1. Introduction

Pulmonary disease is a lung disease or a systemic disease of pulmonary manifestation. According to a report of the World Health Organization (WHO), chronic obstructive pulmonary disease was predicted that it would rank from the 5^th to the 3^rd widely death and disability of the world in 2030 [1]. In China, lung cancer increased to the most incidence cancer [2], tuberculosis increased to the number one infectious disease, as well as pneumoconiosis accounted for 90% of occupational diseases, with the increased air pollution, smoker population, aging population, and newly emerged drug-resistant pathogens [3]. The diagnosis and treatment of pulmonary diseases become more and more significant, so non-invasive real-time lung detection method will be important to the prevention and treatment of pulmonary diseases.

At present, the detection of pulmonary diseases mainly relied on imaging approaches, including X-ray tomography and computed tomography. These methods provided a clearer observation of the lung shadows shape and mass [4]. However, both approaches required the patient to stay still and stopped breathing as much as possible to prevent artifacts caused by slight movements or activities [5], which was tough for severe patients. Particularly, both methods required X-ray radiation, which was invasive, harmful, and could not be used in continuing way. The detection of pulmonary diseases could also use pathological, bacterial and biochemical tests through skin puncture and lung biopsy tissue [6]. Although these detection methods could improve the diagnosis rate, they were invasive and could not provide timely feedback from the disease [7]. As you knew, there were so many patients with pulmonary diseases requiring noninvasive, continues, and ease monitoring in a ward or Intensive Care Unit (ICU), such as lung cancer, severe pneumonia, emphysema, and pulmonary embolism.

Considering near-infrared spectroscopy (NIRS) has been successfully utilized in monitoring functional activity of brain, shock disease or tumor in neck, hemodynamics of muscle and breast [8–11]. These parts have good Monte Carlo (MC) simulations in the biomedical optics community. it was not unreasonable to ask if near-infrared light could be used in noninvasive monitoring of pulmonary diseases. Of note, NIRS could provide noninvasive measurements of multiple hemodynamic parameters, including oxy-hemoglobin concentration, deoxy-hemoglobin concentration, and blood volume, which was quite important and potential to be a useful complement of above X-ray or CT imaging methodologies [12–14]. In light of this, we proposed to explore the possibility of noninvasive monitoring of lung hemodynamics by near-infrared light. In this paper, we attempted to modeling the near-infrared light propagation in the lung in the framework of NIRS monitoring placed on the skin covering the lung. By this way, we explored if the light could penetrate into the lung from the skin.

Compared to diffusion theory, Monte Carlo (MC) method was more accurate and flexible, especially for structured tissue [15]. The pioneers Steven L. Jacques’s group developed the MC model for light transport in multi-layered tissue (called MCML) [16], which was widely used in the community of biomedical optics and was accepted as the golden standard. Since most portion of human tissue presented complex three-dimension (3-D) anatomical structure, researchers devoted to develop MC model for 3-D structured tissue to acquire a more precise modeling of light propagation in human body. After that, L. Wang’s group developed a MC simulation software for light transport in soft tissue within heterogeneous geometry [17–19]. Then, Boas’ group developed a MC simulation model for light migration in MRI slice described structures [20]. Followingly, T. Li’s group developed a MC modeling approach to light migration in multi-voxel media (called MCVM) [21], which had been fully tested and applied for biomedical optics. In recent years, Dholakia's group had proposed the signed MC radiation transfer (sMCRT) algorithm, a geometry-based method that used a signed distance function (SDF) to represent the geometry of the model. SDFs were capable of accurately modeling smooth surfaces as well as complex geometries [22].

