The effect of percutaneous transluminal angioplasty of superficial femoral artery on pulse wave features

The effect of percutaneous transluminal angioplasty of superficial femoral artery on pulse wave features Mikko Peltokangasa,∗, Velipekka Suominenb, Damir Vakhitovb, Jarmo Verhoa, Janne Korhonenc, Jukka Lekkalaa, Antti Vehkaojaa, Niku Oksalab,d aBioMediTech Institute and Faculty of Biomedical Sciences and Engineering, Tampere University of Technology, Tampere, Finland. Email addresses: [email protected]. Postal address: Faculty of Biomedical Sciences and Engineering, Tampere University of Technology, Korkeakoulunkatu 3, FI-33720 Tampere, Finland. tel. +3583311511. bDivision of Vascular Surgery, Department of Surgery, Tampere University Hospital, Tampere, Finland. cDivision of Interventional Radiology, Department of Radiology, Tampere University Hospital, Tampere, Finland dFinnish Cardiovascular Research Center Tampere, Surgery, Faculty of Medicine and Life Sciences, University of Tampere, Tampere, Finland. Abstract We aimed to analyze the effects of percutaneous transluminal angioplasty (PTA) of the superficial femoral artery (SFA) on arterial pulse waves (PWs). Altogether 24 subjects i.e. 48 lower limbs were examined including 26 treated lower limbs having abnormal ankle-to-brachial pressure index (ABI) (ABI<0.9 or ABI>1.3) and 22 non-treated lower limbs. The measurements were conducted in pre-, periand post-treatment phases as well as in follow-up visit after 1 month. Both ABI and toe pressures measured by standard equipment were used as reference values. PW-derived parameters include ratios of different peaks of the PW and time differences between them as well as aging index. Both treated and non-treated limbs were compared in preand post-treatment as well as follow-up visit conditions. The results were evaluated in terms of statistical tests, Bland-Altman-plots, free-marginal multirater κ-analysis and multiple linear regression analysis. PTA was found to cause small changes to the studied PW-derived parameters of the treated limb which were observed immediately after the treatment, but the changes were more pronounced in the follow-up ∗Corresponding author Email address: [email protected] (Mikko Peltokangas) Preprint submitted to Computers in Biology and Medicine March 5, 2018 This is the post-print version of the article, which has been published in Computers in Biology and Medicine 2018, 96, 274-282. The final publication is available at https://doi.org/10.1016/j.compbiomed.2018.04.003 visit. In addition, we observed that the endovascular instrumentation itself does not cause significant changes to the PW-derived parameters. The results show that PW-analysis could be a useful tool for monitoring the treatment-effect of the PTA. However, because the pre-treatment differences of the treated and non-treated limb were small, further studies with subjects having no arterial diseases are required. The study demonstrates the potential of the PW analysis in monitoring vascular abnormalities. Keywords: Atherosclerosis, Electromechanical sensors, Photoplethysmography, Peripheral arterial disease, Pulse wave measurements 1. Introduction Atherosclerosis may be present as stiffening, stenosis or occlusion of the arteries. Peripheral arterial disease (PAD) is a specific form of atherosclerosis affecting mainly the lower limbs. PAD may be asymptomatic, but often it causes symptoms such as intermittent claudication, or conditions threatening the vitality of the limb. These conditions may be critical limb ischemia with rest pain and tissue defects and acute limb ischemia. PAD is also seen as a risk factor of acute cardiovascular events such as a stroke or myocardial infarction. PAD is treated commonly by means of percutaneous transluminal angioplasty (PTA) in which the stenosed artery is dilated in a minimally invasive endovascular treatment procedure. According to the past experience, the PTA results in the disruption of the atherosclerotic plaque and initiates the histological remodeling of the arterial wall [1]. In some cases, inward remodeling and a restenosis occur, resulting in the worsening of the symptoms. The patency rates of the PTA are reported to vary between 75 %–97 % and 60 %–84 % for oneand two-year follow-up times, respectively[2]. Commonly, the patients are subjected to a 1-month follow-up to ensure the technical success of the treatment. The re-examination includes the measurement of the ankle-to-brachial pressure index (ABI). However, ABI has several limitations, such as varying sensitivity and specificity from study-to-study [3] and challenges especially with the dia2 betic and mediasclerotic patients [4, 5, 6]. As the ABI measurement has several drawbacks, imaging methods such as magnetic resonance angiography (MRA) or X-ray contrast-agent based angiography, could be alternatives. Still, MRA has high costs and X-ray angiography causes exposure to radiation. Systolic toe pressures or toe-to-brachial pressure index (TBI) have also been suggested as an alternative to ABI especially to diabetic patients, but also they have problems with the reliability [6]. Many studies have suggested pulse wave (PW) measurement and analysis for finding vascular abnormalities or abnormal vascular aging [7, 8, 9, 10, 11, 12]. Earlier, we have studied the repeatability of the PW-measurements and if there exist differences in the PW-morphologies between different-aged subjects as well as between healthy subjects and atherosclerotic patients in the signals recorded from various locations [11]. The experimental set-up of these studies limits the results only to provide information on the overall condition of the arterial tree without data on the effect of the stenosis itself. In this study, we test the suitability of the PW measurement and analysis for the monitoring of a superficial femoral artery (SFA) stenosis treated by PTA and provide information on the effects of PTA on PW. We hypothesize that PTA causes quantitative changes in the observed PW and that the effect of the treatment is observed immediately after the treatment and more clearly in the follow-up visit within one month due to remodeling initiated by the PTA. We also test if the studied PWanalysis methods are suitable for diagnostic use, e.g. if there are pre-treatment differences in the PW-features between the treated and non-treated lower limb. A benefit of the PW measurement compared with the static ABI measurement or angiographs is that it provides information also on the vascular dynamics. 2. Materials and methods 2.1. Measurement hardware and sensor placement The measurement data was collected in supine position by using electret sensors made of electromechanical film (EMFi) (Emfit S-series, Emfit Ltd, Fin3 land) and PPG probes (S0010A, Shenzhen Med-link, China) connected to a wireless body sensor network (WBSN) [13]. The sampling frequency was 250 Hz for the EMFi signals and 500 Hz for the PPG signals. The EMFi signals were interpolated to have a sampling frequency of 500 Hz in further signal processing. EMFi-sensors, which are sensitive only to dynamic but not to static loading, were placed at the wrist on the top of the radial artery, and both ankles on the top of the posterior tibial artery in order to record dynamic pressure PW signals, i.e. a signal proportional to the varying AC-component or pulse pressure of the blood pressure signal. Transmission mode PPG probes having an excitation wavelength of 905 nm were placed on index finger and both 2nd toes for recording blood volume PW signals. In addition to the PW-signal, bipolar ECG (electrocardiogram) was also measured by using conventional silver-silver chloride (Ag/Cl) electrodes placed under the clavicles and an ECG-device compatible with the WBSN [13]. The ECG was utilized in the extraction of the PWs. The measurements were conducted during the pre-, peri-, and post-treatment phases of the PTA and during the follow-up visit after one month (median 33 days, inter-quartile range (IQR) 30–36 days). For peri-procedural data collection, the sensors were placed before the normal preparation of the PTA and were removed within 5–10 minutes after the PTA. The sensor placement was similar in the follow-up visit, but the duration of the measurement was 5 minutes and was done before the ABI and toe pressure measurements. The ankle PW-signals were excluded from the analysis since they were disturbed heavily especially during the treatment mainly for two reasons: First, the PAD patients commonly have extremely low-amplitude pressure PW-signals and therefore the measurement is also very sensitive to the correct positioning of the sensors. Second, the operating table in the PTA room was narrow, which often forced patient’s lower limbs in such position that the structures containing EMFi sensors touched each other or the operating table, and this caused major artifacts to the signal. 