Attention: Restrictions on use of AUA, AUAER, and UCF content in third party applications, including artificial intelligence technologies, such as large language models and generative AI.
You are prohibited from using or uploading content you accessed through this website into external applications, bots, software, or websites, including those using artificial intelligence technologies and infrastructure, including deep learning, machine learning and large language models and generative AI.

JU INSIGHT Identifying Bladder Phenotypes After Spinal Cord Injury With Unsupervised Machine Learning

By: Blayne Welk, MD, MSc*, Western University, London, Ontario, Canada; Tianyue Zhong, MD, MSc*, Western University, London, Ontario, Canada; Jeremy Myers, MD, University of Utah, Salt Lake City; John Stoffel, MD, University of Michigan, Ann Arbor Western University, London, Ontario, Canada; Sean Elliot, MD, MSc, University of Minnesota, Minneapolis; Sara M. Lenherr, MD, MSc, University of Utah, Salt Lake City; Daniel Lizotte, PhD, Western University, London, Ontario, Canada; *Co-first authors | Posted on: 17 Jul 2024

Welk B, Zhong T, Myers J, et al. Identifying bladder phenotypes after spinal cord injury with unsupervised machine learning: a new way to examine urinary symptoms and quality of life. J Urol. 2024;212(1):114-123. doi:10.1097/JU.0000000000003984

Study Need and Importance

Spinal cord injury (SCI) results in heterogenous and interrelated urinary and bowel symptoms, physical impairments, and psychosocial changes. Relationships between these variables have traditionally been defined by adjusted regression equations; however, with the advent of artificial intelligence, more complex models can examine how these factors are related.

What We Found

We used the Neurogenic Bladder Research Group SCI registry to identify 1263 people with SCI who had complete data on demographics, patient-reported outcome measures of pain, bowel function, autonomic dysreflexia, and positive affect/independence quality of life (QOL) scores. Using an unsupervised clustering algorithm that did not consider urinary symptoms or QOL, we found that people sorted into 4 groups (Figure), which were qualitatively described as 1, High functioning, low SCI complications; 2, Older, high SCI complications; 3, Quadriplegia with bowel/bladder morbidity; and 4, Female predominant. Some of these groups had small but significant differences in bladder symptoms at baseline, and in their change in bladder-related QOL after 1 year.

image

Figure.
Overview of the study methodology for the Neurogenic Bladder Research Group (NBRG) spinal cord injury (SCI) registry and how it was used for the current study. The 4 identified clusters are summarized in the bottom right corner, and statistically significant relationships with the outcome variables of interest are shown. ML indicates machine learning; NBSS, Neurogenic Bladder Symptom Score; QOL, quality of life. 1Compared to the “high function, low SCI complication” group. 2Compared to the “high function, low SCI complication” group and the “quadriplegia with bowel/bladder morbidity” group. 3Compared to the “high function, low SCI complication” group and the “older, high SCI complication” group.

Limitations

This was an exploratory study to determine if previous factors related to urinary symptoms and QOL may interact to form phenotypic clusters of patients. The machine learning algorithm that we used is unsupervised, so it may cluster people in way that is not clinically relevant.

Interpretation for Patient Care

This study utilized machine learning to identify phenotypical clusters among individuals with SCI; some of these clusters behaved differently with respect to QOL, both at baseline and after 1 year of follow-up. We hope that continuing this work will help target more specific groups for follow-up and QOL interventions, and that we will be able to determine if these clustering algorithms can better predict symptom and QOL outcomes compared to the use of individual risk factors.

advertisement

advertisement