In order to accommodate the development of MC methods, a well-known limitation of MC methods in terms of computation time also urgently needed to be addressed. Benefiting from the emergence of massively parallel computing techniques and the rapid development of many core processors such as graphics processing units (GPUs), the speed of MC simulations had increased significantly. Steven L. Jacques’ team developed hardware-accelerated MC simulation algorithms capable of accurately simulating polarized light within three-dimensional (3D) heterogeneous tissues [23]. P. Li’s team proposed a compute unified device Architecture (CUDA)-based MC algorithm for polarized light that uses a single-core scheme and is hundreds of times faster than the CPU version [24]. Later, Oulhaj's group reported a GPU-accelerated MC algorithm that efficiently computes the sensitivity distribution of polarized light within a homogeneous medium. refinement of various aspects of the MC approach had spawned a number of series of MC-based simulation platforms [25]. Doronin's group reports on the development of a unified MC-based simulation platform and considers its practical application in the creation of novel optical diagnostic, imaging, and sensing models aided by artificial intelligence (AI) methods [26–28]. Q. Fang's group developed a scalable, high-performance, in-browser Monte Carlo simulation platform with cloud computing (called Monte Carlo eXtreme (MCX) Cloud) [29]. J. Jönsson’s group has optimized a Monte Carlo software based on Lorentz-Mie theory phase functions using a CUDA parallelized model [30,31]. This validated software called “Multi-Scattering” is now directly accessible online through a webpage. [32,33].

It was worth noting that most Visible Man Project chose to record anatomical structure of human body in voxelated way. Thereinto, Visible Chinese human (VCH) was most faithfully represented the human anatomical structures [34]. It was developed according to the most high-resolution cryosectional color photographs of a reference adult, following with a fine segmentation supervised by anatomists [35]. Thus, the VCH thoracic tissue with lung in it was regarded as a most realistic thoracic tissue model containing the precise folding of geometry lung and superficial tissue. In addition, MCVM had been successfully combined with VCH in a serials of Monte Carlo modeling on light propagation studies [36]. Therefore, we proposed to use MCVM-VCH to address the issue on exploring if the lung could be reached and detected by the near-infrared light.

Accordingly, this study proposed to research on the near-infrared light propagation in all five lung lobes by MCVM-VCH modeling on light propagation and experimental measurements in NIRS framework. For the MCVM-VCH modeling, the photon propagation properties were extracted out from statical analysis, including signal sensitivity distribution, partial path-length factor (PPF), differential path-length factor (DPF), and the lung signals contribution ratio. The penetration depths of detected near-infrared light signal were accumulated for all lung lobes, which indicated the strong potential for noninvasive optical monitoring of lung hemodynamics. Plus, a custom NIRS for lung monitoring was developed. Then we carried out a prior experiment with the NIRS to measure the detected diffused light reflectance on 14 volunteers and compared these values quantitatively among different placement of NIRS probe into the different lung lobes. These measurements showed a good agreement with the simulation results. Both simulation and experimental studies convinced us the promising optical monitoring of lung hemodynamics in noninvasive and comfortable way. This will support further research human lungs in the biomedical optics community. Except for quantitatively visualizing the near-infrared light migration in human lung, this study also recommended the optimization of source-detector separations and the placements of source-detector for noninvasive optical monitoring in five different lung lobes.

