Can artificial intelligence support or even replace physicians in measuring sagittal balance?

European Spine Journal (2022) 31:1943–1951

Sagittal balance (SB) plays an important role in the surgical treatment of spinal disorders. The aim of this research study is to provide a detailed evaluation of a new, fully automated algorithm based on artificial intelligence (AI) for the determination of SB parameters on a large number of patients with and without instrumentation.

Methods Pre- and postoperative sagittal full body radiographs of 170 patients were measured by two human raters, twice by one rater and by the AI algorithm which determined: pelvic incidence, pelvic tilt, sacral slope, L1-S1 lordosis, T4-T12 thoracic kyphosis (TK) and the spino-sacral angle (SSA). To evaluate the agreement between human raters and AI, the mean error (95% confidence interval (CI)), standard deviation and an intra- and inter-rater reliability was conducted using intra-class correlation (ICC) coefficients.

Results ICC values for the assessment of the intra- (range: 0.88–0.97) and inter-rater (0.86–0.97) reliability of human raters are excellent. The algorithm is able to determine all parameters in 95% of all pre- and in 91% of all postoperative images with excellent ICC values (PreOP-range: 0.83–0.91, PostOP: 0.72–0.89). Mean errors are smallest for the SSA (PreOP: −0.1° (95%-CI: −0.9°–0.6°); PostOP: −0.5° (−1.4°–0.4°)) and largest for TK (7.0° (6.1°–7.8°); 7.1° (6.1°–8.1°)).

Conclusion A new, fully automated algorithm that determines SB parameters has excellent reliability and agreement with human raters, particularly on preoperative full spine images. The presented solution will relieve physicians from timeconsuming routine work of measuring SB parameters and allow the analysis of large databases efficiently.

Coronal balance with circumferential minimally invasive spinal deformity surgery for the treatment of degenerative scoliosis

J Neurosurg Spine 34:879–887, 2021

Coronal malalignment (CM) in adult spinal deformity is associated with poor outcomes and remains underappreciated in the literature. Recent attempts at classifying CM indicate that some coronal shifts may be more difficult to treat than others. To date, outcomes for circumferential minimally invasive surgery (cMIS) of the spine in the context of these new CM classifications are unreported.

METHODS A retrospective evaluation of patients with degenerative scoliosis (Cobb angle > 20 ) consecutively treated with cMIS at a single institution was performed. Preoperative and 1-year postoperative standing radiographs were used to make the comparisons. Clinical outcome measures were compared. Patients were subgrouped according to the preoperative distance between their C7 plumb line and central sacral vertical line (C7-CSVL) as either coronally aligned (type A, C7-CSVL < 3 cm); shifted ≥ 3 cm toward the concavity (type B); or shifted ≥ 3 cm toward the convexity (type C) of the main lumbar curve.

RESULTS Forty-two patients were included (mean age 67.7 years). Twenty-six patients (62%) were classified as type A, 5 patients (12%) as type B, and 11 patients (26%) as type C. An average of 4.9 segments were treated. No type A patients developed postoperative CM. All type B patients had CM correction. Six of the 11 type C patients had CM after surgery. Overall, there was an improvement in the C7-CSVL (from 2.4 to 1.8 cm, p = 0.04). Among subgroups, only type B patients improved (from 4.5 to 0.8 cm, p = 0.002); no difference was seen for type A patients (from 1.2 to 1.4 cm, p = 0.32) or type C patients (from 4.3 to 3.1 cm, p = 0.11). Comparing type C patients with postoperative CM versus those without postoperative CM, patients with CM had worse visual analog scale back scores at 1 year (5 vs 1, p = 0.01). Moreover, they had higher postoperative L4 tilt angles (11  vs 5 , p = 0.01), indicating inadequate correction of the lumbosacral fractional curve.

CONCLUSIONS cMIS improved coronal alignment, curve magnitudes, and clinical outcomes among patients with degenerative scoliosis. It did not result in CM in type A patients and was successful at improving the C7-CSVL in type B patients. Type C patients remain the most difficult to treat coronally, with worse visual analog scale back pain scores in those with postoperative CM. Regional coronal restoration of the lumbosacral fracture curve should be the focus of correction in cMIS for these patients.


Asymmetrical pedicle subtraction osteotomy for correction of concurrent sagittal-coronal imbalance in adult spinal deformity

J Neurosurg Spine 33:822–829, 2020

Rigid multiplanar thoracolumbar adult spinal deformity (ASD) cases are challenging and many require a 3-column osteotomy (3CO), specifically asymmetrical pedicle subtraction osteotomy (APSO). The outcomes and additional risks of performing APSO for the correction of concurrent sagittal-coronal deformity have yet to be adequately studied.

