Open Access
Add to BibSonomy
Abstract
Assessing the metastatic status of axillary lymph nodes is a common clinical practice in the staging of early breast cancers. Yet sentinel lymph nodes (SLNs) are the regional lymph nodes believed to be the first stop along the lymphatic drainage path of the metastasizing cancer cells. Compared to axillary lymph node dissection, sentinel lymph node biopsy (SLNB) helps reduce morbidity and side effects. Current SLNB methods, however, still have suboptimum properties, such as restrictions due to nuclide accessibility and a relatively low therapeutic efficacy when only a single contrast agent is used. To overcome these limitations, researchers have been motivated to develop a non-radioactive SLN mapping method to replace or supplement radionuclide mapping. We proposed and demonstrated a clinical procedure using a dual-modality photoacoustic (PA)/ultrasound (US) imaging system to locate the SLNs to offer surgical guidance. In our work, the high contrast of PA imaging and its specificity to SLNs were based on the accumulation of carbon nanoparticles (CNPs) in the SLNs. A machine-learning model was also trained and validated to distinguish stained SLNs based on single-wavelength PA images. In the pilot study, we imaged 11 patients in vivo, and the specimens from 13 patients were studied ex vivo. PA/US imaging identified stained SLNs in vivo without a single false positive (23 SLNs), yielding 100% specificity and 52.6% sensitivity based on the current PA imaging system. Our machine-learning model can automatically detect SLNs in real time. In the new procedure, single-wavelength PA/US imaging uses CNPs as the contrast agent. The new system can, with that contrast agent, noninvasively image SLNs with high specificity in real time based on the unique features of the SLNs in the PA images. Ultimately, we aim to use our systems and approach to substitute or supplement nuclide tracers for a non-radioactive, less invasive SLN mapping method in SLNB for the axillary staging of breast cancer.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
The annual report of the International Agency for Research on Cancer (IARC) showed that breast cancer (BC) was the most common malignancy in women worldwide in 2020 [1]. Since nearly 50% of patients with BCs show no lymph node (LN) metastasis, sentinel lymph node biopsy (SLNB) has, over the last decade, replaced axillary lymph node dissection (ALND) as the gold standard for the axillary staging of BC because of SLNB’s much-reduced invasiveness [2,3]. SLNB requires the injection of either 99mTc-tilmanocept sulfur colloid or blue dye for SLN tracing, a procedure recommended by the National Comprehensive Cancer Network guidelines for SLN mapping. However, this scheme is still potentially hazardous due to its radioactive nature, and hospitals could have limited access to the radioisotope contrast agent due to local nuclear use policies and nuclide storage requirements. Single-agent mapping has also limited the therapy efficacy [4]. A mapping method that relies on nuclide tracers is incompatible with standard preoperative imaging modalities as well.
Thus, researchers have been motivated to develop a non-radioactive SLN identification method. To this end, photoacoustic [5]/ultrasound (PA/US) dual-modality imaging systems were developed. Photoacoustic imaging (PAI), also referred to as optoacoustic imaging, is an emerging imaging technology with many new biomedical applications [5–8]. The advantages of using PAI include (a) non-radioactive imaging at clinically relevant depths (penetrating up to several centimeters), (b) fast and real-time imaging, and (c) high spatial resolution defined by acoustic reconstruction. PAI can image the distributions of intrinsic tissue chromophores (e.g., hemoglobin, melanin, and lipids) and exogenous contrast agents [9–15]. For example, optical contrast agents are typically administered to differentiate SLNs from the surrounding tissues in PAI. Among the contrast agents used today, blue dye, indocyanine green (ICG), and carbon nanoparticle (CNP) are approved by the National Medical Products Administration (NMPA) of China for clinical use. Alejandro et al. also developed a dual-modality PA/US imaging system to noninvasively detect SLNs in patients (13 patients, 6 detected) with BCs by employing methylene blue (MBD) as the contrast agent [16]. The researchers’ PA system worked at two wavelengths to unmix the contrast agent’s signal from that of the background tissue. In another clinical study involving patients with melanoma, Stoffels et al. used multispectral (5 wavelengths) PA imaging to identify SLNs labeled by ICG [13]. However, small-molecule dyes like MBD and ICG are not ideal contrast agents for the preoperative detection of SLN since they quickly drain into the next-level LNs. Those contrast agents are instead commonly used intraoperatively to find the SLNs along the stained lymph ducts, yet complications still occur due to the confusion between blood vessels and lymph vessels (LVs), and between SLNs and higher-echelon nodes. Besides, using blue dye (isosulfan blue and methylene blue) for SLNB is discouraged during pregnancy [17]. Another problem with these clinical studies is that spectroscopic PA imaging is costly, slow, and subject to spectral errors [18].
