Multimodal photoacoustic/ultrasonic imaging system: a promising imaging method for the evaluation of disease activity in rheumatoid arthritis

Objectives We aimed to assess the clinical value of multimodal photoacoustic/ultrasound (PA/US) articular imaging scores, a novel imaging method which can reflect the micro-vessels and oxygenation level of inflamed joints of rheumatoid arthritis (RA). Methods Seven small joints were examined by the PA/US imaging system. A 0–3 scoring system was used to semi-quantify the PA and power-Doppler (PD) signals, and the sums of PA and PD scores (PA-sum and PD-sum scores) of the seven joints were calculated. The relative oxygen saturation (SO2) values of the inflamed joints were measured and classified into 3 PA+SO2 patterns. The correlations between the PA/US imaging scores and the disease activity scores were assessed. Results Thirty-one patients of RA and a total of 217 joints were examined using the PA/US system. The PA-sum had high positive correlations with the standard clinical scores of RA (DAS28 [ESR] ρ = 0.754, DAS28 [CRP] ρ = 0.796, SDAI ρ = 0.836, CDAI ρ = 0.837, p < 0.001), which were superior to the PD-sum (DAS28 [ESR] ρ = 0.651, DAS28 [CRP] ρ = 0.676, SDAI ρ = 0.716, CDAI ρ = 0.709, p < 0.001). For the patients with high PA-sum scores, significant differences between hypoxia and hyperoxia were identified in pain visual analog score (p = 0.020) and patient’s global assessment (p = 0.026). The PA+SO2 patterns presented moderate and high correlation with PGA (ρ = 0.477, p = 0.0077) and VAS pain score (ρ = 0.717, p < 0.001). Conclusion The PA scores have significant correlations with standard clinical scores for RA, and the PA+SO2 patterns are also related with clinical scores that reflect pain severity. PA may have clinical potential in evaluating RA. Key Points • Multimodal photoacoustic/ultrasound imaging is a novel method to assess micro-vessels and oxygenation of local lesions. • Significant correlations between multimodal imaging parameters and clinical scores of RA patients were verified. • The multimodal PA/US system can provide objective imaging parameters, including PA scores of micro-vessels and relative SO2 value, as a supplementary to disease activity evaluation. Supplementary Information The online version of this article (10.1007/s00330-020-07353-z) contains supplementary material, which is available to authorized users.


Introduction
Rheumatoid arthritis (RA) is a chronic, autoimmune disease marked by symmetrical arthritis that usually involves synovial joints and causes further erosive changes and joint deformities [1,2]. At present, the widely recognized treatment strategy for RA is "treat-to-target," in which RA patients are treated and assessed strictly in periods of every 1-3 months until complete remission (CR) or low disease activity (LDA) is achieved [3]. Nevertheless, disease relapses occasionally occur in patients tapering or discontinuing disease-modifying anti-rheumatic drugs (DMARDs), even when they have achieved the treatment goals. This suggests that a more sensitive and objective method is needed to detect subclinical synovitis [4].
Noninvasive imaging methods, including ultrasound (US) and magnetic resonance imaging, are commonly used for the evaluation of joint inflammation [5]. As a convenient and bedside-accessible imaging tool, the role of US has been well established in diagnosing and assessing RA with the development of high-frequency gray-scale ultrasonic probes and power-Doppler ultrasound (PDUS) imaging techniques [6,7]. Researchers also validated that US played a substantial part in optimizing management of RA patients, as an addition to clinical assessment [8]. Notably, the use of US in RA is still questioned by some researchers [9]. Several studies demonstrated that inflamed synovium could not be completely observed by US, and some PDUS-negative patients might have obvious joint inflammation, which might be ascribed to the insensitivity of PDUS in detecting micro-vessels of the inflammatory lesions [10]. Therefore, new approaches that can enhance the diagnostic performance of US in visualizing arthritis are in great demand.
Photoacoustic imaging (PAI), a hot spot of medical imaging, has broad clinical applications in identifying a variety of different diseases, including thyroid nodules [11], breast cancers [12][13][14], and inflammatory bowel diseases [15]. The principle of PAI is generating thermal expansion of local tissues induced by laser irradiation, which in turn produces ultrasonic waves that can be detected by common ultrasonic probes. PAI combines the merits of optical imaging and US imaging and reflects optical characteristics of local tissue, including hemoglobin and melanin content, with a deep penetration depth [16,17]. Moreover, it is applicable to integrate the PA modality into a well-established ultrasonic technique to facilitate dualmodality PA/US imaging after image post-processing, because of the acoustic features of PA signals. A dualmodality PA/US system based on a commercial ultrasonic platform and a handheld probe can be a good way to promote the clinical translation of PAI. By detecting the hemoglobin content in local tissue, PAI can visualize micro-vessels in the hyperplastic synovium of RA-involved joints. Oxygenation can be assessed by calculating the signals of oxygenated and deoxygenated hemoglobin in dual-wavelength PAI and is an important functional indicator of inflamed tissues.
The feasibility of PAI in detecting minor inflammatory lesions in RA joints has been validated by previous studies, in which dual-modality PA/US imaging systems were used [18][19][20][21][22][23][24][25]. In the study by Wang et al, patients with synovitis were recruited for PA scanning of their metacarpophalangeal joints (MCPs) with a dual-modality PA/US system, and significant hyperemia and hypoxia were found in the inflamed synovial tissues [22]. According to van den Berg et al, increased PA signals were detected by a portable PA/US imaging tool in 10 RA patients with clinically evident synovitis and had a good correlation with PDUS [23]. However, the sample sizes of the previous studies were relatively small, and a comprehensive comparison between PA/US imaging and clinical evaluation methods of RA has not been conducted. To explore further the clinical applications and value of PAI in RA, more studies focusing on the clinical evaluation and validation of PAI are necessary.
In this study, we performed imaging the joints of RA patients with a multimodal PA/US imaging system that combined a light device and an ultrasonic transducer. We aimed to assess the correlation of multimodal PA/US articular imaging scores and relative oxygen saturation (SO 2 ) values of lesions with different disease activity measurements of RA and to evaluate the potential value of multimodality PA/US imaging of RA in clinical practice.