2. Materials and method

2.1 Lung model and Monte Carlo simulation

In this experiment, the VCH slice specimens were from an adult male, that were sliced horizontally at regular intervals in a standing position. Each slice was a digital color photograph that was distinguished in tissues. This enabled VCH to truly represent the anatomical structure of the human body [34]. For the research of pulmonary diseases, it was an optimal choice to use VCH to establish a lung model. Figure 1(a) showed a representative slice of a raw lung photo from the VCH model [37]. The related segmentation image to the Fig. 1(a) was shown in Fig. 1(b). When simulating the lung, it was divided into five lobe regions (Fig. 1(c)). 200 such images of the Fig. 1(b) combined with image processing methods to construct a three-dimensional matrix to present the entire lung tissue structure (Fig. 1(d)). According to the location of the lung lobes, five lung lobe models of 420 × 436 × 200 voxels were segmented from Fig. 1(d). The results of the simulation also showed the specific situation of the five lung lobes. Each voxel was a 0.04 × 0.04 × 0.04 cm³ cube. The lung model included 10 tissue types: skin, muscle, bone, fat, liver, lung, myocardium, stomach, arterial blood and venous blood [38]. The MC method was used to simulate the migration of light in the lungs. In the theory, after a large number of photons entered the tissue medium, photons were randomly scattered or absorbed by particles in the medium during the movement until they “die” or “escape”. The scattering and absorption were simulated and calculated by a series of possibility functions and survival criteria. After the surviving photons were collected by the light detector on the skin surface, important information, such as their intensity and propagation path length, were recorded and calculated. This study used the software MCVM which was targeted for 3D voxelized media. In the simulation, the light source was set to 800 nm point light source, and the optical properties of 10 tissues in the near-infrared band of 800 nm were shown in Table 1 [39]. The initial position of the light source was set as shown in Fig. 2(a) and (b). Different lung lobes required different positions of light source. The light source was located near the second left rib of the front sternum for the superior lobe of left lung. The light source of the superior lobe of right lung lobe was located near the second right rib of the front sternum. The light source of the middle lobe of right lung was located near the fourth right rib near the sternum body. The light source of the inferior lobe of left lung was located on the eighth left rib. The light source of the inferior lobe of right lung was located on the eighth right rib. In the simulation, the detection area of the detector was 2 mm × 2 mm, and the detection result will show the light fluence distribution of the detected photons. 1 × 10⁶ photons were emitted in the human body. 10 times of simulations were performed and the average was accumulated in the following data analysis. Two output files were then obtained from the software MCVM, which represented absorbed photon amounts and the distribution of output photons, respectively.

 

Fig. 1. Illustration of the lung image model of the Visual Chinese Human (VCH). (a) VCH original lung slice; (b) tissue segmentation from the (a) image; (c) the five lobes of the lung; (d) The 3D image is an arrangement of 200 processed (b) images.

Download Full Size | PDF

 

Fig. 2. The light fluence distribution in five lung lobes. “S” is source, “D” is detector. (a) the light source positions for detecting superior lobe of left lung, superior lobe of right lung, and middle lobe of right lung; (b) the schematic diagram of the detection light source of five lung lobes; (d) the 3D space coordinate system used in the simulation; (c, e, f, g, h) represented light fluence distribution under X-Z view in the right lung inferior lobe, the left lung inferior lobe, the left lung superior lobe, the right lung superior lobe, and the right lung middle lobe, respectively.

Download Full Size | PDF

Table 1. Optical properties of the tissue of the lung model under 800 nm light

View Table

2.2 Spatial sensitivity distribution

The spatial sensitivity distribution (SSD) was the volume of tissue contributing to the change in the intensity of the detected light, which was calculated from the cumulative trajectory of the detected light [40]. The calculation functions were introduced in the previous paper. In brief, the formula (1) of the three-point Green's function was used to calculate the SSD [36]. Among them, $F({{{\vec{r}}_S},{{\vec{r}}_m}} )$ represented the signal amount from the m-th voxel to the light source, and $F({{{\vec{r}}_D},{{\vec{r}}_m}} )$ represented the signal amount from the m-th voxel to the detector. $F({{{\vec{r}}_S},{{\vec{r}}_m}} )$ and $F({{{\vec{r}}_D},{{\vec{r}}_m}} )$ were calculated by the formula (2). Among them, $A({{{\vec{r}}_x},{{\vec{r}}_m}} )$ represented the absorption of voxel ${\vec{r}_m}$ at ${\vec{r}_x}$, which come from the absorbed photon file. Besides, $\mu ({{{\vec{r}}_m}} )$ represented the absorptance of voxel ${\vec{r}_m}$, which could be found in Table 1.

2.3 Distance between the light source and the detector

In this research, the optimal L_SD in the migration of photons in lungs was also calculated. The setting of the light source should reduce the absorption of muscles as much as possible, and retain more photons to reach the lung. The detector was located at left side of the light source for detecting two lobes of the left lung. The detector was located at right side of the light source for detecting three lobes of the right lung.

