Project overview
Kidney stone disease has a UK lifetime prevalence of ~14% and is increasing [1]. Surgical management including Flexible Ureteroscopy with Lithotripsy (FURSL) and Percutaneous Nephrolithotripsy (PCNL) is indicated for stones that fail to spontaneously pass or when interventional management (shockwave lithotripsy) is unsuccessful. However, there is no clear consensus on which technique is best for any individual patient with only a few limited published studies focusing on predictive factors for stone-free rate or postoperative complications following FURSL [2].
Machine learning (ML) /Artificial Intelligence (AI) techniques are being increasingly employed in endourology to predict treatment outcomes and complications, benefiting from large data inputs such as CT imaging, patient characteristics, and laboratory variables. At the core of this success is the ability to integrate diverse sources of information via computational and mathematical modelling. For example, in shockwave lithotripsy, machine learning techniques have demonstrated better treatment success classification than standard statistical approaches [3].
The aim of our research is to apply machine learning techniques to preoperative data to predict FURSL/PCNL outcomes . This will allow clinicians to make better-informed decisions on treatment modalities, facilitating patient-specific decision making in clinical scenarios where guidelines consider treatment modalities to be equivocal. This will also allow better-informed consent for patients, as complication rates could be more refined to their clinical presentation.
Machine learning (ML) /Artificial Intelligence (AI) techniques are being increasingly employed in endourology to predict treatment outcomes and complications, benefiting from large data inputs such as CT imaging, patient characteristics, and laboratory variables. At the core of this success is the ability to integrate diverse sources of information via computational and mathematical modelling. For example, in shockwave lithotripsy, machine learning techniques have demonstrated better treatment success classification than standard statistical approaches [3].
The aim of our research is to apply machine learning techniques to preoperative data to predict FURSL/PCNL outcomes . This will allow clinicians to make better-informed decisions on treatment modalities, facilitating patient-specific decision making in clinical scenarios where guidelines consider treatment modalities to be equivocal. This will also allow better-informed consent for patients, as complication rates could be more refined to their clinical presentation.
