Developing a new electromyography-based algorithm to diagnose the etiology of fecal incontinence

Abstract

Purpose: For surface electromyography (sEMG) to become widely used in fecal incontinence (FI) etiology assessment, one would have to create a simple, step-by-step, computer-aided, electromyography-based algorithm that would become the basis for a computer-aided diagnosis (CAD) system. Thus, the aim of this work was to develop such an algorithm.

Methods: Each patient included in the study underwent a structured medical interview, a general physical examination, and a proctological examination. Patients that scored more than 10 points on the fecal incontinence severity index (FISI) underwent further tests that included rectoscopy, anorectal manometry, transanal ultrasonography, multichannel sEMG, and assessment of anal reflexes. Patients with fully diagnosed FI were included into the study group. The control group consisted of healthy volunteers that scored five or less points on the FISI and had no known anal sphincters dysfunction.

Results: Forty-nine patients were qualified to the study group (age ± SD 58.9 ± 13.8). The control group was number- and gender-matched (age ± SD 45.4 ± 15.1). The sensitivity and specificity of classification tree number I, to diagnose neurogenic FI, were 89.5 and 86 %, respectively. For patients with idiopathic FI, these values were 82 and 91 %, respectively. The sensitivity and specificity of classification tree number III, to diagnose neurogenic FI, were 84 and 78 %, respectively. For patients with idiopathic FI, these values were 78 and 87 %, respectively.

Conclusions: The relative simplicity and low classification costs allow to assume that algorithms based on classification trees I and III will serve to be the basis for a FI etiology CAD system.

Fig. 1

Classification tree number I (11 decision (nonterminal, blue boxes) nodes and 12 endpoints (terminal nodes, red boxes)) fecal incontinence etiology. Tree optimized taking into account minimal classification costs. Bars inside each box depict classified patients, divided by etiology (controls, neurogenic, myogenic, and idiopathic fecal incontinence). Green bars depict decisions with the numerical value below and the name of the variable above. Results presented as means. FI fecal incontinence, RMS root-means-square, MF median frequency, R external anal sphincter relaxed, C external anal sphincter maximally contracted. Numbers after RMS or MF refer to the depth of the ring that was used to obtain the measurements (1 closest to the anus; 3 the deepest ring)

Fig. 2

Classification tree number II (five decision (nonterminal, blue boxes) nodes and six endpoints (terminal nodes, red boxes)) fecal incontinence etiology. Tree optimized taking into account the minimal number of decision nodes. Bars inside each box depict classified patients, divided by etiology (controls, neurogenic, myogenic, and idiopathic fecal incontinence). Green bars depict decisions with the numerical value below and the name of the variable above. Results presented as means. FI fecal incontinence, RMS root-mean-square, MF median frequency, R external anal sphincter relaxed, C external anal sphincter maximally contracted. Numbers after RMS or MF refer to the depth of the ring that was used to obtain the measurements (1 closest to the anus; 3 the deepest ring)

Fig. 3

Classification tree number III (seven decision (nonterminal, blue boxes) nodes and eight endpoints (terminal nodes, red boxes)) fecal incontinence etiology. Tree optimized taking into account classification costs and the number of decision nodes. Bars inside each box depict classified patients, divided by etiology (controls, neurogenic, myogenic, and idiopathic fecal incontinence). Green bars depict decisions with the numerical value below and the name of the variable above. Results presented as means. FI fecal incontinence, RMS root-mean-square, MF median frequency, R external anal sphincter relaxed, C external anal sphincter maximally contracted. Numbers after RMS or MF refer to the depth of the ring that was used to obtain the measurements (1 closest to the anus; 3 the deepest ring)