Study | Clinical Purpose | AI Methods | Number of Datasets/Subjects | Performance Results of Model |
Burton, 2019 | Diagnosis of urinary tract infection | Random forest; neural network; extreme gradient boosting | 212,554 reports | Classification sensitivity above 95% |
Barzegari, 2018 | Diagnosis of stress UI | MLP ANN | 6453 data points (22 patients) | N/A |
Geramipour, 2015 | Real-time estimation of bladder pressure and stimulation to reduce incontinence | MLP feed forward network levenberg-marquardt (for training) | 3 rats | Tracking performance at 8.89% (neural network model); 5-s bladder prediction error is about 10.54% |
Laurikkala, 1999 | Diagnose female urinary incontinence | Genetic algorithm-based galactica; C.4.5 | Trained on 1105 sets; tested on 485 | Mean prediction accuracy of 91% (C.4.5 rule-based) and 90% (Galactica); Mean descriptive accuracy of 92% (C.4.5 rule-based) and 88% (Galactica) |
Hung et al., 2019 | Prediction of urinary continence after robot-assisted radical prostatectomy (RARP) | Deep learning model (DeepSurv) | Automated performance metrics acquired from 100 patients | Achieved the lowest mean absolute error (MAE) of 85.9; CI of 0.6 |
Karem et al., 2016 | Classification of bladder events such as detrusor contraction for urinary incontinence diagnosis | Context-Aware Thresholding (CAT) algorithm | 14 human subjects; vesical and abdominal pressure data from 64 tracings from the human subjects | 97% accuracy (CAT); 1.34 false positivity rate (CAT) |
Huang, 2007 | Diagnosis of urodynamic Stress incontinence (USI) with computer-aided perineal ultrasound | MLP neural network with backpropagation algorithm for training | 48 women—36 with USI; 12 normal | 91.7% accuracy; 94.4% sensitivity; 94.4% positive predictive value; 83.3% negative predictive value |