim 15\mathrm{mm}$. The value of $y_0$ in the sigmoid curve (1) was 11.1 for tHb concentration, 7.9 for water content, and the value of $y_0+a$ was 69 for lipid.

Fig. 3

tHb concentration, water content, and lipid content in the two-layer structure versus skin-to-chest wall distance. Experimental data ($n=38$) and fitted curves are shown for tHb concentration ($R=0.978$), water content ($R=0.965$), and lipid content ($R=0.931$).

3.2.

tHb Concentration, Water Content, and Lipid Content in Normal Breast Tissue

The tHb concentration, water content, and lipid content in normal breast tissue for all 133 data in 37 patients are shown in Fig. 4. In the figure, the relationship between each parameter and the skin-to-chest wall distance is displayed. Both tHb concentration and water content increased as the skin-to-chest wall distance decreased, whereas the lipid content decreased as the skin-to-chest wall distance decreased. The multiple correlation coefficients ($R$ values) for the fitted curves were 0.938 for tHb, 0.774 for water content, and 0.838 for lipid content. The coefficients of variation were calculated to compare the dispersion of tHb concentration and water content in normal breasts for data with a skin-to-chest wall distance of 20 mm or more, assuming that the chest wall has little effect on each parameter. The coefficients of variation were 24.3% for tHb concentration and 41.4% for the water content.

Fig. 4

tHb concentration, water content, and lipid content of normal breast tissue versus skin-to-chest wall distance. Experimental data ($n=133$) and fitted curves are shown for (a) tHb concentration ($R=0.938$), (b) water content ($R=0.774$), and (c) lipid content ($R=0.838$).

Figure 5 shows the relationship between tHb concentration, water content, and lipid content and the skin-to-mammary gland distance in normal breast tissue for all 133 data from 37 patients. Both the tHb and water content increased as the skin-to-mammary gland distance decreased. Lipid content decreased as the skin-to-mammary gland distance decreased. For skin-to-mammary gland distances $<10\mathrm{mm}$, the dispersion of tHb values was large, whereas the values of water and lipid content lie near the fitted curves, showing a distribution similar to that of the two-layer structure of a lipid layer and chest wall. The values of $R$ for the fitted curves were 0.795 for tHb, 0.949 for water content, and 0.923 for lipid content.

Fig. 5

tHb concentration, water content, and lipid content of normal breast tissue versus skin-to-mammary gland distance. Experimental data ($n=133$) and fitted curves are shown for (a) tHb concentration ($R=0.795$), (b) water content ($R=0.949$), and (c) lipid content ($R=0.923$).

3.3.

tHb Concentration, Water Content, and Lipid Content in Normal Breast Tissue Based on Breast Density

The breast density in MMG was classified as “almost entirely fat” (category a) in three patients, having “scattered fibroglandular densities” (category b) in 20 patients, “heterogeneously dense” (category c) in 11 patients, and “extremely dense” (category d) in three patients. Breast tissue in categories c and d was defined as dense, whereas that in the other categories was defined as non-dense.

Boxplots of the tHb concentration, water content, and lipid content in normal breast tissue for each category are shown in Fig. 6. There were 14 data in category a, 69 in category b, 36 in category c, and 14 in category d. The tHb concentration in category a was significantly lower than that in category d ($p=0.017$). Although there was no significant difference between the water content in categories c and d, significant differences in water content were found for all other pairwise group comparisons ($p<0.01$). The lipid content in category b was significantly higher than that in category c ($p<0.01$) and category d ($p=0.027$). Figure 7 shows the relationship between the skin-to-chest wall distance and tHb concentration, water content, and lipid content of normal breast tissue in non-dense ($n=83$) and dense breasts ($n=50$). As the skin-to-chest wall distance decreased, the tHb concentration and water content increased for both non-dense and dense breasts. There was a tendency for the water content to be higher in dense breasts than in non-dense breasts for a given skin-to-chest wall distance. The lipid content decreased as the skin-to-chest wall distance decreased. Moreover, there was a tendency for the lipid content to be lower in dense breasts than in non-dense breasts for a given skin-to-chest wall distance. The skin-to-mammary gland distance was shorter in dense breasts than in non-dense breasts ($p<0.001$), with a median distance of 10 mm [interquartile range (IQR): 7–14, $n=50$] in dense breasts and 14 mm (IQR: 11–18, $n=83$) in non-dense breasts.

