1.

Introduction

Reconstructive surgery aims to restore anatomical defects with autologous tissues called “flaps.” A flap includes the harvested tissues (skin, muscle, bone, etc., alone or combined) and their blood-supplying vessels (arteries and veins that constitute the pedicle) which are essential for the flap survival. In case of major reconstruction, the flap is completely separated from the donor site and the vessels are anastomosed (reconnected) to the vessels in the recipient site. In this case, surgeons execute the so-called “free-flap.” Patency of the arterial and venous anastomosis is mandatory for free-flap survival. Human flap thickness is around 1.5 to 3 cm (in some cases, e.g., deep inferior epigastric flap, they can be thicker than 3 cm) and vessels are buried deep below the skin, muscle, and fat layers.

The major complication of this type of surgery is thrombosis,¹ i.e., vessel obstruction by a clot (usually a venous one) and it can lead to flap failure (necrosis). Even if this complication is not frequent (between 3% to 7% depending on the surgeon’s experience and the type of flap²) the functional and esthetic consequences are dramatic if not treated rapidly. Salvage of the flap depends on the rapidity of restoration of the vessels permeability: 60% to 73% of complicated flaps could be salvaged if the thrombosis is treated before 6 h of evolution.³

Postoperative monitoring of the flap is commonly carried out by a clinical examination (inspection and palpation), a procedure that is highly operator-dependent. Moreover, this examination is even impossible when deep buried flaps are involved, e.g., in the case of mastectomy where an extra layer of biological tissue is used over the flap to reconstruct the volume of the breast or in the case of pharyngeal reconstruction.^4,5 Importantly, changes in the appearance of the flap due to microvascular complications are delayed as compared to the instrumental detection and they will become visible only 1 or 2 h later.

Monitoring buried free flaps is difficult because the standard methods of clinical examination and surface Doppler monitoring are not possible. Buried free flaps, which represent less than 5% of all free flaps, are often monitored using an implantable 20-MHz ultrasonic Doppler probe. Implantable Doppler probe monitoring⁶ has a 100% sensitivity rate in detecting loss of flap perfusion, making it a good screening test for free-flap viability. However, it suffers from a high false-positive rate of 88%, which results in a high proportion of subsequent negative surgical explorations. In addition, the use of implantable Doppler implies the need to perforate the flap which exposes it to the risk of infection and bleeding or vessel damage. Moreover, the quality and effectiveness of the monitoring depend on the stability of the probe position into the flap. Another solution consists of the externalization of a flap component, but it is often unaesthetic and it requires a supplementary surgery to remove the externalized part at the end of the monitoring period.

Buried flaps could also be monitored by microdialysis,⁷ laser-Doppler flowmetry,⁸ or tissue oxygen monitoring. These methods are very sensitive but invasive. In this case also, there is a risk due to the need to implant a probe perforating the flap. In addition, similarly to implantable Doppler, the quality of the monitoring is very much dependent on the stability of the probe into the flap, which is sometimes difficult because of breath movements or in the neck because of deglutition. Moreover, most of these systems are expensive: in addition to the cost of the machine it is necessary to add the cost of the consumables that varies from 400 to 600 € per patient. Finally, monitoring is provided throughout the period during which the risk of vascular occlusion is the highest, i.e., four to six days following the operation, as frequently as possible—ideally continuously—but in practice every 4 to 6 h. Since the chances of recovery (reperfusion of the flap) after the occlusion are directly proportional to the rapidity of diagnosis and subsequent surgical remedial treatment, an ideal continuous monitoring of flap viability is desired. In addition, it would be beneficial for the surgeon to have an idea of the position and the depth at which the occlusion is situated so as to better evaluate the more convenient and safer intervention to do.

In this context, “time-resolved near-infrared spectroscopy” appears to be an attractive technique to noninvasively assess tissues’ oxygenation and deep perfusion thanks to its capability to encode depth in temporal delays of the time-resolved photon distributions.⁹ Indeed, photons that arrive later have traveled deeper, thus carrying information about oxygenation of deep layers. In recent years, it has been also demonstrated that the adoption of a small distance between the injection and collection fibers in reflectance geometry allows one to increase the spatial resolution and the overall number of detected photons without any loss in terms of depth sensitivity.¹⁰ Recently, we also showed that the use of a fast-gated single-photon avalanche diode (SPAD) coupled with a time-correlated single-photon counting (TCSPC) system permits to reach a dynamic range up to eight orders of magnitude¹¹ (a limit set by the memory effect of silicon SPADs¹²), thus increasing retrieved information about deeper layers.