4 2.2. Reference values The ABI is defined as a ratio of the systolic blood pressures measured by a cuff from an arm and an ankle, i.e. ABI = Pankle/Pbrachial. ABI was computed by dividing the highest ankle pressure (either ADP (arteria dorsalis pedis) or ATP (arteria tibialis posterior)) by the highest arm pressure (left or right). Toe pressures refer to the systolic pressures measured from the hallux. Both reference values were collected by using an automated measurement device, Falcon Pro (Viasonix, Israel). Pre-treatment (visit at polyclinics before the PTA) and follow-up visit ABI and toe pressure measurements were conducted as a part of normal clinical practice and used as the reference values in the study. 2.3. Study subjects The inclusion criteria of the study were the abnormal ABI reading, i.e. ABI<0.9 or ABI>1.3, relevant symptoms, stenosis in SFA based on magnetic resonance imaging angiography and a patient considered as a candidate for the PTA of the SFA. A pacemaker and a possible risk that the study interferes the patient’s treatment process were considered as exclusion criteria. Altogether 27 volunteer patients undergoing PTA of the SFA were recruited for the study. Three of them were excluded for the following reasons: research personnel unavailable, a patient was found to be unsuitable for PTA during Wrist: EMFi Index finger: PPG Ankles: EMFi Second toes: PPG ECG Figure 1: Sensor placement. 5 the imaging and extremely low peripheral perfusion. 14 left and 12 right lower limbs were treated, including two patients with both lower limbs treated. More detailed data on the study subjects is shown in Table 1. A clear majority of the included patients met the criterion ABI<0.9 — only one of the patients had pre-treatment ABI higher than 1.3. 15 non-treated lower limbs out of 22 lower limbs (68 %) also had the ABI value outside the normal range of 0.9<ABI<1.3 at least in one of the pre-treatment or follow-up visit ABI measurements. Table 1: Data describing the study subject population including clinical reference values, ABI and toe pressure. Parameter Median (IQR) Mass (kg) 75.5 (67.0...93.0) Height (cm) 170.0 (165.5...176.0) BMI (kg/m2) 26.7 (24.4...30.0) Age (years) 71.5 (67.5...76.0) Treated lower limbs: 26 ABI, pre-treatment 0.61 (0.50...0.75) * ABI, follow-up 0.96 (0.82...1.04) # Toe pressure, pre-treatment (mmHg) 59.0 (42.0...86.0) ** Toe pressure, follow-up (mmHg) 94.5 (76.0...122.0) ## Non-treated lower limbs: 22 ABI, pre-treatment 0.88 (0.75...1.02) * ABI, follow-up 0.94 (0.72...1.08) # Toe pressure, pre-treatment (mmHg) 75.5 (53.0...117.5) ** Toe pressure, follow-up (mmHg) 99.5 (66.0...122.0) ## Parameter Number (%) Males 16 (66.7 %) Diabetes 11 (45.8 %) Dyslipidemia 22 (91.7 %) Rheumatoid arthritis 3 (12.5 %) Current smoker 2 (8.3 %) Ex-smoker 11 (45.8 %) p-values for the pre-treatment ABI and toe pressure values between treated and non-treated limb: ∗:p < 0.001, **: p < 0.1, # and ##: not signigicant BMI: Body-mass index; IQR: inter-quartile range. 2.4. Ethics and patient safety The study was approved by the local ethical review board of the hospital district (R15107), the Finnish National Supervisory Authority of Health and 6 Welfare (Valvira, ID 309) and the technical department of the hospital. The study was registered to a public clinical trials register (ClinicalTrials.gov ID: NCT02725307). Written informed consents were obtained from the volunteer patients participating in the study. 2.5. Pulse wave analysis The PW-extraction was implemented first by finding R-peaks from the ECG by an algorithm proposed in [14]. After that, local mimima following each detected R-peak were extracted from the pulse wave signals. The pulse wave signal between the successive local minima was considered as a PW candidate. The validity of the PW candidates was tested by using adaptive thresholds. A PW candidate was rejected if 1) the slope of the rising edge of the PW candidate was not positive for 50 ms or if its average slope was less than 30% of the maximum slope; 2) the difference of the minimum and maximum value of the PW-candidate was outside certain thresholds; or 3) amplitude-normalized PW candidate was outside predefined limits for at least 20% of its duration. In pre-processing, the signals were filtered with a Savitzky-Golay smoothing filter having a window length of 91 samples (182 ms) and a polynomial order of 2. In addition, the signals were forward-backward filtered with a finite impulse response low-pass filter having a cutoff frequency of 10 Hz, transition band of 10–12 Hz, pass band ripple of 0.05 dB, and stop band attenuation of 100 dB [15], as implemented in [16, 11]. After pre-processing and feature extraction, the PWs were characterized by computing altogether 10 different PW-parameters. 