Fig. 6

Boxplots of tHb concentration, water content, and lipid content for each mammography category. The median value of tHb was $14.9\mu \mathrm{M}$ in category a (almost entirely fat), $16.8\mu \mathrm{M}$ in category b (scattered fibroglandular densities), $18.3\mu \mathrm{M}$ in category c (heterogeneously dense), and $23.3\mu \mathrm{M}$ in category d (extremely dense). The tHb concentration in category a was significantly lower than that in d ($p=0.017$). The median water content was 7.9% in category a, 11.9% in category b, 21.8% in category c, and 19.5% in category d. Although there was no significant difference between the water content in categories c and d, significant differences were found for all other pairwise group comparisons ($p<0.01$). The median values of lipid content were 70.9% in category a, 72.9% in category b, 63.8% in category c, and 65.3% in category d. The lipid content in category b was significantly higher than that in category c ($p<0.01$) and category d ($p=0.027$).

Fig. 7

(a) tHb concentration, (b) water content, and (c) lipid content versus skin-to-chest wall distance in normal breast tissue for non-dense ($n=83$) and dense ($n=50$) breasts.

Fig. 8

Boxplots of (a) water content and (b) lipid content of non-dense ($n=52$) and dense ($n=31$) breasts with a skin-to-chest wall distance of 15 mm or more and a skin-to-mammary gland distance of 15 mm or less, assuming that these data reflect the content of breast tissue with a mammary gland and are minimally influenced by the chest wall. The water content of normal breast tissue was higher in dense breasts than in non-dense breasts ($p<0.001$). The lipid content of normal breast tissue was lower in dense breasts than in non-dense breasts ($p<0.001$).

According to Fig. 3, the depth sensitivity for water and lipid content was $\sim 15\mathrm{mm}$. Figure 8 shows boxplots of water and lipid content for non-dense ($n=52$) and dense ($n=31$) breasts with a skin-to-chest wall distance of 15 mm or more and a skin-to-mammary gland distance of 15 mm or less, assuming that these data reflect the content of breast tissue with a mammary gland and are minimally influenced by the chest wall. The water content in normal breast tissue was higher in dense than in non-dense breasts ($p<0.001$), with a median content of 21.0% (IQR: 16.2–27.1, $n=31$) in dense breasts and 11.9% (IQR: 10.5–13.4, $n=52$) in non-dense breasts. The lipid content in normal breast tissue was lower in dense breasts than in non-dense breasts ($p<0.001$), with a median content of 63.5% (IQR: 58.2–67.6, $n=31$) in dense breasts and 72.9% (IQR: 70.1–74.3, $n=52$) in non-dense breasts.

3.5.

ROC Analysis for Distinguishing Dense from Non-Dense Breasts

ROC curve analysis was used to evaluate the possibility of distinguishing dense from non-dense breasts based on optical parameters. Figure 9 shows the ROC curve for normal breast tissue for 83 data (31 dense and 52 non-dense breasts) with a skin-to-chest wall distance of 15 mm or more and a skin-to-mammary gland distance of 15 mm or less, assuming that these data reflect the content of breast tissue with a mammary gland and are minimally influenced by the chest wall. The AUC was 0.91 for both water content and lipid content. With an optimal cutoff of 14.0% for water content, the results were: diagnostic accuracy, 83.1%; sensitivity, 83.9%; specificity, 82.7%; PPV, 74.3%; and NPV, 89.6%. With an optimal cutoff of 69.1% for lipid content, the results were: diagnostic accuracy, 80.7 %; sensitivity, 80.6%; specificity, 80.8%; PPV, 71.4%; and NPV, 87.5%.

Fig. 9

ROC curves for 83 data corresponding to a skin-to-chest wall distance of 15 mm or more and a skin-to-mammary gland distance of 15 mm or less. The AUC was 0.91 for water content and 0.91 for lipid content.