The following is a summary of “Development of an Artificial Intelligence Model Based Solely on Computer Tomography Successfully Predicts Which Patients Will Pass Obstructing Ureteral Calculi,” published in the APRIL 2023 issue of Urology by Katz, et al.

For a study, researchers sought to develop a machine learning algorithm using a 3D-convolutional neural network (CNN) to predict the passage of obstructing ureteral stones based solely on CT scan images at the initial presentation, aiming to improve upon previous prediction models.

A retrospective study was conducted after obtaining Institutional Review Board approval. Data were extracted from patients with obstructing ureteral stones measuring 3-10 mm. Only patients with sufficient data categorizing them as having either passed or failed to pass the stone were included. A 3D-CNN model was developed using a dynamic learning rate, the Adam optimizer, and early stopping with 10-fold cross-validation. The model’s performance was evaluated by calculating the area under the curve (AUC) and developing a model confusion matrix, which was then compared to a model based solely on the largest dimension of the stone.

A total of 138 patients met the inclusion criteria and had suitable images for preprocessing and inclusion in the study. Among them, 70 patients failed to pass their ureteral stones, while 68 patients successfully passed their stones. The 3D-CNN model achieved a mean AUC of 0.95, with an overall mean sensitivity of 95% and a mean specificity of 77%. This outperformed the model based on stone size alone.

The developed 3D-CNN model can predict the passage of obstructing ureteral stones based solely on CT scan images without requiring additional measurements. The model provided valuable clinical information that can help avoid unnecessary delays in care for patients who will eventually require surgical intervention.

Source: goldjournal.net/article/S0090-4295(23)00072-9/fulltext