The most optimal L_SD depended on the differential path-length factor (DPF), partial path-length factor (PPF), and the ratio of PPF to DPF. DPF was calculated from the average path length of light propagation in lung [35]. PPF was the average path length that converted the L_SD into the focal region of light passing through the chromophore change [52,53]. The PPF of the lung tissue was calculated using the output photon distribution file, which indicated the sensitivity of the detected light intensity signal to the lung tissue. The ratio of PPF to DPF indicated the contribution of the lung tissue to the detected light signal. In general, the PPF/DPF represented the contribution of target tissue to the absolute absorption measurement.

2.4 Experiment

A preliminary experimental study aimed to detect light intensity signals in the lungs. A device based on near-infrared technology was developed by our research group [54]. Similar near-infrared devices have demonstrated performance reliability and stability [55–57]. The device consisted of three modules: near-infrared detectors, a control and processing module, and a data acquisition application. In the probe, the 800 nm LED (AlGaAs material) was the light source and the OPT101 (OPT101 Monolithic Photodiode and Single-Supply Transimpedance Amplifier) was the photoelectric sensor. The spacing of 800 nm LED and near-infrared detectors was 2.8 cm. This spacing was the best L_SD for most lung lobes. The detector was assembled from a flexible circuit board, which adhered better to the skin and was less disturbed by ambient light. The control and processing module took the signal from the probe and sent it to the mobile app via Bluetooth.

In the preliminary experimental study, light intensity signals were measured in the lungs of 14 young healthy volunteers. The 14 volunteers included four women and 10 men. Their age was concentrated between 25-35 years old. During the experiment, the light sources were placed in the corresponding positions of the five lung lobes in turn (Fig. 3). Volunteers’ prothorax detected light intensity at three locations (Fig. 3(a)). Back of the volunteer detected light intensity at two locations (Fig. 3(b)). All subjects were asked to lie down quietly and rest for 2-3 minutes before measurement to allow their breathing to stabilize. The collection instructions were set on the near-infrared device application. The main instructions are the wavelength of light source and the collection frequency. These instructions were transmitted to the near-infrared device using Bluetooth. Then the light intensity signals at each position were collected by the near-infrared device. The collection instruction was issued to the end of the collection for 2 minutes.

 

Fig. 3. Preliminary experiment of lung light intensity signal. ①-⑤ represented the position of superior lobe of right lung, superior lobe of left lung, middle lobe of right lung, inferior lobe of left lung, and inferior lobe of right lung, respectively. (a) the volunteer was detected in positions ①, ② and ③ in the front of body (b) volunteer’s back was detected in positions ④ and ⑤.

Download Full Size | PDF

3. Results

3.1 Lung SSD

The SSD of the lung was obtained by performing Monte Carlo simulation calculations using the VCH lung model as previously described. The represented light fluence distribution images of the five lobes were shown in Fig. 4. According to the both Formula (1) and Formula (2), the SSD in five lung lobes were calculated (inferior lobe of left lung, 0.0235%; superior lobe of left lung, 0.0252%; inferior lobe of right lung, 0.0368%; middle lobe of right lung, 0.0348%; superior lobe of right lung, 0.0270%). Before photons reached the lung, they passed through four tissues, including skin, fat, muscle, and bone. Due to the variation of the fat distribution, photons passed the fat tissue for different times to achieve different lobes [58]. The light fluence intensity in the log10 scale gradually decreased from the skin layer to the lungs. In details, the light source emitted photons in the inferior lobe of left lung from the left back. The penetration depth was 33.8 mm (Fig. 4(a1)). When the photons achieved the left lung inferior lobe, the light fluence intensity decreased to 2/voxel and penetrated 7.4 mm in lung tissue (Fig. 4(a2)). The light source entered the superior lobe of left lung from the left chest. The penetration depth was about 32 mm (Fig. 4(b1)). When the photons achieved the superior lobe of left lung, the light fluence intensity decreased to 12/voxel and penetrated 8.4 mm in lung tissue (Fig. 4(b2)). The light source emitted photons in the inferior lobe of right lung from the right back. The penetration depth was 36 mm (Fig. 4(c1)). When the photons achieved the right lung inferior lobe, the light fluence intensity decreased to 10/voxel and penetrated 7.2 mm in lung tissue (Fig. 4(c2)). The light source emitted photons in the middle lobe of right lung from the right chest. The penetration depth was 34.8 mm (Fig. 4(d1)). When the photons achieved the right lung middle lobe, the light fluence intensity decreased to 3/voxel and penetrated 6.0 mm in lung tissue (Fig. 4(d2)). The light source emitted photons in the superior lobe of right lung from the right chest. The penetration depth was 33.2 mm (Fig. 4(e1)). When the photons achieved the right lung superior lobe, the light fluence intensity decreased to 11/voxel and penetrated 6.8 mm in lung tissue (Fig. 4(e2)). In general, photons at 800 nm could be transmitted as deep as 32-36 mm in the human body, that the last 6-8.4 mm was in the lung. The light fluence intensity dropped sharply by a factor of about 10⁵ from skin to lung. Though the intensity was quite low at the lung, it still could be accurately detected. It indicated that the non-invasive optical detection was available.

