European Urology

1. Introduction

Bladder outlet obstruction (BOO) is commonly associated with lower urinary tract symptoms (LUTS) in men [1]. BOO is a mechanical phenomenon; therefore, it is normally studied using mechanical parameters, such as pressure and flow. Intuitively, if obstruction in a fluid-transporting system is present, pressure must be increased to transmit the usual flow [2]. This is physiologically justified because the detrusor satisfies the Hill model of muscle contraction [2], and [3], which expresses the inverse relationship between detrusor strength and flow at a given bladder volume. This translates into the bladder-output relation (BOR) [4], the approximately inverse relationship between detrusor pressure (P_det) and flow. Due to the BOR, obstruction is associated with high P_det and/or low flow.

Urodynamic tests are the current standard for classification of the degree of obstruction [1]. Pressure–flow studies constitute the gold standard for diagnosing BOO, through the use of the Abrams-Griffiths [5] and Schäfer [6] nomograms. With the Abrams-Griffiths nomogram, obstruction is diagnosed by plotting the values of maximum flow (Q_max) and P_det at Q_max (P_det[Q_max]). This point can fall in one of the three regions of the nomogram: obstructed, unobstructed, and equivocal for obstruction.

Recently, a noninvasive [7], [8], [9], [10], and [11] and invasive [12] optical method for monitoring the bladder detrusor using near-infrared spectroscopy (NIRS) has been described. NIRS is an established, basic science technique that was first applied to in situ human tissue in 1977 [13], [14], [15], and [16] and was used in urological research in 1990 [17]. The technology has evolved to enable several tissues and organs to be studied and a range of parameters to be measured [14]. NIRS uses near-infrared (NIR) light introduced into tissue transcutaneously to monitor changes in tissue oxygenation and haemodynamics [13], and [16]. This is achieved by measurement of changes in the concentration of naturally occurring compounds called chromophores that absorb NIR light, depending on their concentration and chemical structure [18]. The principal chromophores of interest in urologic applications of NIRS are oxyhaemoglobin (O₂Hb) and deoxyhaemoglobin (HHb) [8], [9], and [11]. Two physics principles underlie NIRS: (1) light in the NIR spectrum, unlike visible light, penetrates skin and scatters in tissue, and (2) oxygen-dependent changes in chromophore concentration can be measured because the absorption spectra of O₂Hb and HHb are different at specific wavelengths in the NIR region [13], and [18]. Chromophore concentration (expressed in μmol/100 ml) is measured in real time as a change relative to baseline. Changes in total haemoglobin concentration (tHb), derived from the sum of the O₂Hb and HHb, indicate a change in blood volume [9], [10], [13], and [16].

Applications in a variety of urologic conditions have principally evaluated vascular and haemodynamic aetiologies [8], [9], and [17]. The bladder detrusor can be monitored via a noninvasive suprapubic sensor during simultaneous urodynamic studies of filing and emptying [7], [8], and [9]. Following preliminary animal and human trials [8], our hypothesis was that NIRS-derived patterns of change in O₂Hb and HHb concentrations observed during voiding would differ between patients presenting with LUTS who were classified as obstructed or unobstructed using pressure–flow studies [11]. This study was performed to evaluate the ability of mathematical modelling and analysis of NIRS data via a classification and regression tree (CART) algorithm [19] independently to classify subjects assessed as obstructed or unobstructed using invasive urodynamic data and nomogram classification.

2. Methods

This study received ethical approval (HO6-70108) from the University of British Columbia ethics board.

3. Results

Seventy patients were enrolled, and 64 NIRS data sets were analysed. Four were incomplete because of technical difficulties relating to the NIRS data: fibre-optic cable failure (n = 3) and a data saving error (n = 1). Two were excluded by haematuria (n = 2), a potential confounder due to NIR light absorption. Mean age was 62 yr (range: 49–91 yr). IPSS scores ranged from 12 to 34 (mean: 19). BMI ranged from 18 to 35 (mean: 25). There were no subjects with diabetes.

By Abrams-Griffiths classification of the urodynamic data, 30 of the 64 subjects (47%) were obstructed; 16 of the subjects (25%) were unobstructed; and 18 of the subjects (28%) were equivocal for obstruction. Subjects in the obstructed zone had a Q_max ranging from 2 cm³/s to 18 cm³/s (mean: 11 cm³/s), and a P_det(Q_max) ranging from 7 cm to 150 cm of water (mean: 73 cm). For subjects in the unobstructed zone, Q_max ranged from 2 cm³/s to 38 cm³/s (mean: 15 cm³/s) and the P_det(Q_max) ranged from 5 cm to 87 cm of water (mean: 27 cm). Subjects that were equivocal for obstruction had a Q_max ranging from 2 cm³/s to 19 cm³/s (mean: 10 cm³/s) and a P_det(Q_max) ranging from 0 cm to 44 cm of water (mean: 34 cm).