2.5.1. PW-curve derived features The features extracted from the PWs were originally proposed either for radial pressure PW-analysis or digital volume (PPG) PW-analysis as e.g. in [17, 18]. In earlier studies [16, 11], we extracted similar features also from the lower limb PWs by using algorithms that were originally implemented for upper limb PWs and proposed in [15, 19]. In this study, we analyzed how these features, extracted from lower limb PWs, are changed as a result of the PTA of 7 the SFA. Numerous studies have presented results that the morphology of both pressure and volume pulse wave depends on the status of the vasculature. The resistance and compliance of the arterial pathway affects the wave propagation and reflection [9], but the exact physiological origin and significance of the differences requires further studies especially in case of lower limbs. Different parameters were computed based on the amplitudes and time differences between the systolic and diastolic waves as in [11] and illustrated in Fig. 2. When finding the fiducial points for P1,P2, and B, the incisura or dicrotic notch dividing the individual PW into systolic and diastolic parts was found as the last zero-crossing from negative to positive of the 1st derivative f0 of the PW in the search window limited by a point 80 ms after the maximum of the PW and a point which corresponds to 65% from the total length of the PW. In case of the absence of this zero-crossing, the location of incisura was detected at the location of the highest peak of the 2nd derivative f00 found from the same interval [16]. Based on these features, the following parameters were computed: •R1as a ratio of the amplitude of diastolic wave Band the systolic maximum (the maximum of P1and P2in Fig. 2), i.e. R1=B/ max(P1, P2), named as reflection index for index-finger PPG in [18], •R2as a ratio of the amplitude of diastolic wave Band early systolic wave P1, i.e. R2=B/P1 •R3as a ratio of the amplitude of diastolic wave Band late systolic wave P2, i.e. R3=B/P2 •R4as a ratio of the amplitude of late (P2) and early (P1) systolic wave, i.e. R4=P2/P1, named as peripheral augmentation index for wrist pressure pulses e.g. in [17], •T1as a time difference between systolic maximum (max(P1, P2)) and the peak of the diastolic wave (B) 8 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Treated lower limb Toe PPG (normalized) 0 0.2 0.4 0.6 0.8 1R1=0.61 R2=0.82 R3=0.61 R4=1.33 T1=0.184 s T2=0.296 s T3=0.184 s B P1 P2R1=0.54 R2=0.61 R3=0.54 R4=1.14 T1=0.228 s T2=0.310 s T3=0.228 s B P1 P2R1=0.38 R2=0.41 R3=0.38 R4=1.08 T1=0.326 s T2=0.386 s T3=0.326 s B P1 P2 Before treatment After treatment Follow-up visit T2 T1,T3 Time (s) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Non-treated lower limb Toe PPG (normalized) 0 0.2 0.4 0.6 0.8 1R1=0.35 R2=0.37 R3=0.35 R4=1.07 T1=0.298 s T2=0.350 s T3=0.298 s B P1 P2R1=0.38 R2=0.41 R3=0.38 R4=1.06 T1=0.300 s T2=0.346 s T3=0.300 s B P1 P2R1=0.39 R2=0.43 R3=0.39 R4=1.11 T1=0.296 s T2=0.358 s T3=0.296 s B P1 P2 Figure 2: Examples of the PWs, locations of fiducial points and time intervals, and values of the derived parameters from one test subject before and after the PTA as well as during the follow-up visit. Also T1–T3are illustrated in the PW shown in the right upper corner. T1and T3are equal in all these examples, since P2is the maximum of the PW in all the examples. Each pair of the PWs from treated and non-treated lower limb are recorded concomitantly. •T2as a time difference between early systolic peak (P1) and the diastolic peak (B) •T3as a time difference between late systolic peak (P2) and the diastolic peak (B) The peaks of early and late systolic waves P1(Fig. 2) and P2as well as diastolic wave Bwere found based on the method described in [16, 19]. These peaks are often overlapped in the PWs recorded from atherosclerotic patients, but they are more obvious in radial or carotid PWs and in lower limb PWs recorded from healthy subjects [16, 11]. For these reasons, the fiducial points were found by implementing 5th-order derivative analysis as in [16, 19]. After finding the boundary between the systolic and diastolic parts, the following procedure was implemented for finding P1and P2[16, 19]: If the sign of the 5th derivative