Materials and methods
This study was designed as a cross-sectional study. All the procedures in this study were approved by the Institutional Review Board of Peking Union Medical College Hospital (PUMCH) (approval number: JS-1923). Written informed consent was received from all recruited patients.

Patient recruitment
RA patients aged older than 18 years old were recruited from the outpatient clinic of rheumatology at PUMCH from December 2018 to October 2019. The patients were diagnosed with RA according to the 2010 American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) classification criteria. Patients complicated with other types of arthritis, including osteoarthritis (OA), psoriatic arthritis (ReA), and gouty arthritis, were excluded.

Multimodal PA/US imaging system
The schematic of the dual-modality imaging system is presented in Fig. 1a. Commercially available ultrasonic equipment (Resona 7, Mindray Bio-Medical Electronics Co., Ltd.), equipped with an OPO tuneable laser (SpitLight 600-OPO, InnoLas laser GmbH) and a handheld linear probe (L9-3U, Mindray Bio-Medical Electronics Co., Ltd.) (Fig. 1b) centered at 5.8 MHz, was utilized as the fundamental platform of this novel imaging system. Gray-scale US (GSUS) and PDUS could be displayed simultaneously with the PA imaging mode in real time. For PAI, the wavelengths are 750 nm and 830 nm, at which the deoxygenated hemoglobin and oxygenated hemoglobin could reach peak absorption, respectively. Synchronized gray-scale US and 830 nm PAI were provided by the system at a frame of 10 Hz. For power-Doppler US, the following imaging settings were adopted: pulse repetition frequency of 600-1000 Hz, wall filter of 50-100 Hz, maximum gain of 85-90%, scale of 3 cm/s, a rectangle sampling box with no angulation. Detailed information on the imaging system is provided in Supplementary Data S1.

Imaging protocols
Examination procedure A seven-joint ultrasound scoring system (US7) proposed by Backhaus et al [26] was utilized as the reference to choose joints for the multimodal PA/US examinations. The 2nd metacarpophalangeal joint (MCP2), 3rd metacarpophalangeal joint (MCP3), 2nd proximal interphalangeal joint (PIP2), 3rd proximal interphalangeal joint (PIP3), 2nd metatarsophalangeal joint (MTP2), 5th metatarsophalangeal joint (MTP5), and wrist of the clinically dominant side were chosen for multimodal imaging. GSUS and PDUS imaging of the joints, followed by real-time PA/US imaging, were carried out by a US operator who had 1 year of musculoskeletal US experience and received 1 month of training on operating the system. Detailed explanations of the imaging procedures are presented in Supplementary Data S2, and the training project of the participated radiologists in Supplementary Data S4.

Semi-quantitative PDUS and PAI scoring
The semi-quantitative scoring system scored on a scale of 0-3 proposed by Szkudlarek et al [27] for PDUS was utilized as the reference to grade PD and PA images in the study. The global sums of the PDUS (PD-sum, 0-21) and PA (PA-sum, 0-21) scores of the seven joints were calculated for statistical analysis. The images were assessed by two radiologists with 2 years of experience in musculoskeletal US who were blinded to the patients' information and clinical manifestations of the examined joints. When discrepancies occurred, reconfirmations of the images were conducted until a consensus was reached. After 1 month, the two raters scored the PD and PA images again. The intra-observer of the two times of scoring and inter-observer agreement of the two raters in the first time of scoring were evaluated. The scoring method is explained in detail in Supplementary Data S3.

