Toward development of PreVoid alerting system for nocturnal enuresis patients: A fuzzy-based approach for determining the level of liquid encased in urinary bladder

Highlights

• The manuscript discusses a machine-learning based technique for determining the level of liquid encased in urinary bladder.

• The proposed method is trained over each individual’s voiding-filling pattern of the bladder, and therefore is independent of gender, BMI and level of obesity of a patient.

• The method assumes the maximum capacity of the bladder is the amount of urine it contains when one feels the urge to void, regardless of the size of bladder.

• Four extracted features obtained via echoed-back pulses of ultrasound are fed to a novel fuzzy error correcting output classifier to see which proportion of bladder is filled.

Abstract

Preventive and accurate assessment of bladder voiding dysfunctions necessitates measuring the amount of liquid encapsulated within urinary bladder walls in a non-invasive and real-time manner. The real-time monitoring of urine levels helps patients with urological disorders such as Nocturnal Enuresis (NE) by preventing the occurrence of enuresis via a pre-void stage alerting system. Although some advances have been achieved toward developing a non-invasive approach for determining the amount of accumulated urine inside the bladder, there is still a lack of an easy-to-implement technique which is suitable to embed in a wearable pre-warning device. This study aims to develop a machine-learning empowered technique to quantify to what extent an individual's bladder is filled by observing the filling-voiding pattern of a patient over a training period. In this experiment, a pulse-echo sonar element is used to generate ultrasound pulses while the probe surface is positioned perpendicular to the bladder's position. From the reflected echoes, four features which show sufficient sensitiveness and therefore could be modulated noticeably by different levels of liquid encased in the bladder, are extracted. The extracted features are then fed into a novel intelligent decision support system– known as FECOC – which is based on hybridization of fuzzy inference systems (FIS) and error correcting output codes (ECOC). The proposed scheme tends to achieve better results when examined in real case studies.

Introduction

Nocturnal Enuresis (NE) is a multifactorial condition which - according to American Paediatric Academy – is defined as involuntary episodes of urination at night, mainly in absence of organic disease and after the age when a person expected to be able to control his/her urination [1]. The aetiology of NE has turned out to be complex and its underlying cause could include urological, neurological, sleep abnormalities, genetic and even psychosocial factors [4]. NE is a common disorder in children, with an estimated prevalence of 20 % in young children (aged <5 years) and 2% in young adults. Children who wet the bed may experience parental disapproval, sibling teasing, emotional stigma and social disruption which may interfere with the development of their self-esteem [2].

Analysing bladder dysfunctions such as NE hugely depends on the data related to the volume of fluid encapsulated in the urinary bladder. In addition to NE, literature discussed two other major urinary deficits in which, measuring the level of urine within the bladder and producing a pre-warning alert could be beneficial: First, the patients suffering from Bladder Control Problem (BCP). BCP is resulted mainly because of the injury in the nervous system responsible for regulating the interaction between the brain (and spinal cord) and organs of the lower urinary tract. BPC patients are usually stimulated by a sensation that alerts the person the need for voiding a substantially filled bladder. Such patients are at risk of bladder over-distention. Second; it is also of interest for patients with an increased-in-size prostate, which residual urine may be retained in the bladder even after voiding.

Accurate assessment of bladder voiding dysfunctions often necessitates invasive urodynamic studies [2]. Although these methods also have some advantages, fear of invasive studies prevented those methods to be widely utilized in clinical routines [3]. On the other hand, the non-invasive quantitative measurements tends to be more promising in ambient clinical practice [5]. Among the non-invasive methods, convenience, efficiency, and safety of ultrasound-based methods make them more beneficial and justifiable in a wider variety of clinical applications.

Numerous attempts have been made in favour of bladder urodynamics to find the volume within in recent years. Most of those non-invasively try to sense complex synergies of bladder wall muscle activity and infer the volume of urine inside. Some advances have been done by measuring the volume of the bladder via ultrasound A-mode ultrasound (US) scanning of the sagittal plane. In A-scan the x-axis represents the time/distance required for the return of the echo and y-axis corresponds to the strength of the echo. It was shown that the number of A-scan intersect the bladder, as well as the measured distance between the anterior and posterior bladder walls, correlates with the bladder volume [6,7]. Although it was proved that it is a promising alternative to catheterization but most of the methods which are based on bladder cavity volume measurements merely rely on measuring antero-posterior diameter that, on its own, revealed a quite poor correlation with exact bladder volume. In approaches patented by [8] and [9], ultrasonic transducers used with different methodologies to estimate the volume of urine encapsulated in the bladder. In former, the aim is accomplished by sequentially scanning the bladder with ultrasonic beams that split the bladder into a number of transverse planes. After determining the area in each of the planes, the system determines the volume of the bladder by summing a weighted version of the planar areas. In both methods, deterministic approaches lead to a semi-robust system. Despite being simple and rapid, these approaches suffer from significant limitation i.e. sensitiveness to determining precisely the size and shape of the bladder. Most of these methods proposed to present the bladder by different geometrical shape. The work by [10] is based on the characteristics of the sensor signal observed during a natural voiding event. It stimulates the bladder to void based on some voiding pattern. Its main advantage could be limited to recognise when urinary has been initiated and not provide information needed during the critical time which is the preliminary need to void. In the research published in [11], a US scanner to evaluate the amount of urine inside the bladder is introduced. The design consists of a 2D matrix of transducers to cover as much as possible the bladder. The final estimation assumes that the bladder can be approximated by a sphere.

Some researchers managed to implement machine learning and artificial intelligence schemes to detect and diagnose diseases in the bladder, i.e. cancers and tumours in bladders [[12], [13], [14], [15], [16], [17], [18]]. The signal processing techniques also demonstrated comparatively well in measuring urinary bladder volume, as shown by the work of Padmapriya et al. [19]. The same authors encompass different non-invasive methods to measure the bladder volume via ultrasound images in their recent publications [20].

In this study, we present a machine-learning based system working on a discriminant feature analysis of ultrasound back-scattered signals. Along with every feature, a brief motivation which justifies using those features is explained. The features derived, are then fed into a decision support system to quantify the proportion of liquid occupying bladder. The system can be self-trained for each individual first and then start being used during clinical monitoring. This study considers the volume of liquid within the cavity at the time of sensation-to-void as the maximum capacity of the bladder.

This article is organized as follows: succeeding to introduction section, section 2 outlines the proposed scheme and explains discriminant features and the motivation to extract them. Subsequent to that, section 3 delves with introducing the novel decision support system which combines fuzzy inference system and error correcting output codes to fuse, interpret and quantify the extracted data. Section 4 is summarising the data sampling experiments. The comparison results and analysis are noted in section 5. Finally, we conclude with discussion in section 6.