METHODS The authors performed a retrospective review of all ASD patients who underwent 3CO during the period from 2006 to 2019. All cases involved either isolated sagittal deformity (patients underwent standard PSO) or concurrent sagittal-coronal deformity (coronal vertical axis [CVA] ≥ 4.0 cm; patients underwent APSO). Perioperative and 2-year follow-up outcomes were compared between patients with isolated sagittal imbalance who underwent PSO and those with concurrent sagittal-coronal imbalance who underwent APSO.

RESULTS A total of 390 patients were included: 338 who underwent PSO and 52 who underwent APSO. The mean patient age was 64.6 years, and 65.1% of patients were female. APSO patients required significantly more fusions with upper instrumented vertebrae (UIV) in the upper thoracic spine (63.5% vs 43.3%, p = 0.007). Radiographically, APSO patients had greater deformity with more severe preoperative sagittal and coronal imbalance: sagittal vertical axis (SVA) 13.0 versus 10.7 cm (p = 0.042) and CVA 6.1 versus 1.2 cm (p < 0.001). In APSO cases, significant correction and normalization were achieved (SVA 13.0–3.1 cm, CVA 6.1–2.0 cm, lumbar lordosis [LL] 26.3°–49.4°, pelvic tilt [PT] 38.0°–20.4°, and scoliosis 25.0°–10.4°, p < 0.001). The overall perioperative complication rate was 34.9%. There were no significant differences between PSO and APSO patients in rates of complications (overall 33.7% vs 42.3%, p = 0.227; neurological 5.9% vs 3.9%, p = 0.547; medical 20.7% vs 25.0%, p = 0.482; and surgical 6.5% vs 11.5%, p = 0.191, respectively). However, the APSO group required significantly longer stays in the ICU (3.1 vs 2.3 days, p = 0.047) and hospital (10.8 vs 8.3 days, p = 0.002). At the 2-year follow-up, there were no significant differences in mechanical complications, including proximal junctional kyphosis (p = 0.352), pseudarthrosis (p = 0.980), rod fracture (p = 0.852), and reoperation (p = 0.600).

CONCLUSIONS ASD patients with significant coronal imbalance often have severe concurrent sagittal deformity. APSO is a powerful and effective technique to achieve multiplanar correction without higher risk of morbidity and complications compared with PSO for sagittal imbalance. However, APSO is associated with slightly longer ICU and hospital stays.

Intermediate-term clinical and radiographic outcomes with less invasive adult spinal deformity surgery

Acta Neurochirurgica (2020) 162:1393–1400

Little information exists regarding longer-term outcomes with minimally invasive spine surgery (MISS), particularly regarding long-segment and deformity procedures. We aimed to evaluate intermediate-term outcomes of MISS for adult spinal deformity (ASD).

Methods This retrospective review of a prospectively collected multicenter database examined outcomes at 4 or more years following circumferential MIS (cMIS) or hybrid (HYB) surgery for ASD. A total of 53 patients at 8 academic centers satisfied the following inclusion criteria: age > 18 years and coronal Cobb > 20°, pelvic incidence-lumbar lordosis (PI-LL) > 10°, or sagittal vertical axis (SVA) > 5 cm.

Results Radiographic outcomes demonstrated improvements of PI-LL from 16.8° preoperatively to 10.8° and coronal Cobb angle from38° preoperatively to 18.2° at 4 years. The incidence of complications over the follow-up periodwas 56.6%.A total of 21 (39.6%) patients underwent reoperation in the thoracolumbar spine, most commonly for adjacent level disease or proximal junctional kyphosis, which occurred in 11 (20.8%) patients. Mean Oswestry Disability Index (ODI) at baseline and years 1 through 4 were 49.9, 33.1, 30.2, 32.7, and 35.0, respectively. The percentage of patients meeting minimal clinically important difference (MCID) (defined as 12% or more from baseline) decreased over time, with leg pain reduction more durable than back pain reduction.

Conclusions Intermediate-term clinical and radiographic improvement following MISS for ASD is sustained, but extent of improvement lessens over time. Outcome variability exists within a subset of patients not meeting MCID, which increases over time after year two. Loss of improvement over time wasmore notable in back than leg pain. However, average ODI improvement meets MCID at 4 years after MIS ASD surgery.