Wu et al. proposed using CNP as the contrast agent for SLN mapping and demonstrated CNP’s superiority over the blue dye in a pilot study involving 36 BC patients [19]. The research, however, did not involve PA imaging. Later, Liu et al. [15] demonstrated the feasibility of using CNPs for PA-imaging-based SLN mapping in rats. CNPs have an average diameter of ∼150 nm after aggregation, which usually make them stay in the SLN for a relatively long time before draining into higher-echelon nodes. The large optical absorption of CNP also enhances the PA contrast for improved signal strength and deeper imaging depth [15]. In this paper, our goal is to propose a new clinical procedure to precisely locate SLNs to offer preoperative surgical guidance and validate its effectiveness in a pilot study involving patients with BCs. The new procedure is based on single-wavelength PA/US imaging using CNP as the contrast agent, and the system can noninvasively image SLNs with high specificity in real time based on the unique features of the SLNs in the PA images.
2. Materials and methods
2.1 Patients
Our prospective study was approved by the Medical Ethics Committee of Beijing Tsinghua Changgung Hospital and was compliant with the Health Insurance Portability and Accountability Act (Ethical approval No. 21181-0-01). Written informed consent was obtained from all participants. From December 2021 to April 2022, 11 participants were enrolled in our study at Beijing Tsinghua Changgung Hospital. The participants were diagnosed with BCs and needed ALND or SLNB. Our patient population has not been reported previously.
2.2 Limitations of the current SLNB procedure
Current SLNB guidelines recommend using dye tracers and radioactive tracers simultaneously during SLNB, yet the nuclide tracers have limited accessibility due to the potential ionizing hazard and high cost.
However, the single-agent mapping method has limited therapeutic efficacy. A study based on radioactive tracers showed a false-negative rate of 10.1% (7 false-negative events in 69 patients with residual disease) [4]. While blue dyes or fluorescent dyes are agents currently used to replace dual-tracer approaches, these methods still suffer from limited therapeutic efficacy. However, the false-negative rate can be reduced to 1.4% [4] by removing ≥ 3 SLNs and using a dual-tracer approach; nuclide-guided SLNB must be included as one of the methods (The radioisotope contrast agent is used preoperatively followed by a dye trace used intraoperatively for positioning.) [20].
If the radioisotope contrast agent is unavailable, neoadjuvant therapy cannot be performed since more than two SLN tracing methods are necessary; nuclide-guided SLNB must be included as one of the methods.
SLN mapping based on nuclide tracers is poorly compatible with traditional biomedical imaging modalities for information fusion since the former provides sound information and the latter provide image information.
2.3 Carbon nanoparticles
CNPs are approved by the NMPA of China for clinical use as a new type of contrast agent. Normally, 25 mg of CNPs will be injected; most nanoparticles will stay at the injection area, but a small portion will enter the lymphatic system to stain the SLNs. Researchers have clinically proven that the accumulation of CNPs will not affect the normal functions of LNs [19]. Different from small-molecule dyes such as MBD and ICG, CNPs are much larger in size (average diameter of 150 nm [15]), so they selectively enter the lymphatic vessels (100∼500 nm) rather than the blood capillaries (30∼50 nm). However, CNPs are more expensive than MBD and can stay at the injection site for a relatively longer time.
2.4 Patient imaging protocol
In the current study, CNPs used as the contrast agent were prepared with clinically recommended concentrations. CNP suspension injections (50 mg/ml) were purchased from Chongqing Lummy Pharmaceutical Co. Ltd (Chongqing, China) in a standard package of 0.5 ml and used for in vivo administration without any dilution and reduction. Figure 1 shows the overview of the PA/US imaging approach for SLN staging. Before the imaging, the skin of the target axillary area was cleaned. Typically, any excessive amount of hair in the scanned area was removed. CNPs (0.5 ml) were injected intradermally in the 9:00 and 12:00 o’clock positions in the areola area, followed by PA/US imaging after 20 minutes. Before imaging, the injection site was massaged to facilitate the spread of the drug into the LNs in the axillary regions. CNPs captured and retained in the LNs stained the corresponding LNs black, while no LVs were found to be stained. PAI was then performed by irradiating the tissue with pulsed laser light (at a single wavelength of 1064 nm) and detecting the generated PA signals on the tissue surface using a linear US probe to pinpoint the SLNs and provide the surgeons with guidance to perform SLNB. During PA imaging, US images were also obtained to supplement additional information for cross-validation. The ex vivo specimens were imaged again using PA after the SLNs were removed and were then sent for routine pathological analysis.
