A Fully Automated Artificial Intelligence-Based Approach to Predict Renal Function after Radical or Partial Nephrectomy - Beyond the Abstract
A total of 293 patients (111 RN and 182 PN) who had preoperative CT with solitary renal mass and underwent extirpative surgery for renal tumors (2010-2018) were included. Preoperative GFR was collected just before surgery, and NB-GFR was 3-12 months postoperatively. Split-renal function (SRF) was determined in a fully automated way from preoperative CT, combining our deep learning segmentation model, then using those segmentation masks to estimate postoperative GFR=1.24×GFRPre-RN×SRFContralateral for RN and 89% of GFRpreoperative for PN.
The median age was 60 years, with 41% being female, and 62% undergoing PN. The median tumor size was 4.2 cm, and 92% were malignant. The correlation coefficients for the AI and clinical models, compared to the measured postoperative GFR, were 0.75 and 0.77, respectively. The AI and clinical models both performed similarly in predicting GFR <45 ml/min/1.73 m², with areas under the curve of 0.89 and 0.9,0, respectively.
The fully automated prediction of new baseline renal function demonstrates accuracy comparable to that of a validated clinical model. These AI-generated predictions can be utilized for decision-making without requiring extensive clinical details, significant clinician time, or additional measurements.
Written by: Chalairat Suk-Ouichai, MD, MPH, Department of Surgery, Siriraj Hospital, Mahidol University, Bangkok, Thailand
Read the Abstract