Feasibility study of intelligent autonomous determination of the bladder voiding need to treat bedwetting using ultrasound and smartphone ML techniques : Intelligent autonomous treatment of bedwetting

Abstract

Unsatisfactory cure rates for the treatment of nocturnal enuresis (NE), i.e. bed-wetting, have led to the need to explore alternative modalities. New treatment methods that focus on preventing enuretic episodes by means of a pre-void alerting system could improve outcomes for children with NE in many aspects. No such technology exists currently to monitor the bladder to alarm before bed-wetting. The aim of this study is to carry out the feasibility of building, refining and evaluating a new, safe, comfortable and non-invasive wearable autonomous intelligent electronic device to monitor the bladder using a single-element low-powered low-frequency ultrasound with the help of Machine Learning techniques and to treat NE by warning the patient at the pre-void stage, enhancing quality of life for these children starting from the first use. The sensitivity and specificity values are 0.89 and 0.93 respectively for determining imminent voiding need. The results indicate that customised imminent voiding need based on the expansion of the bladder can be determined by applying a single-element transducer on a bladder in intermittent manner. The acquired results can be improved further with a comfortable non-invasive device by adding several more features to the current features employed in this pilot study. Graphical Abstract Ultrasound device design: echoed US pulses reflected from the bladder and related tissues around the bladder is detected. These pulses are analysed, and an alarm is triggered when needed to treat nocturnal enuresis.

Keywords: Bladder volume determination; Machine Learning; Nocturnal enuresis; Ultrasound physics.

Graphical Abstract

Ultrasound device design: echoed US pulses reflected from the bladder and related tissues around the bladder is detected. These pulses are analysed, and an alarm is triggered when needed to treat nocturnal enuresis.

Fig. 1

Interaction of US beams and the bladder

Fig. 2

Interaction of US beams and the bladder with respect to an empty bladder: a single-element US transducer and respective propagation and attenuation signals; the pulsed signal is almost completely reflected (99%) and the remaining propagated signal after the anterior wall of the bladder is scattered in all directions in a non-uniform manner due to emptiness within the bladder

Fig. 3

Interaction of US beams and the bladder with respect to a full bladder: single-element US transducer and respective propagation and attenuation signals where the reflection is 5% and propagation signals after the anterior wall of the bladder do not lose their strength because the urine inside the bladder causes little attenuation

Fig. 4

Ultrasound Design: the echoed US pulses reflected from the bladder and related tissues around the bladder is detected using a skin-interfacing gel between the transducer and the skin

Fig. 5

US-Key ultrasonic Transmitter/Receiver/Digitizer (Leoceur Electronique) (left) by which data can be acquired using a computing device and ISONIC utPOD device (right) by which data can be acquired using both itself and a computing device

Fig. 6

Settings: General settings and personal settings

Fig. 7

Training is aimed to be implemented at the background automatically when needed without user intervention: a selection of the training dataset and training approach to train the system; b statistical results of the three ML techniques after training using 10-fold cross-validation

Fig. 8

Testing: a selection of the test dataset (left); b statistical results of the three ML techniques and voting scheme (right)

Fig. 9

Phases in training: classifiers in four groups with respect to three ML techniques are built

Fig. 10

Phases in testing: classification of the samples are performed with respect to the comparison between the features of each sample and pre-trained models

Fig. 11

Phases in data acquisition and decision making: classification of the acquired samples are performed with respect to the comparison between the features of the sample and pre-trained models

Fig. 12

Determination of bladder status and control of the undergarment placement

Fig. 13

Se and Sp values with respect to the techniques

Fig. 14

Scenario for settings

Fig. 15

Scenario for bladder assessment

Fig. 16

Self-adhesive gel pads: both sides to stick the hypo-gastric region at the proximal side and probes at the distal side

Fig. 17

Design concept for a wearable support garment: A performance critical aspect of this technology is its adoption by the end-user across various psychophysiological requirements. This approach provides this in two critical areas, more specifically: i. Ease of application/use by a non-technical person to the body, by user or carer offering instant accurate location of device. ii. Offers a discrete undergarment opportunity that can be designed so as not to convey a medical condition in everyday life events. i.e. school environments. a General framework of the concept; b Positioning with respect to bladder; c Representation with respect to morphologic types: i. the devices shape in relationship to the pubic region, this will be graded to fit population types. ii. the garment will provide tension to the rear of the device maintaining contact with the skin. iii. the inner pocket of the proposed garment has a window which enables the self-adhesive gel pad which is adhered to the body side of the device to protrude through and adhere to the skin. Various fit ranges can be developed to accommodate population morphological types