Measurement of SO 2 and PA+SO 2 patterns
By calculating the ratio of the pixels of the PA signals in the target areas at wavelengths of 750 nm and 830 nm, the relative SO 2 values of the inflamed region were determined. The patients were divided into hyperoxic and hypoxic subgroups according to the distribution of the relative SO 2 values. After calculating the SO 2 values, the SO 2 value and the PA-sum score were incorporated as a new index for RA patients, named the "PA+SO 2 pattern." The patients with PA-sum scores < 3 were considered to have minimal PA signals, and those with PA scores ≥ 3 were considered to have evident PA signals. According to the PA-sum scores and SO 2 values, all the patients were divided into 3 groups: pattern 1, no or minimal PA signals; pattern 2, evident PA signals and hyperoxia; and pattern 3, evident PA signals and hypoxia. The associations of the PA+SO 2 patterns and clinical scores were also evaluated. Detailed explanations of the identification of PA+ SO 2 patterns are presented in Supplementary Data S5.

Clinical assessment
The relevant clinical information was recorded, including age, sex, disease duration from onset and from diagnosis confirmation, duration of morning stiffness, symptoms, and current medications. The laboratory parameters of the patients were collected, such as the erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), anti-cyclic citrullinated peptide (anti-CCP) antibodies, and rheumatoid factor (RF). For each patient, 28 joints (bilateral PIPs, MCPs, wrists, elbows, shoulders, and knees) were clinically assessed for swelling and tenderness (the swollen joint count (SJC) and the tender joint count (TJC)) by a rheumatologist with 18 years of experience in clinical rheumatology. The visual analog scale (VAS) scores of joint pain and the patient global activity (PGA), evaluator global activity (EGA), disease activity score in 28 joints (DAS28), clinical disease activity index (CDAI), and simplified disease activity index (SDAI) values were also obtained.

Results
Multimodal imaging scores, laboratory data, and clinical scores A total of 31 patients with RA were recruited for this study, including 24 females and 7 males aged 25-71 years (mean age 51.8, median age 51). A total of 217 joints were examined using the PA/US system. The detailed clinical information of the patients is summarized in Table 1. For the intra-observer agreement of rater 1, κ value was 0.87 (0.76-0.96) for two times of PD scoring (p = 0.041), and 0.88 (0.79-0.96) for two times of PA scoring (p = 0.043), respectively. For the intraobserver agreement of rater 2, κ value was 0.92 (0.84-1.00) for PD scoring (p = 0.043) and 0.88 (0.81-0.97) for PA scoring (p = 0.045), respectively. The intra-observer agreement for both raters in scoring the PD and PA images was very good. And the inter-observer agreement of the two radiologists in scoring the images was also very good (κ = 0. A total of 217 joints were examined using the PA/US system, and the numbers of each PA grading of the corresponding PD grading (0-3) are shown in Table 2. Among all those small joints, a total of 16 joints were divided into the highest level of both PA and PD. There were 21 joints scored as grade 1 in PA, showing no signals in PD. The imaging interpretation of the PD and PA results is shown in Figs. 2, 3, and 4.
The correlation between PA/PD scores and RA disease activity measurements The correlations among the PA-sum scores, PD-sum scores, and RA disease activity measurements are listed in Table 3.   ESR erythrocyte sedimentation rate, CRP C-reactive protein, DAS28 disease activity score in 28 joints, SJC swollen joint count, TJC tender joint count, VAS visual analog scale, PGA patient global activity, EGA evaluator global activity, SDAI simplified disease activity index, CDAI clinical disease activity index, PD-sum sum of power-Doppler scores, PA-sum sum of photoacoustic scores  Fig. S1). And 10 patients with no or minimal PA signals were classified as pattern 1, 12 patients with evident PA signals and hyperoxia classified as pattern 2, and 9 patients with evident PA signals and hypoxia  (Fig. 6).