In a previous work,¹³ we presented a setup for diffuse optical tomography (DOT) based on fast-gated SPADs. High-dynamic range acquisitions were analyzed using the Mellin–Laplace transform,¹³ thus obtaining tomographic reconstructions of an absorbing inclusion within the turbid medium with better depth sensitivity with respect to classical free-running (i.e., “nongated”) acquisitions. Also, we demonstrated that the adoption of short source-detector distances, combined with a fast-gated SPAD, can increase spatial resolution in DOT.

The goal of this work is to explore with a proof-of-principle test the suitability of an optical tomography system to assess flap viability. Indeed, DOT can provide a three-dimensional (3-D) reconstruction of oxy- and deoxygenated hemoglobin changes within the flap. This solution could allow surgeons to ideally continuously monitor the flap perfusion and, in case of complications, to plan the most suitable intervention.

To achieve this goal, we developed a more advanced setup for fast-gated tomography with respect to what was previously shown in Ref. 14. A fast wavelength multiplexing and the exploitation of the photon temporal distribution for a limited set of source-detector couples lead to an acquisition time for a full 3-D tomography of 3 min, which is adequate for following the spatio-temporal evolution of oxy- and deoxy-hemoglobin. As a first test, we used this setup to perform a preclinical study on abdominal fascio-cutaneous flaps in a rat undergoing a vein occlusion to simulate impairments in blood perfusion.

This paper is organized as follows: Sec. 2 deals with the experimental setup, measurement procedure, flap surgery, and data analysis; results are presented in Sec. 3, while in Sec. 4 we summarize the work and discuss the future perspectives.

2.

Material and methods

2.1.

Experimental Setup

A schematic of the setup is shown in Fig. 1. A supercontinuum fiber laser (Fianium LDT, London, UK) provided short light pulses at the repetition rate of 40 MHz. The illuminating source wavelength was selected by means of three interference filters (Coherent Inc.) centered at 750, 800, and 850 nm with a full-width at half maximum of 10 nm. We selected those three wavelengths because of their particular position across the oxy- and deoxygenated hemoglobin spectrum (respectively, before, on, and after the isosbestic point). The filters were mounted on a motorized wheel which was controlled by a dedicated software to speed up the measurement. Light exiting from the interference filter was then attenuated by means of a motorized variable optical attenuator (VOA) and sent into an optical fiber switch (core size: $62.5\mu \mathrm{m}$) so as to address six different output fibers. In the experiments presented in the following, in order to keep the measurement time to a minimum, only two output fibers were used to shine the sample into two different points of the flap [S1 and S2 in Figs. 1a and 2(a)].

Fig. 1

(a) Schematics of experimental setup. VOA is variable optical attenuator; FG-SPAD is fast-gated SPAD module and in the probe the letters “S” and “D” indicate the source (injection) and detector (collection) fibers, respectively. (b) Picture of the probe. (c) Probe placed on the flap during the measurement.

Fig. 2

(a) Rat after surgery: on the left flap are represented axes used for reconstruction maps presented in the followings and positions of injection (S1 and S2) and collection fibers (D1 and D2). (b) Reconstructed volume with axes and representation of $xy$ and $yz$ planes.

Retro-diffused photons were then collected by means of two $400\text{-}\mu \mathrm{m}$ core optical fibers (one for each fast-gated SPAD module) at a source-detector distance of 10.6 mm. Photons were then focused onto the SPAD ($100\text{-}\mu \mathrm{m}$ diameter active area) of the fast-gated module, whose characteristics are described in Ref. 15.

The optical probe [see Fig. 1(b)] consisted of these optical fibers fixed in a 2.5-cm diameter, 17.5-cm long cylinder.

The synchronization signal provided by the laser was used to trigger the fast-gated SPAD modules—after a proper adjustable delay—and the two TCPSC boards (SPC 130, Becker&Hickl GmbH, Germany).

The fast-gated acquisitions were fully automatized by means of a dedicated software. For each wavelength, the delayer was set to three different values. This enabled to acquire the time-resolved curve in three different time regions and to optimize the count rate for each region in order to obtain a good signal-to-noise ratio, while keeping it within the single-photon counting statistics.¹¹ For each delay, light was sequentially injected into two points of the probe using two different outputs of the switch. The attenuation introduced by the VOA was chosen in order to optimize the signal recorded from both detectors at the 10.6 mm interfiber distance.