Fig. 1. Overview of the PA/US imaging approach for minimally invasive SLN staging. (a) Anatomy showing the relevant structures and injection sites of CNPs, which are drained to the SLNs through LVs. (b) Anatomy showing how PA/US imaging pinpoints the SLNs preoperatively. (c) Anatomy showing how PA/US imaging guides SLNB with minimal invasiveness.
2.5 Dual-modality PA/US imaging system
Figure 2(a-c) shows the diagram, point spread function (PSF), and three-dimensional (3D) visualization of the dual-modality PA/US imaging system used in our clinical study. A clinical ultrasound system (M_FE_A0, Stork Healthcare, China) was used for US imaging. A customized optical fiber-based light delivery system was coupled with a lab-designed linear US probe (128-element, center frequency: 5 MHz, −6 dB bandwidth 70%). Aligning with common usage in clinical settings, the focal length of the transducer was 2 cm [4]. The PA excitation light was generated by a neodymium-doped yttrium aluminum garnet laser (Nd: YAG; 1064-nm output, MQ/E, Beamtech Optronics, China), emitting 10 ns pulses at a repetition rate of 10 Hz. The fluence on the skin was about 50 mJ/cm2, below the safety limit of 100 mJ/cm2 at 1064 nm. A low-noise, 128-channel data acquisition system (MarsonicsDAQ128, Tianjin Langyuan Inc., China) was used to record the raw PA data. A time-division multiplexing scheme was enabled by a lab-made multiplexer that fed the signals received from the probe alternately to the US and PA systems. The US images were obtained directly from the ultrasound unit without any additional processing, while the PA images were reconstructed by the delay-and-sum algorithm and directly sent for display. All data processing was accelerated using CUDA to allow for real-time display of the images. According to the PSF shown in Fig. 2(b), the lateral and axial resolutions of the system were measured to be 0.37 mm and 0.17 mm, respectively.
Fig. 2. PA/US imaging system and its point spread function (PSF). (a) Schematic of the dual-modality photoacoustic and ultrasonographic imaging system. (b) The PSF of the PA imaging system (scale bar: 1 mm). (c) The 3D rendering of the system and its probe (inset). PA: photoacoustic. US: ultrasound. DAQ: data acquisition system.
2.6 Data acquisition and image processing
Imaging was performed using the PA/US imaging system. Each laser trigger, activated 100 microseconds after the US trigger, initiated the acquisition of a PA sinogram. Our system can simultaneously capture US and PA data to form co-registered US and PA B-scan images at a frame rate of 10 Hz. During the imaging procedure, a user interface was used to interactively control the sound speed, imaging depth, and gain coefficient. All PA/US signals were stored in a computer for postprocessing. During the offline processing, a digital bandpass filter was applied to the raw RF signals to suppress low-frequency artifacts. The reconstruction images were then Hilbert-transformed along the depth direction, and the absolute value was taken to suppress the negative artifact [21,22]. Notably, all PA images in this paper, except Fig. 2(b) and Fig. 4(d), have undergone post-processing to remove signals from the skin and highlight the SLN and vessel features by thresholding.
Table 1. Summary of patients’ information.a
A CNN model was trained and validated to identify black-stained SLNs in the ex vivo and in vivo PA images, which shows great potential for automatically detecting SLNs. The network architecture, shown in Fig. 3, divided the input image into an S × S grid, and each grid cell was responsible for predicting bounding boxes and their corresponding confidence scores. To combine multi-scale information for detecting SLNs of various sizes, we used a three-layer pyramid top-down structure (P3, P4, P5) to maximize the combination of deep and shallow semantic information. Especially, in the top-down pyramid, we used the nearest neighbor interpolation for up-sampling the feature map (Y3, Y4, Y5) from the deep layer and concatenating it with the feature map from the shallow layer, which fused the semantic and detailed information. To reduce the model size, the Ghost Module was adopted in each branch of the multi-scale pyramid structure. The network consisted of 24 convolution (Conv) layers and 2 fully connected (FC) layers, with some Conv layers and FC layers constructing ensembles of residual modules. The object detection model’s backbone (GhostDarknet53) was pretrained on the COCO dataset (200,000 datasets and 80 class categories), and then the fully connected layer and the last pooling layer were deleted. Our model was trained by freezing the feature extraction backbone and fine-tuning the network weight parameters in the object detection and classification tasks, based on our own dataset. The object detection network can process images in real time at 45 fps.