 

Fig. 4. The light fluence distribution images and Changes of light fluence intensity as tissue depth in five lung lobes. (a1-e1) represented the light fluence distribution of left lung inferior lobe, left lung superior lobe, right lung inferior lobe, right lung middle lobe, right lung superior lobe, respectively. The point of “0” and the arrow represented the starting point of the light source and the photon emission direction, respectively. (a2-e2) represented the change of light fluence intensity in 5 tissues in the inferior lobe of left lung, the superior lobe of left lung, the inferior lobe of right lung, the middle lobe of right lung, the superior lobe of right lung, respectively.

Download Full Size | PDF

3.2 Photon absorption

When light entered the body, the photons were either absorbed, annihilated or scattered by tissues. Most photons were absorbed by the tissue medium during migration. The photon absorption in lobes was investigated (Fig. 5). In the left lung, the photon absorption of the inferior lobe was 6%-8%, as well as the photon absorption of superior lobe was 7%-8% (Fig. 5(a)). The photon absorption of the left lung lobes changed with the L_SD. At the L_SD of 2.5 cm, the maximum photon absorption achieved in both lobes. In the right lung, the photon absorption of the inferior lobe was 9.5%-11% (Fig. 5(b)). At the L_SD of 2 cm, there was the maximum absorption. The photon absorption of the middle lobe was 8%-9.5% (Fig. 5(b)). When the L_SD was 2.5cm, the maximum photon absorption achieved; the overall photon absorption of the superior lobe was 7.8%-8.2%. When the L_SD was 3cm, the absorption was maximum. In conclusion, the average photon absorption of lung was about 9%. After achieving the maximum photon absorption, the photon absorption decreased when the L_SD increased. It was indicated that the photon absorption changed with the L_SD.

 

Fig. 5. Photon absorption in five lobes with the change of L_SD. (a) Photon absorption in the left lung lobes changed with L_SD; (b) Photon absorption in the right lung lobes changed with the L_SD.