3.

Results

Figure 2 (a) depicts rat 1 after the flap surgery in both sides (left and right). On the left flap, the positions of injection and collection fibers are represented together with the orientation of $x$- and $y$-axes used for the tomographic reconstruction. The $z$-axis refers to the depth, as can be seen in Fig. 2(b).

Figure 3 shows the reconstructed maps of variation in oxy- and deoxygenated hemoglobin in the $yz$ and $xy$ planes, obtained for the left flap of rat 1 at different times from the reference measurement (start of the occlusion). The acquisitions are performed every 3 min, starting immediately after the occlusion. The clamp on the vein was removed after 57 min (19 series of acquisitions). To reduce the number of panels, Fig. 3 shows maps at steps of 12 min. Maps for both planes at all times are presented in Video 1.

Fig. 3

Reconstructed maps of variation of oxy- and deoxygenated hemoglobin for the left flap of rat 1 at different times from the start of venous occlusion (see Video 1 for all times). (a) The left columns display the maps of HHb and ${\mathrm{HbO}}_2$ in the $yz$ plane (thus enlightening the depth distribution of chromophores) while (b) the right columns present the evolution of oxy- and deoxygenated hemoglobin in the $xy$ plane at a depth of 3 mm. (Video 1, MPEG, 5.72 MB) [URL: http://dx.doi.org/10.1117/1.JBO.21.2.025004.1].

Figure 3(a) shows the evolution of HHb and ${\mathrm{HbO}}_2$ changes in the $yz$ plane [see Fig. 2(b)] for $x=-4\mathrm{mm}$. Looking at the maps, it is possible to notice that the HHb accumulation is mostly localized between the surface and a depth of 5 mm, while the maximum increase of HHb is at $z=3\mathrm{mm}$. This value is reasonable, considering that the thickness of the flap is about 3 to 5 mm. Small variations in concentration can be also noticed up to 10 to 15 mm in depth since, after the occlusion, the layer under the flap can also experience some modifications. In addition, the possible formation of a thrombus within the flap can increase its thickness, thus enlarging the volume where variations in oxygenation are detected.²⁴

In reference to the time evolution, the accumulation of HHb starts 9 min after the clamping of the vein and increases with time. On the other hand, ${\mathrm{HbO}}_2$ concentration does not vary significantly with respect to HHb values. This is due to the fact that the arterial circulation was not massively altered by the specific vein occlusion. Even in the case of a total venous occlusion, part of the venous congestion can possibly be drained by the edge of the flap. The consequences on the arterial flow might be delayed causing the effect on ${\mathrm{HbO}}_2$ to appear later. It is also true that the small size of the flap (and consequently the small vascular flow) made it more difficult to identify flow alterations. However, this issue on flap dimensions is related to the use of rats, and will be overcome in the next studies with a model closer to humans.

In the two columns of Fig. 3(b), we show the reconstructed maps in the horizontal $xy$ plane at the depth $z=3\mathrm{mm}$, where the maximum HHb change was observed. After 6 min from the start of the occlusion, no significant accumulation of HHb is present, while after 18 min, an increase of HHb, can be clearly seen, which evolves in time. It is also possible to notice an increase of the area where HHb is concentrated. After 54 min the accumulation of HHb is higher and more spread out in the space as compared to the beginning. This phenomenon is most probably due to an accumulation of HHb that could not be washed out because of the occlusion. For that reason, the HHb concentration increased, leading to the formation of a thrombus. This hypothesis was verified after the euthanasia of the rat: the flap was unsewn, confirming the presence of a thrombus.

For ${\mathrm{HbO}}_2$, there is little accumulation in the same region where HHb concentration is increased, as already shown for the $yz$ plane.

In the HHb and ${\mathrm{HbO}}_2$ maps in the $xy$ plane, some artifacts may be observed: these are possibly due to both the geometry of the probe that was reduced in these experiments to four source-detector couples as well as to the assumption of a homogeneous medium as a reference state.

The temporal evolution of HHb and ${\mathrm{HbO}}_2$ up to 54 min after the vein occlusion is shown in Fig. 4 for three different pixels: one where the increase in HHb concentration is maximum (a), another where it is less pronounced (b), and the last where no major changes happened (c). The plots were extracted from the reconstructed 3-D maps at a depth of $z=3\mathrm{mm}$ and binning of adjacent pixels (bin factor: two).