A New Piece of the Puzzle to Understand Cervical Sagittal Alignment: Utilizing a Novel Angle δ to Describe the Relationship among T1 Vertebral Body Slope, Cervical Lordosis, and Cervical Sagittal Alignment

Neurosurgery 2019 DOI:10.1093/neuros/nyz088

Cervical alignment has become increasingly important in the planning of spine surgery. A relationship between the slope of T1 (T1S), the cervical lordosis (CL), and the overall cervical sagittal vertical axis (cSVA) has previously been demonstrated, but the exact nature of this relationship is poorly understood. In this study, we derive theoretical and empirical equations to better understand how T1S and CL affect cSVA. The first equation was developed on a theoretical basis using inherent trigonometric relationships of the cervical spine. By treating the cervical spine as the arc of a circumference, and by taking into account the cervical height (CH), the geometric relationship between theT1S, CL, and cSVA was described via a trigonometric identity utilizing a novel angle δ subtended by the CH and cSVA (δ = T1S-CL/2). The second equation was developed on an empiric basis by performing a multiple linear regression on data obtained from a retrospective review of a large multicenter deformity database. The theoretical equation determined that the value of cSVA could be expressed as: cSVA = CH ∗ tan(π/180 ∗ (T 1S −CL/2)) . The empirical equation determined that value of cSVA could be expressed as: cSVA = (1.1 ∗ T 1) −(0.43 ∗ CL) + 6.69. In both, the sagittal alignment of the head over the shoulders is directly proportional to the T1S and inversely proportional to CL/2. These 2 equations may allow surgeons to better understand how the CL compensates for the T1S, to accurately predict the postoperative cSVA, and to customize cervical interbody grafts by taking into consideration each individual patient’s specific cervical spine parameters.

Outcomes of Operative Treatment for Adult Cervical Deformity

Neurosurgery 83:1031–1039, 2018

Despite the potential for profound impact of adult cervical deformity (ACD) on function and health-related quality of life (HRQOL), there are few high-quality studies that assess outcomes of surgical treatment for these patients.

OBJECTIVE: To determine the impact of surgical treatment for ACD on HRQOL.

METHODS: We conducted a prospective cohort study of surgically treated ACD patients eligible for 1-yr follow-up. Baseline deformity characteristics, surgical parameters, and 1-yr HRQOL outcomes were assessed.

RESULTS: Of 77 ACD patients, 55 (71%) had 1-yr follow-up (64% women, mean age of 62 yr, mean Charlson Comorbidity Index of 0.6, previous cervical surgery in 47%). Diagnoses included cervical sagittal imbalance (56%), cervical kyphosis (55%), proximal junctional kyphosis (7%) and coronal deformity (9%). Posterior fusion was performed in 85% (mean levels = 10), and anterior fusion was performed in 53% (mean levels = 5). Three-column osteotomy was performed in 24%of patients.One year following surgery, ACD patients had significant improvement in Neck Disability Index (50.5 to 38.0, P<.001), neck pain numeric rating scale score (6.9 to 4.3, P<.001), EuroQol 5 dimension (EQ-5D) index (0.51 to 0.66, P< .001), and EQ-5D subscores: mobility (1.9 to 1.7, P=.019), usual activities (2.2 to 1.9, P=.007), pain/discomfort (2.4 to 2.1, P < .001), anxiety/depression (1.8 to 1.5, P = .014).

CONCLUSION: Based on a prospective multicenter series of ACD patients, surgical treatment provided significant improvement in multiple measures of pain and function, including Neck Disability Index, neck pain numeric rating scale score, and EQ-5D. Further follow-up will be necessary to assess the long-term durability of these improved outcomes.


The value of sitting radiographs: analysis of spine flexibility and its utility in preoperative planning for adult spinal deformity surgery

J Neurosurg Spine 29:414–421, 2018

Preoperative planning of thoracolumbar deformity (TLD) surgery has been shown to improve radiographic and clinical outcomes. One of the confounders in attaining optimal postoperative alignment is the reciprocal hyperkyphosis of unfused thoracic segments. Traditional planning utilizes standing radiographs, but the value of sitting radiographs to predict thoracic flexibility has not been investigated. Authors of the present study propose that alignment changes from a sitting to a standing position will predict changes in unfused thoracic segments after TLD correction.

