A Fully Automated Artificial Intelligence-Based Approach to Predict Renal Function After Radical or Partial Nephrectomy - 24/02/25

Doi : 10.1016/j.urology.2025.01.073

Nour Abdallah ^a,^⁎ , Nityam Rathi ^a , Nicholas Heller ^b , Andrew Wood ^a , Rebecca Campbell ^a , Tarik Benidir ^a , Fabian Isensee ^c , Resha Tejpaul ^b , Chalairat Suk-ouichai ^d , Diego Aguilar Palacios ^a , Alex You ^e , Satish Viswanath ^f , Brennan Flannery ^f , Jihad Kaouk ^a , Samuel Haywood ^a , Venkatesh Krishnamurthi ^a , Nikolaos Papanikolopoulos ^b , Joseph Zabell ^g , Robert Abouassaly ^a , Erick M. Remer ^a,^h , Steven Campbell ^a , Christopher J. Weight ^a
^a Glickman Urological and Kidney Institute, Cleveland, OH
^b Department of Computer Science and Engineering, University of Minnesota, Minneapolis, MN
^c German Cancer Research Center (DKFZ) Heidelberg, University of Heidelberg, Heidelberg, Germany
^d Department of Surgery, Siriraj Hospital, Mahidol University, Bangkok, Thailand
^e Case Western Reserve University, Cleveland, OH
^f Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH
^g Department of Urology, University of Minnesota Medical School, Minneapolis, MN
^h Department of Diagnostic Radiology, Imaging Institute Cleveland Clinic, Cleveland, OH

^⁎Address correspondence to: Nour Abdallah, M.D., Research Fellow, Glickman Urological and Kidney Institute, Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH, 44195.Research Fellow, Glickman Urological and Kidney Institute, Cleveland Clinic Foundation9500 Euclid AveClevelandOH44195

Sous presse. Épreuves corrigées par l'auteur. Disponible en ligne depuis le Monday 24 February 2025
Cet article a été publié dans un numéro de la revue, cliquez ici pour y accéder

Résumé

Objective

To test if our artificial intelligence (AI)-postoperative glomerular filtration rate (GFR) prediction is as accurate as a validated clinical model. The American Urologic Association recommends estimating postoperative GFR in patients with renal masses and prioritizing partial nephrectomy (PN) when GFR would be <45 ml/minutes/1.73 m² if radical nephrectomy (RN) was performed. Previously validated models have limited clinical uptake.

Methods

We included 300 patients undergoing nephrectomy for renal tumors from the KiTS19 challenge. Preoperative GFR was collected just before surgery, and new baseline GFR 3-12 months postoperatively. Split-renal function (SRF) was determined in a fully automated way from preoperative computed tomography, combining our deep learning segmentation model, then using those segmentation masks to estimate postoperative GFR = 1.24 × GFR_Pre-RN × SRF_{Contralateral} for RN and 89% of GFR_preoperative for PN. A clinical model estimated postoperative GFR = 35 + GFR_preoperative x 0.65–18 (if RN)–age x 0.25 + 3 (if tumor>7 cm)⁻² (if diabetes). We compared the AI and clinical model GFR estimations to the measured postoperative GFR using correlation coefficients and their ability to predict GFR < 45 using logistic regression.

Results

Median age was 60 years, 41% were female, and 62% had PN. Median tumor size was 4.2 cm, and 92% were malignant. Compared to the measured postoperative GFR, correlation coefficients were 0.75 and 0.77 for the AI and clinical models, respectively. The AI and clinical models performed similarly for predicting GFR < 45 (areas under the curve 0.89 and 0.9, respectively).

Conclusion

Our fully automated prediction of new baseline renal function is as accurate as a validated clinical model without needing clinical details, clinician time, or measurements.

Le texte complet de cet article est disponible en PDF.