Fig. 3. The convolutional neural network (CNN) architecture for SLN detection in PA images. Each Ghost Module (GM) applies a series of low-cost linear transformations to a high number of ghost feature maps, which can greatly reduce the amount of computation. Ghost Bottleneck (GB) is designed by replacing the conventional convolution layers in the bottleneck of DarkNet53 with GM. Conv: Convolution layers. P3, P4, P5: pyramid top-down structure. Y3, Y4, Y5: up-sampling the feature map.
2.7 Ex vivo imaging protocol
The SLNs stained with CNPs were dissected and placed in agarose phantoms for ex vivo imaging to verify the image features observed in vivo corresponded to the SLNs dissected in the surgery. To ensure the LNs were roughly at the acoustic focus (2 cm below the linear array surface), they were placed at a depth of 2 cm inside the phantom, to guarantee that they were roughly at the acoustic focus (2 cm below the linear array’s surface). Medical ultrasound couplant was used as the coupling medium, and only PA imaging was performed. During the imaging process of each ex vivo sample, the probe was linearly translated across the entire phantom, and a video of the entire scanning process was recorded as a 3D data cube for further analysis.
2.8 Statistical analyses
Our statistical analyses were based on comparing the PA/US imaging results with the intraoperative visual findings. The sensitivity and specificity of the imaging results were calculated as the number of true-positive and true-negative decisions divided by the total number of positive and negative cases, respectively. To determine the overall accuracy, we divided the number of correct detections and classifications by the total number of cases. All statistical analyses were performed using the Statistics Toolbox of MATLAB R2021a (MathWorks).
3. Results and discussion
3.1 Patient characteristics
Some information about the patients and the pathological results are disclosed in Table 1. In total, 11 patients with BCs were evaluated. The average participant age was 63.8 ± 24.8 years, while the average participant BMI was 27.4 ± 8.3. All participants were scheduled for ALND or SLNB. The average PA/US scanning time ranged from 5 to 10 minutes. The LNs of 2 participants (#8 and #10) were not stained successfully as they had undergone SLNB, which resulted in lymphatic obstruction. In these two cases, we defined the abnormally enlarged and palpationally hard LNs as SLNs. We defined the CNP-stained LNs as SLNs in other situations. In Table 1, we provide explanations for the various cases with failed SLN detection.
3.2 In vivo imaging and ex vivo validation of SLNs
During in vivo imaging, PA can distinguish SLNs from the downstream nodes based on the strong PA signals from the CNPs. An SLN appeared as a double-horizontal-line pattern, stemming from the PA signals of the top and bottom surfaces of the SLN. In practice, a CNP-stained-SLN may look similar to a blood vessel under PAI. We found, however, that the SLN can be distinguished from blood vessels for two reasons: (a) the strong light absorption of CNPs makes the SLN appear sharper, and (b) an SLN showed a distinctive pattern of feature-evolvement when the transducer was scanned across the SLN. The scan result of Patient 4 (Visualization 1) clearly illustrates this observation.
Once an SLN was located, surgeons marked the skin above it with a marker pen. The process was repeated until the entire region of interest was scanned. After the SLNs were dissected, the PA/US system was used again to image the ex vivo samples. Figure 4(a–c) show an example of the US, PA, and merged PA/US images of a representative SLN, captured in vivo. In the PA image, the unique double-line pattern of a stained SLN distinguished it from blood vessels (e.g., the two bright spots above the SLN). Figure 4(d) illustrates in-vivo SLN detection by machine learning during the imaging process. The surgeon can, based on the US image, further verify that the detected object was an LN. Figure 4(e–h) show how we validated the in vivo imaging results. Figure 4(e) and (f) show the PA images of an SLN before and after it was removed from the patient, respectively. To confirm the PA image features observed during the scan came from the excised LN, ex vivo imaging was performed. The photos of the SLN before and after it was cut are shown in Fig. 4(g) and (h).