Fig. 4

Binned maps of HHb variation in concentration in the $xy$ plane (for $z=3\mathrm{mm}$) and time evolution of HHb and ${\mathrm{HbO}}_2$ concentrations in three different points (named a, b, and c).

Looking at the time evolution in Fig. 4, we can notice that for pixel a there is an increase in HHb which is steeper for the first 25 min while later it is much slower (this last part is probably due to the formation and then the slow increase in dimensions of the thrombus). On the other side, ${\mathrm{HbO}}_2$ does not vary significantly in its concentration.

In pixel b, the amount of HHb is quite stable throughout the whole experiment as well as the ${\mathrm{HbO}}_2$ content, which varies without following a particular trend.

In pixel c, no significant changes as compared to the background are present: fluctuations of HHb and ${\mathrm{HbO}}_2$ (see inset) are negligible with respect to the changes in HHb induced by the vein occlusion.

The experiment performed on the other flap (right) of the same rat displays a temporal and spatial trend similar to the one shown in Fig. 3 (see Video 1), just with a specular pattern, meaning that the maximum increase of HHb is observed on the right. This is consistent with the expected specular symmetry of blood perfusion in the left and right flaps.

The same overall behavior is observed in a second rat (see maps shown in Video 1), thus confirming the observed temporal and spatial evolution of concentrations of HHb and ${\mathrm{HbO}}_2$ after a venous occlusion.

The results obtained in all three flap experiments are consistent, thus allowing us to state that the proposed setup is able to properly detect the variation in perfusion after a venous occlusion, which mimics the possible complications arising after a flap surgery.

In these preclinical tests, we do not fully exploit all potentialities of the fast-gated high dynamic range acquisitions since the flap thickness on rats is pretty small (3 to 5 mm) with respect to human flaps (1.5 to 3 cm). However, the improvement in reconstruction given by the high dynamic range will be a crucial feature in future clinical tests on humans where blood vessels are deeper, thus requiring the better depth sensitivity given by fast time-gated technique¹⁰ to monitor the permeability of veins and arteries after flap surgery.

4.

Conclusions and Future Perspectives

Monitoring flap viability is an unmet clinical need. Early diagnosis is needed in order to quickly act to save the flap by rapidly restoring the vessels’ permeability. In this paper, we presented a new prototypal and fully automatized setup for fast time-gated optical tomography based on four source-detector couples. The aim of this work was to validate this setup for the noninvasive assessment of flap viability by means of preclinical in vivo measurements on rats. We demonstrated the possibility to get dynamic 3-D tomographic reconstruction of oxy- and deoxygenated hemoglobin thus detecting changes in concentrations caused by a venous occlusion. Indeed, a venous clamp is a relevant model²⁵ to simulate complications that can arise after flap surgery. In this case, since the vessels are not well reconnected, an increase in deoxygenated hemoglobin and the formation of a thrombus occurred.

At present, the acquisition time was about 3 min for a full 3-D image, which is already compatible with clinical needs (continuous check in the first 6 h and then measurements every 6 h for a few days). In the case of human flaps, the monitoring lasts for six days with control measurements every 6 h as routine. Considering the present clinical monitoring protocol, it is clear that our overall acquisition time is not a real concern. The total duration of a complete tomography is about 3 min and the time needed for the data analysis is about 2 min on a standard PC without any optimization process. Furthermore, in the future, sampling time can be reduced down to 1 min by reducing the dead time to a few seconds when changing the gate delay.

These positive results on the venous occlusion of superficial flaps in a rat encourage the further pursuit of this work. The next step is to tackle an animal model (e.g., pigs) closer to human flaps in terms of composition and thickness. Both venous and arterial occlusions should be investigated in this preclinical case before tackling clinical tests. In that situation, we will exploit at best the potentialities of the fast-gated SPAD since flaps will have thicknesses closer to the human case.

Future works will also consider the optimization of the probe and an increase in the number of source-detectors couples. This increase in measurement points will ameliorate 3-D reconstruction of oxy- and deoxygenated hemoglobin concentrations, thus possibly reducing artifacts in reconstructed maps and improving spatial resolution.

In addition, the development of low-cost and miniaturized detectors (e.g., silicon photomultipliers),^26,27 compact light sources (e.g., vertical-cavity surface-emitting lasers),²⁸ and improved electronics, replacing the expensive TCSPC instruments,²⁹ will enable the realization of low-cost and high-performances diffuse optics devices opening the way to the clinical use of this technique.