Data from the 46 subjects classified as obstructed or unobstructed were used for subsequent analysis. Scatter plots of the three relative chromophore concentrations (O₂Hb, HHb, and tHb [O₂ Hb + HHb]) relative to P_det(Q_max) are shown in Fig. 1, Fig. 2, and Fig. 3.

Create Platinum Slide Series

Fig. 1 Detrusor pressure (P_det) at maximum flow (Q_max) versus relative concentration of deoxyhaemoglobin (HHb) at Q_max. Patients classified as obstructed according to the Abrams-Griffiths nomogram are plotted using crosses (+), whereas patients classified as unobstructed are plotted using circles (o). The dashed line represents the best line dividing the data set in two regions, obstructed and unobstructed, according to the linear discriminant analysis.

Create Platinum Slide Series

Fig. 2 Detrusor pressure (P_det) at maximum flow (Q_max) versus relative concentration of oxyhaemoglobin (O₂Hb) at Q_max. Patients classified as obstructed according to the Abrams-Griffiths nomogram are plotted using crosses (+), whereas patients classified as unobstructed are plotted using circles (o). The dashed line represents the best line dividing the data set in two regions, obstructed and unobstructed, according to the linear discriminant analysis.

Create Platinum Slide Series

Fig. 3 Detrusor pressure (P_det) at maximum flow (Q_max) versus relative concentration of total haemoglobin (tHb) at Q_max. Patients classified as obstructed according to the Abrams-Griffiths nomogram are plotted using crosses (+), whereas patients classified as unobstructed are plotted using circles (o). The dashed line represents the best line dividing the data set in two regions, obstructed and unobstructed, according to the linear discriminant analysis.

LDA of each of these NIRS data sets gave the same misclassification error of 16%, while application of the CART algorithm resulted in a misclassification error of 4%. The CART calculation used a learning sample in which a learning set consisted of the relative concentrations of O₂Hb and HHb for patients classified according to the Abrams-Griffiths nomograms classifications of obstructed or unobstructed. Elements in the learning set were paired [HHb(t), O₂Hb(t)] for every time t in the time series of recorded data. The total number of pairs in the learning set was 37 540.

The following classification responses were obtained: true-positive fraction (TP) = 30, true-negative fraction (TN) = 14, false-positive fraction (FP) = 2, false-negative fraction (FN) = 0. From these fractions, the specificity and sensitivity of the method were calculated: sensitivity = TP/(TP + FN) = 1; specificity = TN/(TN + FP) = 0.8750; and precision = TP/(TP + FP) = 0.9375.

4. Discussion

This study reports simultaneous, noninvasive NIRS monitoring of changes in detrusor concentration of O₂Hb and HHb and conventional invasive pressure–flow measurements in men with LUTS. The scatter plots derived indicate that this group of patients could be divided into obstructed and unobstructed groups based on P_det(Q_max) and relative change in concentration of O₂Hb and HHb. LDA adds information on the relevance of different variables. Because the LDA misclassification error for O₂Hb and HHb is the same, only one of these equivalent relative concentrations and P_det(Q_max) are required for a linear classification of the presence of obstruction. However, classification is also possible using the noninvasive NIRS data alone, where analysis using a CART algorithm identifies those with and without obstruction with excellent sensitivity and specificity.

Because NIRS measures optical parameters influenced by physiologic changes occurring in the detrusor microcirculation and urodynamic measurements reflect pressure, NIRS chromophore concentration changes and urodynamic pressures can be expected to have a relationship based on current knowledge of the pathophysiology of obstruction, which includes elements of vascular change in the aetiology and evolution of BOO. Animal-model data suggest that a reduction in detrusor blood flow occurs in BOO [21], and [22]). Belenky et al [23] showed that BOO is associated with a reduction in blood flow to the detrusor using pressure–flow and Doppler studies, and Levin [24] described detrusor hypertrophy and accompanying alterations in innervation and metabolic factors as effects of outlet obstruction in humans. We have previously reported significant synchrony between P_det and NIRS chromophore changes in females with LUTS [8].

However, NIRS changes in chromophore concentration reflect the dynamic and sometimes multidirectional relationship that exists between tissue oxygen saturation, venous oxygen saturation, oxygen delivery, and other parameters dictating tissue respiration [16], and [25]. Hence, NIRS data from monitoring changes in detrusor oxygenation and haemodynamics are nonlinear in nature. Therefore, where NIRS data are analysed alone, nonparametric data analysis will be more discriminating. Hence, our application of a CART algorithm in this study and the degree of specificity and sensitivity achieved in classifying subjects with BOO is based on NIRS data alone.