Fig. 4. (a–d). Representative images of in vivo PA/US imaging of SLNs for a 69-year-old woman with breast cancer (Patient 1 in Table 1). (a) US image. (b) PA image (SLN circled). (c) Co-registered PA/US image (SLN circled). (d) SLN is automatically detected by the CNN network in the PA image (scale bars: 10 mm). (e–f) Validation of the detected SLNs by removing the SLN and ex vivo PA imaging: (e) In vivo PA image of the SLN of Patient 1, and (f) Ex vivo PA image of the resected SLN in an agarose phantom. (g) A photo of the stained LN (arrow) during surgery. The incision was guided by the PA/US imaging system. (h) A photo of the ex vivo specimen (arrow points to the black-stained LN, scale bars: 2 mm). SLN: sentinel lymph node.
Figure 5 illustrates the imaging process. The operator typically scanned the probe from the injection site to the axillary region. Once a double-line pattern was found in the PA image, the probe was scanned near the suspicious area transversely and vertically to acquire more information in 3D dimensions. Meanwhile, the operator checked the US image for cross-validation.
Fig. 5. Example of the imaging process. (a) Illustration of the scanning path. POS: positions. In this example, the probe followed the path from POS 1 to POS 6. (b) Columns are the in vivo images obtained at the labeled positions. The first, second, and third rows correspond to PA, US, and merged PA/US images, respectively (scale bar: 10 mm). PA: photoacoustic. SLN: sentinel lymph node.
Figure 6 shows four SLNs stained to various degrees and the corresponding ex vivo PA images. Figure 6(a) shows a completely stained LN, while (b) and (c) show LNs with decreasing degrees of stain. Figure 6(d) shows the case of an LN without staining. We collectively normalized Figs. 6(a2)–(d2) using a single constant. Notably, the PA image feature in this special case comes from the contrast between blood and agar, yet this LN was invisible during the in vivo scan since the background tissue with the same level of blood content eliminated the contrast.
Fig. 6. Specimens of removed LNs stained to various degrees and the corresponding PA images in phantoms. From (a) to (d), the degree of black stain decreases. Rows 1 and 2 show photos of the LNs and their PA images, respectively. The signal in (d2) comes from the contrast between blood and agar, which did not exist in the in vivo case (scale bar: 2 mm).
3.3 Sensitivity and specificity of the current study
The in vivo imaging results of 11 patients and 23 surgically detected SLNs were analyzed. Predictions for SLNs made by the PA/US system were classified as positive or negative based on the PA/US images. Correctness of the predictions made by the surgeons during SLN dissections were classified as true or false based on the surgically confirmed SLN count (Table 2). Four SLNs were classified as negative by both PA/US and surgery (100% true negative) and nine as positive by both PA/US and surgery (52.6% true positive). Overall, this analysis yielded an in vivo sensitivity of 52.6% and a specificity of 100%. The current system and imaging procedure had a high false-negative rate of 47.4%, likely due to the influences of hemorrhage, tattoo pigment, and depth limitation among some patients. Such complications can also explain the difference between the in vivo and ex vivo imaging results of some stained SLNs.
Table 2. Surgical findings and the performance of the proposed imaging method in the current study. Predictions made by the PA/US images were evaluated based on the number of surgically removed SLNs to compute the sensitivity and specificity of SLN mapping. The statistics were based on 23 SLNs in 11 patients. a
We provide some representative imaging results generated during the study in Supplementary Figs. 3–8. All reconstructed PA data were normalized between 0 and 1 after Hilbert-transforming the image and then taking the absolute value of the transformed image. During image co-registration, we removed the superficial skin signals and performed image segmentations based on thresholding.
3.4 Automatic SLN detection
In this study, we utilized the mean average precision to evaluate the proposed methods (Supplementary Fig. 2(b, c)). A video showing the dynamic SLN detection procedure is provided in Visualization 2 and Visualization 3. The detection results show that our network performed well.
In this object detection task, a lightweight SLN detection model based on YOLOv3 was proposed to automatically detect and locate SLNs of different sizes in real-time. However, the dataset is still relatively small with limited features, so we will expand future datasets to make the model more robust and stabler.
