The Gen-2 optical imager primarily consists of 3 components; the forked hand-held probe, the illumination system, and the detection system (see Figure 1). The hand-held probe of the Gen-2 optical imager is a forked probe design capable of conforming to tissue surface areas with minimal compression (see Figure 1(A)). The forked probe design allows for both reflectance as well as transillumination imaging approach, unlike all other hand-held imaging systems that are capable of only reflectance imaging [1–22]. There are 3 sources on each of the forked probe head, configured such that there is maximum area of illumination of the imaged region by the 4 × 5 cm² probe head [26]. The 3 sources and 96 detectors are spaced 0.5 cm apart on each probe head (as shown in Figure 1(C)).

The illumination system allows for both sequential as well as simultaneous illumination of the 3 + 3 sources via 6 laser diodes operated using a multi-channel laser diode controller [26]. Currently, the sources provide an optical output power of ∼5mW from each illumination point as a continuous-wave (CW) signal at 785 nm wavelength. Tissue phantoms are illuminated at single or multiple points via the forked probe heads and the detected optical signals is collected via the probe heads’ detection fibers and imaged using an intensified charge coupled device, ICCD camera (PI-MAX2, (Princeton Instruments, Newton, NJ, USA). Neutral density filters are used appropriately to collect the attenuated optical signal at 785 nm wavelength. The ICCD camera’s exposure time and gain settings are adjusted prior to each experimental study, in order to maximize the intensity of the detected signal. The illumination and detection system of the imager are controlled via custom-developed Labview software. The entire instrumentation is assembled onto a portable cart for ease of mobility during our clinical translational efforts (see Figure 1(B)). Extensive details of the instrumentation are provided elsewhere [26].

For the current studies, CW-based imaging was performed on tissue phantoms, where the exposure time of the ICCD camera was maintained at 0.2 s for all the studies. Ten repeated images were acquired at each location during experiments, and the average optical signal was used for further analysis.

Diffuse optical imaging studies were performed on cubical and slab phantoms filled with 3% Liposyn solution (Hospira, Inc. Lake Forest, IL, USA). Spherical targets of 0.45 cc (0.95 cm diameter) filled with 0.08% (by volume) India Ink (absorbing agent) (Chartpak, Leeds, MA.) and 3% Liposyn solution were employed. Reflectance (simultaneous bilateral) and transillumination experiments were performed under different experimental conditions of target depth (1–5 cm), and India Ink target: background (T:B) contrast ratio (e.g., 0.08%:0, and 0.08%:0.0008% or 1:0 and 100:1, respectively). The details of the experimental cases are given in Table 1. During simultaneous bilateral reflectance imaging studies, two 10 × 10 × 10 cm³ cubical phantoms were used, one with a target and the other without. All the six point sources on both the probe heads are simultaneously illuminated. During transillumination imaging studies, 5.5 × 5.5 × 10 cm³ slab phantoms were used. All the 3 point sources of one probe head are simultaneous illuminated and the point sources on the other probe head (located on the opposite surface of the imaging plane) are not (see Figure 2(B)). In both reflectance and transillumination imaging studies, the target depth is measured w.r.t. the phantom surface that is illuminated (via all simultaneous point sources) by the probe head. These studies will determine if transilluminated measurements detect deeper targets (from 2D surface images) over reflected measurements, upon using a unique hand-held probe capable of acquiring both these measurement types.

The optical properties of 3% Liposyn solution in terms of absorption (μ_a) and reduced scattering (μ_s') coefficient are 0.09 and 9.5 cm⁻¹, respectively, as measured using our single pixel homodyne optical imaging system. The typical optical properties of human breast tissues are μ_a = 0.04 ± 0.02 cm⁻¹, μ_s′ = 8 ± 4 cm⁻¹ [27]. Hence, the phantoms used in the current studies have similar optical properties of a human breast tissue. The absorption coefficient of 0.08% India Ink is ∼0.3 cm⁻¹ as estimated by researchers in the past (at 750 nm wavelength) [28]. The reduced scattering coefficient (μ_s′) of India Ink can vary between 4 to 20 cm⁻¹ as reported by various researchers [28,29], and was challenging to estimate in our studies.

The optical data acquired by the ICCD camera is in terms of a.b.u., which corresponds to the detected intensity signal from the various detection points on the imaged surface. These pixelated optical data acquired for the entire imaged region of the ICCD camera is then used to extract meaningful data corresponding to the detection fiber regions (i.e., for 96 × 2 detectors). The extracted optical data (or intensity data) corresponding to the 96 detected points of each probe head on the imaging surface is averaged across the 10 repeated measurements. The averaged intensity data corresponding to the 96 detection points of each probe head is further normalized to account for any instrument effects.

