Pressure-flow studies are the gold standard for diagnosing bladder outlet obstruction. Still, approximately one-fourth of patients who had prostatectomy do not present symptom amelioration, and, therefore, there would be space for diagnosis improvement. To avoid the disadvantages of invasive pressure-flow studies, which can even carry some morbidity for the patient, noninvasive methods have been suggested . In this study, we will use a urethral connector, a minimally invasive method, to measure pressure-flow data. 
This minimally invasive urodynamics (MIU) device has a conical shape part made of polyvinyl carbon inserted into the fossa navicularis (1-2 cm) which avoids leakage and minimizes use discomfort. The second part has an outlet to connect to the external pressure transducer, and the two parts are connected through a central canal which allows urine to flow [1,2]. Since MIU was already proven to be an efficient method when compared with the invasive gold standard [3] and is already commercially available, a method for pressure-flow data analysis based on MIU data could be used in clinical practice.
Non-invasive and MIU methods rely on the interruption of urinary flow and the measurement of the bladder pressure transmitted along the fluid column between the bladder and the site of urethral occlusion, i.e., isovolumetric pressure (Piso). This pressure differs from the detrusor pressure measured with the invasive procedure, and therefore, using Abrams-Griffiths or the ICS nomogram to classify pressure-flow data is not the most efficient method for data analysis.
Therefore, we propose a nomogram derived from an unsupervised machine-learning algorithm that separates data into two classes, of obstruction and unobstruction. The classification obtained from the developed nomogram will be compared to the gold-standard invasive method.

This study received ethical approval (Nº 1017/2008) for the measurement of pressure-flow data with MIU. When the patient reported reaching the maximum bladder capacity, micturition could occur. The patient would then start micturition and would manually occlude the device outlet for a few seconds wearing gloves. At this moment, Piso could be measured. Interrupted urinary flow rate (Qinter) was measured after releasing the occlusion of the device. Participants also had pressure-flow data measured with the invasive method. These data were obtained by urodynamics performed with a 6F rectal balloon catheter, for abdominal pressure measurement, an 8F catheter, introduced into the urinary bladder for saline solution infusion, and a 6F catheter used for vesical pressure measurement. Urodynamics was performed in accordance with good urodynamic practices . Before voiding, the 8F catheter was removed. Flow parameters and intravesical and abdominal pressures were recorded simultaneously during micturition. There was a total of 68 male participants.
Pressure-flow data from MIU were classified as obstructed and unobstructed according to the k-means clustering algorithm, an unsupervised machine-learning method that learns from data that were not classified previously. The separation of data into these two clusters occurs using the objects’ mean values. We used as an initial random state for the centroids the value of 80 for reproducibility. The training dataset had a total of 45 cases. With the separation of data into obstructed and nonobstructed classes, a linear regression from data classified as obstructed was used to determine the limit of the nomogram to be used. If (Piso -  7.4 x Qinter)> 30.4, the case is of obstruction, otherwise it is unobstructed (Figure 1). The nomogram developed was tested on 23 pressure-flow data.

With the nomogram based on the k-means algorithm, from the 12 cases classified as unobstructed by the gold standard, 11 cases measured with the MIU were coincident, which represents 92% specificity. From 11 cases of obstruction from the invasive method, 9 MIU cases were also obstructed, with an 82% sensitivity.

As can be seen, the classification of pressure-flow data measured with MIU and evaluated by the nomogram developed is highly coincident with the gold standard invasive urodynamics. The only known nomogram developed for noninvasive methods is the one developed by Griffiths et al. [3], which is based on noninvasive penile cuff pressure-flow data. When we classify MIU pressure-flow data, although from the 12 cases of nonobstruction by the gold standard method, all of them were classified as normal by the Griffiths modified nomogram (100% specificity), from the 11 cases of obstruction as from the gold standard method, only 5 MIU cases were classified as obstructed (45% of sensitivity). Therefore, the nomogram developed based on the k-means algorithm is most appropriate for MIU data analysis.

It is important to have a nomogram that is efficient for evaluating noninvasive and MIU data since Piso differs from pressure measured with invasive urodynamics. Moreover, results can be close to what would be expected by the invasive gold standard, but without the difficulties that make this method a painful, time-consuming, and sometimes dangerous method for the patient. The urethral connector could even be used for periodic follow-up of patients that have just mild symptoms and the nomogram developed could be used on the routine for analyzing this type of pressure-flow data.

D’Ancona CAL, Bassani JWM, Querne FAO, Carvalho J, Oliveira RRM, Netto NR Jr. New method for minimally invasive assessment in men with lower urinary tract symptoms. Urol 2008; 71:75–78.
de Almeida JCM, D’Ancona CAL, Bassani JWM. Minimally invasive measurement of vesical pressure for diagnosis of infravesical obstruction. Neurourol Urodyn 2018; 37(2):849-853.
Griffiths CJ, Harding C, Blake C, McIntosh S, Drinnan MJ, Robson WA, et al. A nomogram to classify men with lower urinary tract symptoms using urine flow and noninvasive measurement of bladder pressure. J Urol 2005; 174:1323-1326.

Continence 12S (2024) 101460
DOI: 10.1016/j.cont.2024.101460