f(5) at the point corresponding to the systolic maximum of the PW was 1. positive, this point was considered as a point for the late systolic peak P2 9 and follow-up visit have at least moderate acceptance with the ABI, and the κ-values between ABI and R1,R3,T1and T2parameters are even higher than the κ-value between ABI and toe pressure (0.75). All κ-values between the ABI and PW-derived parameters are higher than corresponding values between toe pressures and PW-derived parameters. Table 3: Free-marginal multirater κvalues for the changes between different values and ABI and toe pressure. Parameter ABI Toe pressure R10.82 0.62 R20.73 0.52 R30.91 0.71 T10.73 0.62 T20.82 0.71 T30.82 0.71 Awt 0.52 0.30 Aft 0.55 0.52 Toe pressure 0.75 3.6. Multiple linear regression The results of multiple linear regression analysis for the changes between pre-treatment situation and follow-up visit are shown in Table 4. According to the results, the changes are dependent on the treatment of the SFA but not on the RR-interval except T2. Surprizingly the toe pressures are less dependent on the treatment of the SFA than PW-derived indices and ABI. 3.7. Pre-treatment repeatability Both ICC and CV are shown in Table 5 for beat-to-beat data in pre-treatment situation. ICCs close to 0.9 or higher and median CVs less or equal than 2% were observed. Both indicators show good repeatability and are inline with the earlier study [11]. 16 1/2#(pfollow-up+pbefore) 0.3 0.4 0.5 0.6 pfollow-up-pbefore -0.4 -0.2 0 0.2 a) R1 1/2#(pfollow-up+pbefore) 0.3 0.4 0.5 0.6 0.7 0.8 pfollow-up-pbefore -0.5 0 0.5 b) R2 1/2#(pfollow-up+pbefore) 0.3 0.4 0.5 0.6 pfollow-up-pbefore -0.4 -0.2 0 0.2 c) R3 1/2#(pfollow-up+pbefore) (s) 0.25 0.3 0.35 0.4 0.45 pfollow-up-pbefore (s) -0.2 0 0.2 f) T1 1/2#(pfollow-up+pbefore) (s) 0.25 0.3 0.35 0.4 0.45 0.5 pfollow-up-pbefore (s) -0.1 0 0.1 0.2 0.3 g) T2 1/2#(pfollow-up+pbefore) (s) 0.15 0.2 0.25 0.3 0.35 pfollow-up-pbefore (s) 0 0.2 0.4 h) T3 1/2#(pfollow-up+pbefore) 0.7 0.8 0.9 1 1.1 pfollow-up-pbefore -0.4 -0.2 0 0.2 0.4 d) Wrist/Toe (Awt) 1/2#(pfollow-up+pbefore) 0.85 0.9 0.95 1 1.05 1.1 1.15 pfollow-up-pbefore -0.5 0 0.5 e) Finger/Toe (Aft) 1/2#(pfollow-up+pbefore) 0.5 1 1.5 2 pfollow-up-pbefore -0.4 -0.2 0 0.2 0.4 0.6 0.8 i) Reference value: ABI 1/2#(pfollow-up-pbefore) (mmHg) 0 50 100 150 200 pfollow-up-pbefore (mmHg) -100 0 100 j) Reference value: Toe pressure Non-operated limb Average difference (non-operated limb) 95% limits of agreement (non-operated limb) Operated limb Average difference (operated limb) Figure 4: Bland-Altman plots for the pre-treatment values and the values found in the followup visit for each PW-parameter and the reference values ABI and toe pressure. The normal range for ABI is also shown in black vertical dashed lines in panel i). 17 Table 4: Multiple linear regression coefficients and significance levels. The marker after parameter name refers to p-value of the F-statistics vs. constant model. Coefficients Parameter Intercept ∆RR SFA PTA R1◦-0.047 -0.164 -0.116• R2•-0.058 -0.257 -0.141• R3•-0.041 -0.134 -0.133/ T1•0.013 -0.056 0.054/ T2/0.010 -0.235•0.047• T3◦0.011 -0.129 0.052• Awt•0.005 0.047 0.086/ Aft•-0.004 0.123 0.108/ ABI/-0.053 0.490 -0.210/ Toe pressure -4.731 19.928 -20.727* ∆RR: change in RR-interval ∗:p < 0.10, ◦:p < 0.05, •:p < 0.025, /:p < 0.01 Table 5: Intra-class correlations (ICC) and coefficients of variation (CV) for pre-treatment measurement. Parameter ICC (95% confidence limit) Median CV (1st...3rd quartile) (%) R10.96 (0.93–0.97) 0.6 (0.4...3.0) R20.96 (0.93–0.97) 0.5 (0.3...3.0) R30.96 (0.94–0.97) 0.6 (0.4...2.9) T10.89 (0.84–0.93) 1.5 (0.9...2.7) T20.89 (0.83–0.92) 1.5 (0.8...3.5) T30.91 (0.87–0.94) 2.0 (1.2...4.0) Awt 0.96 (0.94–0.97) 1.9 (1.4...3.2) Aft 0.97 (0.96–0.98) 1.3 (1.0...1.8) 4. Discussion The presented results suggest that PTA caused significant changes in the features extracted from the PWs. The majority (R1–R3,T1–T3,Awt, and Aft) of the studied parameters showed statistically significant differences between pretreatment condition and a situation at the follow-up visit. Only R2,T2–T3, and Aft showed differences between the pre-treatment and immediate post-treatment values, but other PW-derived parameters still had trends towards healthier condition. The PW-measurement is able to find some immediate changes, but the results suggest that the remodeling of the plaque and arterial wall within the 18 follow-up period makes the changes more visible. Therefore, the behaviour of the post-treatment PW-parameters is similar to the post-treatment ABI as proposed in [23] and especially if the pre-treatment ABI is less than 0.80 [24]. According to the results, the endovascular instrumentation inserted into SFA does not affect the values of the peri-procedural PW-parameters. The independence of the PW-parameters on the insertion of the endovascular instruments could allow to monitor changes in PW morphology caused by the PTA in real time, even though the immediate changes could be small. 