4. Discussion
The present SLNB procedure suffers from the limited accessibility of the nuclide agent and reduced therapeutic efficacy of single-contrast-agent mapping. A simpler, non-radiative SLN mapping technique, with comparable or higher positioning accuracy, has the potential to replace the nuclide tracer and supplement the current clinical procedure for the axillary staging of BC—with the benefit of the complete elimination of radiation hazards, high accessibility, and reduced invasiveness. Here, we developed a dual-modality PA/US imaging system for the mapping of SLNs and evaluated its potential to offer preoperative surgical guidance for SLNBs. The new technology relies on the image features of CNP-stained SLNs to provide the mapping information, thus allowing single-wavelength operation. Based on our results, CNPs have a long retaining time in SLNs, with negligible staining of the nearby LVs. These properties, together with a properly chosen period between CNP injection and PA/US imaging (20–30 minutes in this study), improve the specificity of SLN detection (100% specificity). A deep neural network was developed based on the imaging data and could automatically identify SLNs in a post-processing manner. Once the network accuracy is confirmed by more imaging data in the future, we expect the network to help surgeons detect SLNs in real time. Using the new technology, we pinpointed the SLNs preoperatively in the PA/US images, before the surgical removal of the SLNs. We then performed PA imaging of the ex vivo specimen to verify feature consistency and performed statistical analysis to estimate the sensitivity and specificity of the current study.
Compared with other SLN identification methods [23–25]—such as Geiger counter, MRI, and single photon emission computed tomography (SPECT)/CT—PAI is the only imaging modality that allows surgeons to find stained lymph nodes both preoperatively and intraoperatively. In addition, our method is non-radioactive and relatively economical.
We identified the following limitations to be addressed in the future. Currently, the method has low sensitivity (52.6%). Even though previous works using PAI to detect SLNs have not counted sensitivity, some studies can be compared with our work [16,26,27]. The low sensitivity also creates an obstacle to applying this method in clinical applications. On the one hand, the starting point of our work was to verify the localization effect of PAI on SLNs. We were cautious in calculating the sensitivity, and we only recorded the confirmed results as SLNs. On the other hand, the imaging depth of the system also limited the method’s sensitivity. The deepest CNP-stained SLN was found at a depth of 3.5 cm, and we failed to detect a stained SLN at a greater depth (roughly 4.5 cm). Thus, the maximum imaging depth of CNP-stained SLNs is roughly 3.5 cm in the current system. In the presence of internal hemorrhage, tattoo pigment, and dark skin color, the PA imaging depth will decrease. Ultimately, we are optimistic about this technique eventually being suitable for clinical applications; both the imaging system and the imaging procedure have much room for improvement. To extend the imaging depth in the future, we plan to further improve the sensitivity of the sensing elements and simultaneously increase the focal depth of the probe. We also plan to better differentiate stained LNs from blood vessels by using a planar array transducer for 3D imaging. Meanwhile, we will employ dual-wavelength excitation to further increase the differentiation capability by taking advantage of the spectral-domain information.
In addition, CNPs have been clinically used in China, making the regulatory barrier low for the proposed technology. Once successfully deployed, the new approach offers benefits over the current standard of care in reducing the morbidity and invasiveness of the SLNB procedure, eliminating the need for radioactive isotopes, and bringing down the cost of the overall equipment and procedure.
Funding
Initiative Scientific Research Program, Institute for Intelligent Healthcare, Tsinghua University; Tsinghua-Foshan Institute of Advanced Manufacturing; National Natural Science Foundation of China (61735016, 61971265).
Acknowledgment
We thank Zhe Zhao at the Department of Orthopaedics, Beijing Tsinghua Changgung Hospital, Tsinghua University, for engaging in helpful discussions. We also thank Qi Qin and Wuyang Ji at the Department of General Surgery of Beijing Tsinghua Changgung Hospital, Tsinghua University, for obtaining patient consent and designing the clinical study.