Download Full Size | PDF

3.3 Optimization of the distance between light source and detector

Finally, the optimization of the L_SD was analyzed. With the change of L_SD, the value of DPF, PPF, and PPF/DPF changes. Therefore, the optimal L_SD can be found through comparison of these value changes. The changes of DPF, PPF, and PPF/DPF of the five lung lobes with the L_SD were calculated (Fig. 6). Overall, the DPF of the five lung lobes showed an exponential growth trend as the L_SD increases (Fig. 6(a1) - (a5)). For the PPF index, the inferior lobe of left lung reached a peak at L_SD of 2.7 cm, and then dropped rapidly (Fig. 6(b1)). The superior lobe of left lung had a peak at L_SD of 3.6 cm and a maximum at L_SD of 2.9 cm (Fig. 6(b2)). The inferior lobe of right lung reached a peak at L_SD of 3.5 cm, there was a maximum at L_SD of 2.7 cm (Fig. 6(b3)). The middle lobe of right lung had a peak at L_SD of 2.7 cm, and then dropped sharply (Fig. 6(b4)). The superior lobe of right lung had a peak at L_SD of 3.5 cm, and it dropped significantly when it was greater than this value (Fig. 6(b5)). On the whole, the PPF index of the five lung lobes reflected the sensitivity of the light intensity signal to the lungs, which showed a first increasing trend and then decreasing trend with the L_SD. Optimizing L_SD also relied on the PPF/DPF ratio, which represented the contribution of lung tissue to the detection signal. The PPF/DPF ratio of the inferior lobe of left lung reached its peak when L_SD was about 2.7 cm (Fig. 6(c1)). The ratio of the superior lobe of left lung reached its peak when L_SD was about 2.9 cm (Fig. 6(c2)). The ratio of the inferior lobe of right lung reached its peak when L_SD was about 2.7 cm (Fig. 6(c3)). The ratio of the middle right lobe reached its peak when L_SD was about 2.8 cm, and then decreased fast (Fig. 6(c4)). The superior lobe of right lung reached the peak at about 3.4 cm L_SD, following with a sudden drop (Fig. 6(c5)). The optimal L_SD range of the five lung lobes were calculated (inferior lobe of left lung, 2.8-2.9 cm; superior lobe of left lung, 2.7-2.9 cm; inferior lobe of right lung, 2.7-2.8 cm; middle lobe of right lung, 2.7-2.8 cm; superior lobe of right lung, 3.3-3.5 cm).

 

Fig. 6. DPF (blue) and PPF (green) and PPF/DPF (red) diagrams of five lobes as a function of L_SD. (a1-a5) The DPF of left lung inferior lobe, left lung superior lobe, right lung inferior lobe, right lung inferior lobe, and right lung middle lobe, respectively; (b1-b5) The PPF of left lung inferior lobe, left lung superior lobe, right lung inferior lobe, right lung inferior lobe, and right lung middle lobe, respectively; (c1-c5) The PPF/DPF of left lung inferior lobe, left lung superior lobe, right lung inferior lobe, right lung inferior lobe, and right lung middle lobe, respectively.

Download Full Size | PDF

3.4 Experiment results

The light intensity signals of 14 healthy volunteers collected in the experiment were processed. On the one hand, the light intensity signals of the five positions obtained in the experiment were filtered to eliminate some interference effects. Then the median of the signal light intensity of this section was calculated to represent the light intensity of the position. On the other hand, the photon absorption of the simulated five lung lobes was used to represent the light intensity of each lung lobe. Here, the photon absorption of the same L_SD as the experiment was selected. The light intensity of the five lung lobes of a male volunteer in the experiment was compared with the simulated light intensity. The experimental values were all smaller than the simulated ones (Fig. 7(a)). But the experimental value and the simulation value had a great correlation, p < 0.05 (Fig. 7(b)). Next, the median of the light intensity of the 14 volunteers was selected and compared with the simulated values. It could be seen that both have similar trends (Fig. 7(c)). The experimental and simulation values also had a strong correlation, p < 0.05 (Fig. 7(d)).

 

Fig. 7. Comparison of summed light intensity in detected area on the skin covering the lung between experimental and simulation values. (a) The experimental light intensity of five lung lobes in a male volunteer was compared with the simulation value; (b) Fitting of experimental and simulated values in (a); (c) Comparison of median light intensity per lobe with simulated values in 14 volunteers; (d) Fitting of experimental and simulated values in (c). SR, the superior lobe of right lung; IR, the inferior lobe of right lung, MR, the middle lobe of right lung; SL, the superior lobe of left lung; IL, the inferior lobe of left lung. R² coefficient of determination; p < 0.05*, p < 0.01**.

Download Full Size | PDF

4. Discussion and conclusion

Based on the VCH lung model (Fig. 1), this paper used the Monte Carlo simulation method to quantitatively analyze the photon migration characteristics in the human lung. The light fluence distribution in the lungs and the changes in light fluence intensity indicated that photons could reach the lungs from the light source (Fig. 3). The photon penetration depth from the skin to the lung was 32-36 mm, of which the photon penetration depth in the lung was 6-8.4 mm. The SSD for the five lobes was 0.0235-0.0368%. The average photon absorption of five lung lobes was about 9% (Fig. 4). The differences in photon migration in the five lung lobes were reflected. Besides, we proposed an optimization plan for the L_SD. According to the two path length factors of PPF and DPF, combined with the index of PPF/DPF, the L_SD in the lungs may be the optimal topological location between 2.7-2.9 cm, but the distance from the superior lobe of right lung was about 3.3-3.5 cm. Using VCH-based Monte Carlo simulation of human lungs, the optical migration characteristics of the lungs were quantitatively visualized, the feasibility of non-invasive optical detection of the lungs was proved, and the optimal L_SD for optical detection of the lungs was found.