4.1. PW-analysis and changes caused by PAD To the extent of our knowledge, the presented study is the first study which compares the values of lower-limb PW-derived parameters between the preand post-treatment situations of the PTA of the SFA. However, the results are supported by other studies concentrating on the diagnosis of the PAD: Yokoyama et al. [25] have reported an increase in PW velocity in diabetic patients due to arterial stiffness, but decrease in the PW velocity in a lower limb having PAD symptoms likely due to stenoses which decrease the blood pressure and thus the PW velocity in the diseased lower limb. After the successful PTA, the PW velocity increased indicating the changes in the lower limb vasculature. Similar kind of results were found in this study: the values of the PW-derived parameters change in the healthier direction as reported earlier in [11]. The PW-analysis can also reveal the beneficial wide-scale systemic remote effects of the PTA: Jacomella et al. [12] reported changes towards better condition within 3-month follow-up period of the PTA of the SFA in a carotid artery augmentation index based on a radial artery applanation tonometry. In the present study, the medians of the distributions of R1–R3in non-treated limbs (Table 2) shifted towards healthier values observed in [11] during the 1-month followup period even though the patient-wise differences between pre-treatment and follow-up visit situations were close to zero (Figs. 3h–j). However, the changes in the IQRs were not as dramatic in Table 2. We assume that a successful PTA restores physical activity, such as walking, also improving the condition of the 19 contralateral lower limb and resulting in increased perfusion in the lower limbs. Similar changes were also observed in the clinical references values (Table 1) and Figs. 3d and 3k) which supports the assumption. The area ratios Awt and Aft as well as time intervals T1–T3behave differently: their distributions at pre-PTA and follow-up visit situations did not differ from each other in case of non-operated lower limbs (Table 2), indicating weaker dependence on positive systemic remote effects. Further studies should confirm these assumptions. There are also studies [7, 8, 9, 10] reporting bilateral differences between the lower limbs with and without PAD. Erts et al. [10] have reported time differences in the foot points (the beginnings of the PW) between the lower limb with and without the PAD. Allen et al. [7, 8, 9] have found bilateral differences between toe-PPG signals in unilateral PAD patients. A comparison between different time delays between the peak or foot points of the PWs from different locations and between the patients and control subjects is also in our future interests. 4.2. Limitations of the method In previous study[11], statistically significant differences were found for the study groups consisting of patients having atherosclerotic changes and sameaged healthy control subjects. The experimental setup in the previous study [11] was different than in this study in which the lower limbs that underwent SFA of the PTA were compared with the lower limbs without the PTA. The results of the present study do not show significant bilateral differences between the treated limb and the non-treatment limb if the pre-treatment values of the PW-derived parameters are studied. This possibly limits the diagnostic use of the studied PW-analysis methods. Possible reason is that 68% of non-treated lower limbs provided abnormal ABI reading at least in one of the two measurements. If a patient has the symptoms of PAD in one limb, there is a high likelihood of atherosclerotic changes also contralaterally and elsewhere in the vasculature. The studied PW-derived parameters may also be sensitive to all kinds of degenerative atherosclerotic changes including the stiffening of the large arteries such as the aorta, not only 20 the stenoses and occlusions which are the typical lesions of