CART is a nonparametric classification algorithm based on decision trees [19]. CART use includes applications to other spectroscopic data and clinical situations in which there is no prior knowledge of the probability distribution of the data [19], [26], [27], and [28]. Examples where CART have been used successfully include the segregation of patients with unknown primary carcinoma into groups with similar clinical features and survival [27], and accurate separation of patients with and without myocardial infarction on admission to the emergency department [28]. Because of this capability, we used CART methodology, recognising that NIRS of the bladder detrusor is a new application of this technology, and that the majority of NIRS parameters measured do not have an established gold standard of measurements for comparison [13], [14], and [16].

Currently, pressure–flow studies and their related nomograms provide the gold standard for the classification of patients as obstructed or unobstructed, and they use a single point to establish this classification [2]. The relationship of this point to NIRS data in our study is shown by linear models. However, a proportion of patients fall into an equivocal zone on the Abrams-Griffiths nomogram and are not definitively classified according to this methodology. For this reason this group was excluded from our data analysis. While the CART algorithm could be used to classify the equivocal group, doing so would generate data that could not be verified against any gold standard. In future studies where the CART analysis of NIRS data is used, performing a series of measurements in patient's classified as equivocal by urodynamic testing would allow the clinician to ultimately validate classification of these subjects as unobstructed or obstructed as their LUTS evolve.

In contrast to the single point used for urodynamic classification, the NIRS algorithm uses data from changes occurring in the detrusor throughout the whole voiding cycle. Hence, patterns of change, trends, and magnitudes of change in oxygenation and blood volume are included. This may explain the excellent ability of CART analysis of the NIRS data to discriminate even though the spread of chromophore change appears similar between groups. It is possible that further research with NIRS will enable patients to be classified by other factors, apart from pressure, which are probably involved in the pathophysiology of BOO. For example, factors relating to oxygenation and haemodynamics may be relevant. Prior research using Doppler flow probe studies in humans does indicate that blood flow tends to increase with increasing volume and pressure [29]. Consequently, as NIRS chromophore concentration changes reflect variations in detrusor oxygenation and blood volume [7], [8], [9], [10], and [11], NIRS data collected simultaneously with pressure flow data should provide further insight into vascular and haemodynamic mechanisms related to BOO. Another example would be the potential of NIRS to quantify degrees of obstruction and/or the contribution of specific underlying diseases that are known to influence the vascular system, including diabetes and arteriosclerosis. Future studies would benefit from additional clinical history and screening for markers of vascular disease. Such screening could include, for example, lipid profiles, haemoglobin A1C, blood pressure, and echocardiogram.

Confidence that NIRS monitors change within the detrusor and that chromophore concentration changes observed are a valid reflection of physiologic effect come from observations that include ultrasound confirmation of bladder position related to NIRS sensor placement; knowledge of the depth of penetration of NIR light into tissue; the fact that significant chromophore change only occurs in relation to temporal events during voiding; the fact that simultaneous recording at an abdominal site remote from the bladder yields no evidence of change; and observations from animal and human studies that specific physiologic events (eg, hypoxia, ischaemia, and changes in blood volume) generate similar patterns of change in the detrusor as seen in other tissue (eg, brain and muscle) [8], [10], [13], and [16].

The sensitivity and specificity of the CART nomogram developed in this study suggest that further studies are warranted. The NIRS technique is attractive to patients because it is noninvasive and because the laser-generated light used is not toxic. A prospective study would enable further precision of the CART algorithm to be established and would allow the ability of NIRS to classify equivocal patients to be investigated. Simultaneous NIRS and urodynamic testing would allow the pathophysiology related to the vascular effects of obstruction to be more comprehensively described and would allow pretreatment and posttreatment values to be evaluated. We suggest that, because NIRS monitoring with this algorithm has this degree of discrimination ability and is also noninvasive, it is a promising alternative to current invasive pressure–flow studies that should be evaluated in conjunction with current diagnostic methods. In this way, it will be possible to evaluate the contributions that NIRS can make to more comprehensive diagnosis in BOO and also to establish whether it has a place used alone as a noninvasive screening option.

5. Conclusions

Simultaneous measurements of invasive pressure–flow and noninvasive, NIRS-derived changes in O₂Hb and HHb concentrations in the detrusor were conducted in male patients with LUTS. When P_det(Q_max) values and NIRS changes were analysed using LDA, a relatively good classification of subjects with and without obstruction resulted. However, because the patterns of change in O₂Hb and HHb are nonlinear, the use of a CART algorithm is more applicable. Using CART, the noninvasive NIRS data have an independent ability to distinguish between patients with and without obstruction, with good sensitivity and specificity.