Objective To investigate the feasibility of predicting proximal junctional kyphosis (PJK) in adults after spinal deformity surgery based on back-forward Bending CT localization images and related predictive indicators. Methods A retrospective analysis was performed for 31 adult patients with spinal deformity who underwent posterior osteotomy and long-segment fusion fixation between March 2017 and March 2020. There were 5 males and 26 females with an average age of 62.5 years (range, 30-77 years). The upper instrumented vertebrae (UIV) located at T5 in 1 case, T6 in 1 case, T9 in 13 cases, T10 in 12 cases, and T11 in 4 cases. The lowest instrumented vertebrae (LIV) located at L1 in 3 cases, L2 in 3 cases, L3 in 10 cases, L4 in 7 cases, L5 in 5 cases, and S1 in 3 cases. Based on the full-length lateral X-ray film of the spine in the standing position before and after operation and back-forward Bending CT localization images before operation, the sagittal sequence of the spine was obtained, and the relevant indexes were measured, including thoracic kyphosis (TK), lumbar lordosis (LL), local kyphosis Cobb angle (LKCA) [the difference between the different positions before operation (recovery value) was calculated], kyphosis flexibility, hyperextension sagittal vertical axis (hSVA), T2-L5 hyperextension C7-vertebral sagittal offset (hC7-VSO), and pre- and post-operative proximal junctional angle (PJA). At last follow-up, the patients were divided into PJK and non-PJK groups based on PJA to determine whether they had PJK. The gender, age, body mass index (BMI), number of fusion segments, number of cases with coronal plane deformity, bone mineral density (T value), UIV position, LIV position, operation time, intraoperative blood loss, osteotomy grading, and related imaging indicators were compared between the two groups. The hC7-VSO of the vertebral body with significant differences between groups was taken, and the receiver operating characteristic curve (ROC) was used to evaluate its accuracy in predicting the occurrence of PJK. Results All 31 patients were followed up 13-52 months, with an average of 30.0 months. The patient’s PJA was 1.4°-29.0° at last follow-up, with an average of 10.4°; PJK occurred in 8 cases (25.8%). There was no significant difference in gender, age, BMI, number of fusion segments, number of cases with coronal plane deformity, bone mineral density (T value), UIV position, LIV position, operation time, intraoperative blood loss, and osteotomy grading between the two groups (P>0.05). Imaging measurements showed that the LL recovery value and T8-L3 vertebral hC7-VSO in the PJK group were significantly higher than those in the non-PJK group (P>0.05). There was no significant difference in hyperextension TK, hyperextension LL, hyperextension LKCA, TK recovery value, LL recovery value, kyphosis flexibility, hSVA, and T2-T7, L4, L5 vertebral hC7-VSO (P>0.05). T8-L3 vertebral hC7-VSO was analyzed for ROC curve, and combined with the area under curve and the comprehensive evaluation of sensitivity and specificity, the best predictive index was hC7-L2, the cut-off value was 2.54 cm, the sensitivity was 100%, and the specificity was 60.9%. Conclusion Preoperative back-forward Bending CT localization image can be used to predict the occurrence of PJK after posterior osteotomy and long-segment fusion fixation in adult spinal deformity. If the patient’s T8-L2 vertebral hC7-VSO is too large, it indicates a higher risk of postoperative PJK. The best predictive index is hC7-L2, and the cut-off value is 2.54 cm.