The effect of implementing the proposed data post-processing techniques (described in Section 2.3) on actual experimental data is shown in Figure 3 for a reflectance imaging case. The data is normalized at each stage of the data post-processing technique applied. For a 0.45 cc target located 1.5 cm deep from the tissue surface, the 2D surface map of the detected (unsubtracted) intensity does not differentiate the target from the background. In fact, there is a strong reflected intensity i.e., excitation leakage at the detectors located around the 3 point sources (see Figure 3(A)). Upon subtracting a background image (obtained from a phantom with no target), the effect of the strong excitation leakage is significantly reduced, as shown in the subtracted image in Figure 3(B). The target is detectable (at [2.53, 3.08] cm) in the subtracted image, but appears to be shifted from its true location, [2.5, 2] cm by 1.08 cm. Further removal of all detectors that immediately surround the source allowed for the detected target to locate closer ([2.49, 2.23] cm) to its true location (see Figure 3(C)) with a 0.23 cm distance-off. The 2D surface contour plots for Figure 3(A–C) are based on the Matlab’s linear interpolated contour plots. Upon applying the developed 5th degree polynomial fit to the post-processed 2D experimental (discrete) data points, the target localization is distinct with no artifacts (see Figure 3(D)). From these results, it is obvious that post-processing of the experimental optical data plays a significant role in improved target differentiation from the background with minimal artifacts. The proposed polynomial fitting approach is limited to tissue phantoms with a single target and not multiple targets. Currently, extensive simulation studies are carried out to develop appropriate techniques to better represent contour plots of tissue phantom surfaces with multiple targets as well.

Figure 4 shows the 2D surface contour plots (via polynomial fit) of the reflected (and subtracted) intensity signals obtained from a 0.45 cc target placed at various depths across the 10 × 10 × 10 cm³ phantom. Each experimental case was normalized since the instrumentation settings were different for each depth and T:B contrast ratio cases. From the plots, it is obvious that the target was detectable up to 2.5 cm deep, at both the chosen T:B (India Ink) contrast ratios. At depths of 3 cm and higher, the targets were not detectable in simulation studies and hence not imaged. The measurement errors were consistently ∼0.2 across any given target depth (up to 2.5 cm). The target was localized closer to its true location in all cases. The detected target location, target size, measurement errors and T:B detected contrast are given in Table 2.

Figure 5 shows the 2D surface contour plots (via polynomial fit) of the transilluminated (and subtracted) intensity signals obtained from a 0.45 cc target placed at various depths across the 10 × 10 × 10 cm³ phantom. From the plots, it is obvious that the target was detectable across the entire depth of the phantom (i.e., 5 cm deep), at both the chosen optical contrast ratios. The target was localized closer to its true location in all cases (≤0.5 cm in most cases), although there were obvious artifacts without the data post-processing techniques. The measurement errors were two orders of magnitude smaller for 1:0 contrast cases in comparison to the 100:1 contrast cases. The detected target location, target size, measurement errors and T:B detected contrast are given in Table 3. The measurement error was significantly smaller for 1:0 contrast cases of transillumination studies compared to reflectance imaging studies, and hence reported up to 4 significant digits.

The progression of images exhibit that as the target is closer to the detection surface (i.e., further away from the illumination surface), where the optical detection fibers are placed a reduction in artifacts is visible, as well as an improvement in location estimation. Transillumination measurements in general provide greater depth information of the tissues over reflectance measurements. Typically most studies have focused on using single point source illumination to demonstrate the effectiveness of transillumination measurements. The current studies are different in terms of the source illumination configuration. Unlike the typically used sequential point illumination employed by other researchers, the Gen-2 hand-held imager can perform simultaneous multiple point illumination towards enhanced imaging rates. Thus, even simultaneous illumination, which can have null (intensity) points from interferences, can still detect deep targets based via transillumination imaging over reflectance, independent of the illumination configuration.

Preliminary in-vivo breast imaging study was performed on a normal human subject with a 0.45 cc absorption target (India Ink) placed in the inter-mammography fold of the left breast (at 6 ’0 clock position). CW-based simultaneous bilateral reflectance imaging was performed on both the breasts at symmetric locations. While the unsubtracted images of left and right breasts did not detect the tumors, the subtracted image (left-right) distinctly differentiates the target (even prior to applying other data post-processing techniques described in Section 2.3), as shown in Figure 6. This preliminary study establishes the capability of the Gen-2 imager to detect targets in-vivo human breast tissues via simultaneous bilateral imaging. Extensive clinical studies are currently carried out to determine the performance of the Gen-2 imager towards both reflectance and transillumination imaging on normal and breast cancer subjects.