Disclosures
Regarding activities related to the present article, L.G., H.D., Y.B., X.W., Y.T. J.G., and B.L. have no relevant relationships to disclose. C.M. has a financial interest in TsingPAI Technology Co., Ltd., which provided the data acquisition unit (DAQ) used in this work.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. World Health Organization: Regional Office for Europe, World Cancer Report: Cancer Research for Cancer Development, (IARC, 2020), 382–393
2. D. Krag, S. Weaver, T. Ashikaga, F. Moffat, V. S. Klimberg, C. Shriver, S. Feldman, R. Kusminsky, M. Gadd, J. Kuhn, S. Harlow, and P. Beitsch, “The sentinel node in breast cancer – a multicenter validation study,” N Engl. J Med. 339(14), 941–946 (1998). [CrossRef]
3. D. N. Krag, S. J. Anderson, T. B. Julian, A. M. Brown, S. P. Harlow, J. P. Costantino, T. Ashikaga, D. L. Weaver, E. P. Mamounas, L. M. Jalovec, T. G. Frazier, R. D. Noyes, A. Robidoux, H. McScarth, and N. Wolmarket, “Sentinel-lymph-node resection compared with conventional axillary-lymph-node dissection in clinically node-negative patients with breast cancer: overall survival findings from the NSABP B-32 randomised Phase 3 trial,” Lancet Oncol. 11(10), 927–933 (2010). [CrossRef]
4. A. S. Caudle, W. T. Yang, S Krishnamurthy, E. A. Mittendorf, D. M. Black, M. Z. Gilcrease, I Bedrosian, B. P. Hobbs, S. M. DeSnyder, R. F. Hwang, B. E. Adrada, S. F. Shaitelman, M. Chavez-MacGregor, B. D. Smith, R. P. Candelaria, G. V. Babiera, B. E. Dogan, L Santiago, K. K. Hunt, and H. M. Kuerer, “Improved axillary evaluation following neoadjuvant therapy for patients with node-positive breast cancer using selective evaluation of clipped nodes: implementation of targeted axillary dissection,” J. Clin. Oncol. 34(10), 1072–1078 (2016). [CrossRef]
5. L. Lin and L. V. Wang, “The emerging role of photoacoustic imaging in clinical oncology,” Nat. Rev. Clin. Oncol. 19(6), 365–384 (2022). [CrossRef]
6. S. Zackrisson, S. M. W. Y. van de Ven, and S. S. Gambhir, “Light in and sound out: emerging translational strategies for photoacoustic imaging,” Cancer Res. 74(4), 979–1004 (2014). [CrossRef]
7. E. Zheng, H. Zhang, S. Goswami, I. E. Kabir, M. M. Doyley, and J. Xia, “Second-generation dual scan mammoscope with photoacoustic, ultrasound, and elastographic imaging capabilities,” Front Oncol. 11, 779071 (2021). [CrossRef]
8. L. Lin, P. Hu, J. Shi, C. M. Appleton, K. Maslov, L. Li, R. Zhang, and L. V. Wang, “Single-breath-hold photoacoustic computed tomography of the breast,” Nat. Commun. 9(1), 2352 (2018). [CrossRef]
9. C. Kim, T. N. Erpelding, K. Maslov, L. Jankovic, W. J. Akers, L. Song, S. Achilefu, J. A. Margenthaler, M. D. Pashley, and L. V. Wang, “Handheld array-based photoacoustic probe for guiding needle biopsy of sentinel lymph nodes,” J. Biomed. Opt. 15(04), 1 (2010). [CrossRef]
10. C. Kim, T. N. Erpelding, L. Jankovic, M. D. Pashley, and L. V. Wang, “Deeply penetrating in vivo photoacoustic imaging using a clinical ultrasound array system,” Biomed. Opt. Express 1(1), 278–284 (2010). [CrossRef]
11. J. Kim, E. I. Galanzha, E. V. Shashkov, H. Moon, and V. P. Zharov, “Golden carbon nanotubes as multimodal photoacoustic and photothermal high-contrast molecular agents,” Nat. Nanotechnol. 4(10), 688–694 (2009). [CrossRef]
12. L. Song, C. Kim, K. Maslov, K. K. Shung, and L. V. Wang, “High-speed dynamic 3D photoacoustic imaging of sentinel lymph node in a murine model using an ultrasound array,” Med. Phys. 36(8), 3724–3729 (2009). [CrossRef]
13. I. Stoffels, S. Morscher, I. Helfrich, U. Hillen, U. Leyh, N. C. Burton, T. C. P. Sardella, J. Claussen, T. D. Poeppel, H. S. Bachmann, A. Roesch, K. Griewank, D. Schadendorf, and M. Gunzer, “Klode J. metastatic status of sentinel lymph nodes in melanoma determined noninvasively with multispectral optoacoustic imaging,” Sci. Transl. Med. 7(317), 317ra199 (2015). [CrossRef]
