Background　and　objective:　Pediatric　hydronephrosis　is　a　common　disease　due　to　ureteropelvic　junction　obstruction　(UPJO).　Non-invasive　magnetic　resonance　urography　(MRU)　can　utilize　the　water　in　the　urine　as　a　natural　contrast　agent　to　image　renal　pelvis　and　calyces,　which　could　be　used　to　diagnose　UPJO.　However,　as　invasive　techniques,　MRU　fails　to　provide　detailed　definitive　measurements,　such　as　pressure　and　velocity.　This　study　aims　to　provide　a　non-invasive　method　to　compute　urine　flow　dynamics　based　on　deep　learning　and　computational　fluid　dynamics　(CFD)　techniques.　The　method　could　gain　insights　into　the　urodynamic　characteristics　of　hydronephrosis　patients,　which　facilitates　clinical　decision-making　and　further　contributes　to　prognosis.　Methods:　This　work　proposed　a　method　to　characterize　urine　flow　dynamics　with　MRU　data.　Specifically,　the　method　first　reconstructs　patient-specific　three-dimensional　renal　pelvis　mesh　models　with　deep　learning　techniques.　Then,　boundary　conditions　were　set　according　to　the　mesh　models,　and　the　Viscous　k-epsilon　turbulence　model　was　used　to　perform　CFD　simulation　of　urine.　Finally,　the　method　characterizes　the　dynamics　of　urinary　flows　for　UPJO　diagnosis　and　prognosis.　Results:　Experiments　are　conducted　on　MRU　data　via　our　proposed　method.　By　comparing　the　preoperative　and　postoperative　urine　pressure　cloud　maps　of　patients,　we　found　that　the　pressure　in　the　renal　pelvis　was　significantly　lower　in　the　postoperative　period　than　in　the　preoperative　period.　Furthermore,　we　introduced　the　urine　flow　velocity　ratio　(UFVR)　index　to　assess　the　obstruction　of　urine　in　renal　pelvis.　We　compared　the　UFVR　in　the　preoperative　and　postoperative　phase,　and　indicated　that　the　obstruction　of　urine　flow　has　been　mitigated　following　surgery.　Conclusions:　CFD　can　simulate　the　urinary　flow　field　within　the　renal　pelvis,　offering　the　possibility　of　predicting　renal　pelvic　pressure　without　the　need　for　painful,　invasive　interventions　such　as　the　Whitaker　test.　Moreover,　numerical　simulation　provides　a　new　definitive　way　for　evaluating　the　postoperative　efficacy　of　dismembered　pyeloplasty.　©　2025　Elsevier　B.V.