the PAD. Still, depending on the method, either bilateral comparison or unilateral measurement can provide the better performance, even though the differences are small as in [8, 9]. If the obvious sources of error in the PPG-measurement, such as incorrectly placed PPG-probe and motion artefacts, are excluded, there are a few methodological limitations. Poor peripheral perfusion decreases the amplitude of the PPG-signal and may prevent the proposed analysis. The low-amplitude PPGsignal can be a marker of PAD, but can also be caused by vasoconstriction as the result of low temperature. The body position affects the shape of the PPG-signal and is a topic of further studies. However, in the present study, all the measurements were conducted in supine position and limbs extented. Light from the environment can disturb the PPG-signals, but in our measurements, regular 50 Hz and 100 Hz frequency components caused by the mains and the fluorescent lights were the most significant sources of disturbance and were eliminated by the filtering. Despite the limitations discussed, the results suggest that PTA has an effect on most of the features extracted and evaluated in the present study. The analysis of the PW-features may provide additional information in finding the early signs of PAD or in monitoring the treatment during the follow-up period especially in the cases in which the conventional methods, ABI and toe pressure measurements, provide in unreliable or conflicting results. A part of the features combined with other health data may be utilized also as the inputs of multivariate analysis methods or machine learning which would be a topic of further studies. 5. Conclusions Statistically significant differences were found in the values of the PWderived features between pre-treatment and immediate post-treatment condition. Stronger differences in the treated lower limb were found between the 21 pre-treatment values and the post-treatment conditions after a 1-month followup period. In both cases, the changes in the values of the treated limb differed significantly from the corresponding changes found for the non-treated lower limb. In case of the non-treated lower limbs, the variations of the differences between different measurement events were relatively high but without a consistent trend towards healthier or worse condition. The endovascular instrumentation inserted into SFA did not affect the PW-parameters. The study was not able to show statistically significant differences in the pre-treatment values of the PW-derived parameters between the treated and non-treated lower limbs. This is possibly explained by the study sample: 68 % of the non-treated limbs produced an abnormal ABI-reading at least for one of the two measurements. This indicates that many patients had subclinical atherosclerotic changes also in the other lower limb. Prior to the real utilization of the measurement method, further comparison of the healthy subjects and PAD-patients and having standardized measurement conditions should confirm the findings. Conflict of interest statement The authors declare no conflict of interest. Acknowledgment We would like to thank all the volunteer test subjects for their valuable contribution. The personnel in Tampere University Hospital (Department of Vascular Surgery) are acknowledged for their contribution to collecting the reference values. The work was funded by the Doctoral Programme of the President of Tampere University of Technology, Finnish Funding Agency for Innovation (TEKES) as a part of project VitalSens (decision ID 40103/14), and grants from Finnish Cultural Foundation/Pirkanmaa Regional Fund/Elli and Elvi Oksanen’s Fund, Tekniikan edist¨amiss¨a¨ati¨o and Emil Aaltonen Foundation. 22 References [1] R. Wyttenbach, A. Gallino, M. Alerci, F. Mahler, L. Cozzi, M. Di Valentino, J. J. Badimon, V. Fuster, R. Corti, Effects of percutaneous transluminal angioplasty and endovascular brachytherapy on vascular remodeling of human femoropopliteal artery by noninvasive magnetic resonance imaging, Circulation 110 (9) (2004) 1156–1161. [2] J. R. Laird, Limitations of percutaneous transluminal angioplasty and stenting for the treatment of disease of the superficial femoral and popliteal arteries, Journal of Endovascular Therapy 13 (2 suppl) (2006) II–30–II–40. [3] D. Xu, J. Li, L. Zou, Y. Xu, D. Hu, S. L. Pagoto, Y. Ma, Sensitivity and specificity of the ankle-brachial index to diagnose peripheral artery disease: a structured