14. C. Kim, K. H. Song, F. Gao, and L. V. Wang, “Sentinel lymph nodes and lymphatic vessels: noninvasive dual-modality in vivo mapping by using indocyanine green in rats–volumetric spectroscopic photoacoustic imaging and planar fluorescence imaging,” Radiology 255(2), 442–450 (2010). [CrossRef]
15. S. Liu, H. Wang, C. Zhang, J. Dong, S. Liu, R. Xu, and C. Tian, “In vivo photoacoustic sentinel lymph node imaging using clinically-approved carbon nanoparticles,” IEEE Trans. Biomed. Eng. 67(7), 1 (2019). [CrossRef]
16. A. Garcia-Uribe, T. N. Erpelding, A. Krumholz, H. Ke, K. Maslov, C. Appleton, J. A. Margenthaler, and L. V. Wang, “Dual-modality photoacoustic and ultrasound imaging system for noninvasive sentinel lymph node detection in patients with breast cancer,” Sci. Rep. 5(1), 15748 (2015). [CrossRef]
17. W. J. Gradishar, M. S. Moran, J. Abraham, R. Aft, D. Agnese, K. H. Allison, B. Anderson, H. J. Burstein, H. Chew, C. Dang, A. D. Elias, S. H. Giordano, M. P. Goetz, L. J. Goldstein, S. A. Hurvitz, S. J. Isakoff, R. C. Jankowitz, S. H. Javid, J. Krishnamurthy, M. Leitch, J. Lyons, J. Mortimer, S. A. Patel, L. J. Pierce, L. H. Rosenberger, H. S. Rugo, A. Sitapati, K. L. Smith, M. L. Smith, H. Soliman, E. M. Stringer-Reasor, M. L. Telli, J. H. Ward, K. B. Wisinski, J. S. Young, J. Burns, and Kumar R. Breast Cancer, “Version 3.2022, NCCN Clinical Practice Guidelines in Oncology,” J. Natl. Compr. Canc. Netw. 20(6), 691–722 (2022). [CrossRef]
18. H. Zuo, M. Cui, X. Wang, and C. Ma, “Spectral crosstalk in photoacoustic computed tomography,” Photoacoustics 26(26), 100356 (2022). [CrossRef]
19. X Wu, Q Lin, G Chen, J Lu, Y Zeng, X Chen, and J Yan, “Sentinel lymph node detection using carbon nanoparticles in patients with early breast cancer,” PLoS One 10(8), 0135714 (2015). [CrossRef]
20. A. Taruttis, A. C. Timmermans, P. C. Wouters, M. Kacprowicz, G. M. van Dam, and V. Ntziachristos, “Optoacoustic imaging of human vasculature: feasibility by using a handheld probe,” Radiology 281(1), 256–263 (2016). [CrossRef]
21. G. Li, L. Li, L. Zhu, J. Xia, and L. V. Wang, “Multiview Hilbert transformation for full-view photoacoustic computed tomography using a linear array,” J. Biomed. Opt. 20(06), 1 (2015). [CrossRef]
22. L. Li, L. Zhu, Y. Shen, and L. V. Wang, “Multiview Hilbert transformation in full-ring transducer array-based photoacoustic computed tomography,” J. Biomed. Opt. 22(7), 076017 (2017). [CrossRef]
23. P. Li and D. Sun, “Advanced diagnostic imaging of sentinel lymph node in early stage breast cancer,” J. Clin. Ultrasound 50(3), 415–421 (2022). [CrossRef]
24. J. J. Zhang, W. C. Zhang, C. X. An, X. M. Li, and L. Ma, “Comparative research on 99mTc-Rituximab and 99mTc-sulfur colloid in sentinel lymph node imaging of breast cancer,” BMC Cancer 19(1), 956 (2019). [CrossRef]
25. S. Sreedhar, J. Maloney, and S. Hudson, “Introducing SentiMag in a rural setting: a 5-year experience,” ANZ J. Surg. 91(11), 2404–2410 (2021). [CrossRef]
26. C. Kim, T. N. Erpelding, L. Jankovic, and L. V. Wang, “Performance benchmarks of an array-based hand-held photoacoustic probe adapted from a clinical ultrasound system for non-invasive sentinel lymph node imaging,” Phil. Trans. R. Soc. A. 369(1955), 4644–4650 (2011). [CrossRef]
27. T. N. Erpelding, C. Kim, M. Pramanik, L. Jankovic, K. Maslov, Z. Guo, J. A. Margenthaler, M. D. Pashley, and L. V. Wang, “Sentinel lymph nodes in the rat: noninvasive photoacoustic and US imaging with a clinical US system,” Radiology 256(1), 102–110 (2010). [CrossRef]