Discussion
In this study, a multimodal PA/US imaging system was utilized to evaluate the small joints of RA patients with different disease activities. The PA parameters, including PA-sum scores and PA+SO 2 patterns, correlated well with the clinical scores, suggesting the feasibility of PAI in assessing the disease activity of RA. In addition to conventional gray-scale US and PDUS, PAI may be utilized as a new modality affiliated with US systems to provide new imaging information to assist in assessing the disease activity of RA. We performed an overall imaging evaluation of small joints in patients with RA, including the MCP, PIP, MTP, and wrist, using a 0-3 PDUS grading scale and the simplified US7 scoring system as a reference [26,31]. The PA-sum scores of the seven joints were significantly higher than the PD-sum scores, indicating that PAI was more sensitive to small vessels in the Fig. 4 Example of PD 2, PA 3. The PA/US images of the dorsal aspect of the wrist of a 45-yearold female with a 3-year history of RA. Significant signals were presented in the PD mode, scored as 2. In the PA mode, abundant optical signals were visualized under both of the two wavelengths, representing hyperemia of the inflamed lesion (the region marked with the yellow line). We scored the PA results of the wrist as 1. In the SO 2 part, the signals were demonstrated in the pseudo-color of red, with a relative SO 2 value of 97.33%, classified as hyperoxia  [32][33][34][35][36].
Compared with the PD-sum score, the PA-sum score presented higher correlation coefficient, which had high positive correlation with the standard clinical disease activity scores (ρ > 0.70 for DAS28, SDAI, and CDAI). Therefore, PAI may be able to accurately reflect the disease activity of the individual patients, which is especially useful for active lesions that are difficult to recognize by PDUS. Hypoxia has been believed to be a hallmark of inflammatory diseases in this decade, and RA is no exception [37]. In vivo partial pressure of oxygen (pO 2 ) measurements of RA patients have also been performed using arthroscopy and ion electrodes by Ursula Fearo et al, but the method was relatively time-consuming and invasive [38]. In this study, we classified the patients into hyperoxia and hypoxia subtypes using the relative SO 2 values, identified from pixels of PA signals. We found that, among the joints with higher levels of PA signals, the hypoxic individuals were more likely to have higher VAS and PGA scores than the hyperoxic patients. A high positive correlation was validated between the combined PA+SO 2 patterns and pain VAS, and moderate correlation of PA+SO 2 patterns and PGA scores, indicating that the oxygenation status might be related with the pain severity of the joints. The relative SO 2 value could be used as a supplementary to assessing joint symptoms. No significant differences were detected between the PA+SO 2 pattern and the other indexes. This result was not strong enough to establish the role of oxygenation in evaluating disease activity, which might be caused by the limited cases enrolled in this preliminary study. Studies about measuring oxygenation using PAI with larger sample size are required for further verifying the value of dual-wavelength PA in evaluating RA.
The multimodal PA/US system developed by our team integrated the PA modality into a high-end commercial US unit, making it possible to display real-time multimodal imaging and provide versatile imaging information for target lesions. A handheld probe combining the emission of lasers and ultrasonic waves and reception of US and PA signals was also equipped with the system, making the system easy for radiologists to operate. In addition, the PA and SO 2 scoring system in our study was created using built-in software, with reference to the widely approved 0-3 PD scores and US7 ESR erythrocyte sedimentation rate, CRP C-reactive protein, DAS28 disease activity score in 28 joints, SJC swollen joint count, TJC tender joint count, VAS visual analog scale, PGA patient global activity, EGA evaluator global activity, SDAI simplified disease activity index, CDAI clinical disease activity index, SO 2 oxygen saturation Fig. 5 The fitting curves and scatter plots of imaging scores (PD-sum and PA-sum, Y-axis) and clinical scores (DAS28ESR, DAS28CRP, CDAI, SDAI for the X-axis, respectively). An upward straight line reflecting the correlation between the imaging parameter and the clinical score was observed in each figure, respectively, indicating significant correlations of PA scores and standard clinical scores (DAS28, SDAI, CDAI) scoring systems, thus providing a concise and systematic evaluation of RA patients that could be more convenient and acceptable for clinicians. There still exist several limitations in our study. First, the sample of RA patients was still small, and more patients must be recruited for further validation. Further clinical studies with larger sample sizes are also expected to explore the value of SO 2 in the diagnosis of RA. Second, this study is a preliminary cross-sectional observational study of the multimodal PA/US imaging system, and whether these PA measures can predict treatment response and risk of RA relapse remains to be explored by a prospective cohort study in the future.

Conclusions
The correlations between PA scores of micro-vessels and standard clinical scores for RA were identified, and relative SO 2 was also related with clinical scores that reflect pain severity. The multimodal PA/US imaging system provided comprehensive imaging parameters and might have great potential in the evaluation of disease activity of RA patients. Fig. 6 The fitting curves and scatter plots of the clinical scores (pain VAS and PGA, Y-axis) and the three PA+SO 2 patterns (X-axis). Highest degree of pain VAS and PGA presented in pattern 3. Significant correlations between the clinical scores (pain VAS and PGA) of the three PA+SO 2 patterns could be identified