review, Vascular Medicine 15 (5) (2010) 361–369. [4] V. Suominen, T. Rantanen, M. Venermo, J. Saarinen, J. Salenius, Prevalence and risk factors of pad among patients with elevated abi, European Journal of Vascular and Endovascular Surgery 35 (6) (2008) 709–714. [5] S. Alatab, H. Fakhrzadeh, F. Sharifi, A. Mostashfi, M. Mirarefin, Z. Badamchizadeh, Y. Tagalizadehkhoob, Impact of hypertension on various markers of subclinical atherosclerosis in early type 2 diabetes, Journal of Diabetes & Metabolic Disorders 13 (1) (2014) 1. [6] M. T. Romanos, A. Raspovic, B. M. Perrin, The reliability of toe systolic pressure and the toe brachial index in patients with diabetes, Journal of foot and ankle research 3 (1) (2010) 31. [7] J. Allen, A. Murray, Similarity in bilateral photoplethysmographic peripheral pulse wave characteristics at the ears, thumbs and toes, Physiological measurement 21 (3) (2000) 369. [8] J. Allen, K. Overbeck, A. F. Nath, A. Murray, G. Stansby, A prospective comparison of bilateral photoplethysmography versus the ankle-brachial 23 pressure index for detecting and quantifying lower limb peripheral arterial disease, Journal of vascular surgery 47 (4) (2008) 794–802. [9] J. Allen, C. P. Oates, T. A. Lees, A. Murray, Photoplethysmography detection of lower limb peripheral arterial occlusive disease: a comparison of pulse timing, amplitude and shape characteristics, Physiological measurement 26 (5) (2005) 811. [10] R. Erts, J. Spigulis, I. Kukulis, M. Ozols, Bilateral photoplethysmography studies of the leg arterial stenosis, Physiological measurement 26 (5) (2005) 865. [11] M. Peltokangas, A. A. Telembeci, J. Verho, V. M. Mattila, P. Romsi, A. Vehkaoja, J. Lekkala, N. Oksala, Parameters extracted from arterial pulse waves as markers of atherosclerotic changes: performance and repeatability, IEEE Journal of Biomedical and Health InformaticsIn press. DOI: 10.1109/JBHI.2017.2679904. [12] V. Jacomella, A. Shenoy, K. Mosimann, M. Kohler, B. Amann-Vesti, M. Husmann, The impact of endovascular lower-limb revascularisation on the aortic augmentation index in patients with peripheral arterial disease, European Journal of Vascular and Endovascular Surgery 45 (5) (2013) 497– 501. [13] M. Peltokangas, A. Vehkaoja, J. Verho, M. Huotari, J. Rning, J. Lekkala, Monitoring arterial pulse waves with synchronous body sensor network, Biomedical and Health Informatics, IEEE Journal of 18 (6) (2014) 1781– 1787. doi:10.1109/JBHI.2014.2328788. [14] M. Peltokangas, J. Verho, A. Vehkaoja, Night-time ekg and hrv monitoring with bed sheet integrated textile electrodes, Information Technology in Biomedicine, IEEE Transactions on 16 (5) (2012) 935–942. [15] S. C. Millasseau, R. P. Kelly, J. M. Ritter, P. J. Chowienczyk, The vascular impact of aging and vasoactive drugs: comparison of two digital volume 24 pulse measurements, American journal of hypertension 16 (6) (2003) 467– 472. [16] M. Peltokangas, A. Vehkaoja, J. Verho, V. M. Mattila, P. Romsi, J. Lekkala, N. Oksala, Age dependence of arterial pulse wave parameters extracted from dynamic blood pressure and blood volume pulse waves, IEEE journal of biomedical and health informatics 21 (1) (2017) 142–149. [17] V. Melenovsky, B. A. Borlaug, B. Fetics, K. Kessler, L. Shively, D. A. Kass, Estimation of central pressure augmentation using automated radial artery tonometry, Journal of hypertension 25 (7) (2007) 1403–1409. [18] S. C. Millasseau, J. M. Ritter, K. Takazawa, P. J. Chowienczyk, Contour analysis of the photoplethysmographic pulse measured at the finger, Journal of hypertension 24 (8) (2006) 1449–1456. [19] K. Takazawa, N. Tanaka, K. Takeda, F. Kurosu, C. Ibukiyama, Underestimation of vasodilator effects of nitroglycerin by upper limb blood pressure, Hypertension 26 (3) (1995) 520–523. [20] J. Randolph, Free-marginal multirater kappa (multirater κfree): An alternative to fleiss’ fixed-marginal multirater kappa, in: Presented at the Joensuu Learning and Instruction Symposium, Vol. 2005, 2005. [21] K. O. McGraw, S. P. Wong, Forming inferences about some intraclass correlation coefficients., Psychological methods 1 (1) (1996) 30. [22] C. Lessells, P. T. Boag, Unrepeatable repeatabilities: a common mistake, The Auk (1987) 116–121. [23] S. Ray, T. Buckenham, A. Belli, R. Taylor, J. Dormandy, The nature and importance of changes in toe-brachial pressure indices following percutaneous transluminal angioplasty for leg ischaemia, European journal of vascular and endovascular surgery 14 (2) (1997) 125–133. 25