We found that the SSD of the five lung lobes had a certain difference. The cause of the formation of this difference could be speculated from the formula 2. For these five lung lobes, the larger the SSD showed that the more photons reached the lungs. There were more photons that reach the right lungs than the left lungs, especially the photons in the middle right and lower right leaves are much more than the remaining three lung leaves. The photon penetration depth from the skin to the lung was 32-36 mm. It could be found that muscle had an important impact on the depth of penetration. The muscle layer through the lower right lobe was the smallest, and the photon penetration depth was the largest. The light fluence intensity reached the lungs was also the highest. The muscle layer of the lower left leaves was the thickest, so the light fluence to reach the lungs was the smallest [59]. From the perspective of photon absorption analysis of the five lung lobes, it could be found that the photon absorption of the right lung was much larger than the left lung. From the perspective of lung structure, the overall surface area of the left lung was smaller than that of the right lung. The left lung was affected by the heart away from the skin, so the photon absorption of the right lung was better.

For the more important parameter L_SD in optical non-invasive testing, this study had done optimization for the lungs. The optimal lung L_SD was 2.7-2.9 cm, but the optimal spacing of the superior lobe of right lung was 3.3-3.5 cm. Comparing the currently known research on the detection of other parts of the human body in the biomedical optics community, the optimal distance between the lungs was significantly smaller than the optimal distance in other parts. In previous research, the optimal distance for the heart was between 3.5-4 cm [8]. The optimal spacing of the head was 3-3.5 cm [9,10]. The optimal spacing for these different parts was about 3 cm. In fact, this separation was not the bigger the better. On the contrary, with the increase of the spacing of the source detector, the energy density of the transmitted photon was rapidly decreased, which seriously affected the quality of the detection signal [60]. The optimal separation preference depended on the detection of light density and path length factor PPF and DPF.

In a preliminary lung light intensity experiment, the L_SD spacing was 2.8 cm, which was within the optimal L_SD range for the lung. A better light intensity signal could be obtained by detecting in the optimal L_SD range. The experimental data show that the light intensity values calculated from the detection are much smaller than the simulated light intensity values (Fig. 7(a) and Fig. 7(c)). Although the experimental values are relatively small, they are still within the detectable range. However, it can be seen that there is a multiplicative relationship between the experimental and simulated values, and when the experimental values are multiplied by this multiplier, the simulated and experimental values for the five lung lobes will be very close to each other. The main reason for such a difference is that the detector used is generally not sensitive enough, so it is difficult to detect some of the weaker signals. Another reason is that the photon emission deviates from the detection area, which causes the detector not to detect. From the overall light intensity trend of the five locations, there was a correlation between the experimental value and the simulated value, p < 0.05 (Fig. 7(b) and Fig. 7(d)). It indicated that the detected light intensity signals at five locations may come from five lung lobes. This further proves the feasibility of non-invasive optical detection of human lungs.

In summary, this study was based on MC simulation to visualize the propagation of near-infrared light in the VCH lung model. The photon migration in the lungs was studied. It was analyzed that the photon absorption of each lobe was about 9%, and the optimal source-detector in the superior lobe of right lung was about 3.3-3.5 cm, while the optimal distance for the other lung lobes was about 2.7-2.9 cm. A preliminary experiment demonstrates the feasibility of optical detection of the lungs. This paper also provided a theoretical support for functional near-infrared spectroscopy lung detection and a reference for the optimal L_SD. It is hoped that this paper will promote the research on lung diseases in the biomedical optics community.