METHODS Patients with degenerative spine pathology underwent preoperative sitting and standing full-spine stereotactic radiography. A subset of TLD patients who had undergone corrections with minimum T10-pelvis fusions was analyzed in terms of pre- to postoperative alignment. Radiographic parameters were analyzed, including the T1 pelvic angle (TPA), T1–L1 pelvic angle (TLPA), lumbar pelvic angle (LPA), pelvic tilt (PT), mismatch between pelvic incidence and lumbar lordosis (PI-LL mismatch), and T2–12 kyphosis (TK). Thoracic compensation was calculated as the expected TK minus actual TK (i.e., [2/3 × PI] – actual TK). Statistical analysis consisted of paired and unpaired t-tests and linear regression analysis.

RESULTS The authors retrospectively identified 137 patients with full-body standing and sitting radiographs. The mean age of the patients was 60.9 years old, 60.0% were female, and the mean BMI was 27.8 kg/m2. The patients demonstrated significantly different radiographic alignments in baseline spinopelvic and global parameters from the preoperative sitting versus the standing positions: LL (-34.20° vs -47.87°, p < 0.001), PT (28.31° vs 17.50°, p < 0.001), TPA (27.85° vs 16.89°, p < 0.001), TLPA (10.63° vs 5.17°, p < 0.001), and LPA (15.86° vs 9.67°, p < 0.001). Twenty patients (65.0% female) with a mean age of 65.3 years and mean BMI of 30.2 kg/m2 had TLD and underwent surgical correction (pre- to postoperative standing change in TPA: 33.90° to 24.50°, p = 0.001). Preoperative sitting radiographs demonstrated significant differences in alignment compared to postoperative standing radiographs: larger TPA (39.10° vs 24.50°, p < 0.001), PT (35.40° vs 28.10°, p < 0.001), LL (-11.20° vs -44.80°, p < 0.001), LPA (22.80° vs 14.20°, p < 0.001), and unfused Cobb (T2 to upper instrumented vertebra [UIV] Cobb angle: 19.95° vs 27.50°, p = 0.039). Also in the TLD group, mean thoracic compensation was 6.75°. In the linear regression analysis, the change from sitting to standing predicted pre- to postoperative changes for TK and the unfused thoracic component of TPA (5° change in preoperative sitting to preoperative standing corresponded to a pre- to postoperative change in standing TK of 6.35° and in standing TPA of 7.23°, R2 = 0.30 and 0.38, respectively).

CONCLUSIONS Sitting radiographs were useful in demonstrating spine flexibility. Among the TLD surgery group, relaxation of the unfused thoracic spine in the sitting position predicted the postoperative increase in kyphosis of the unfused thoracic segments. Sitting radiographs are a useful tool to anticipate reciprocal changes in thoracic alignment that diminish global corrections.


After 9 Years of 3-Column Osteotomies, Are We Doing Better? Performance Curve Analysis of 573 Surgeries With 2-Year Follow-up

Neurosurgery 83:69–75, 2018

In spinal deformity treatment, the increased utilization of 3-column (3CO) osteotomies reflects greater comfort and better training among surgeons. This study aims to evaluate the longitudinal performance and adverse events (complications or revisions) for a multicenter group following a decade of 3CO.

OBJECTIVE: To investigate if performance of 3CO surgeries improves with years of practice. METHODS:Patientswhounderwent 3COfor spinal deformity with intra/postoperative and revision data collected up to 2 yrwere included. Patientswere chronologically divided into 4 even groups. Demographics, baseline deformity/correction, and surgical metrics were compared using Student t-test. Postoperative and revision rates were compared using Chi-square analysis.

RESULTS: Five hundred seventy-three patients were stratified into: G1 (n=143, 2004-2008), G2 (n = 142, 2008-2009), G3 (n = 144, 2009-2010), G4 (n = 144 2010-2013). The most recent patients were more disabled by Oswestry disability index (G4=49.2 vs G1=38.3, P=.001), and received a larger osteotomy resection (G4 = 26◦ vs G1 = 20◦, P = .011) than the earliest group. There was a decrease in revision rate (45%, 35%, 33%, 30%, P = .039), notably in revisions for pseudarthrosis (16.7% G1 vs 6.9% G4, P = .007). Major complication rates also decreased (57%, 50%, 46%, 39%, P = .023) as did excessive blood loss (>4 L, 27.2 vs 16.7%, P =.023) and bladder/bowel deficit (4.2% vs 0.7% P=.002). Successful outcomes (no complications or revision) significantly increased (P = .001).

CONCLUSION:Over 9 yr, 3COs are being performed on an increasingly disabled population while gaining a greater correction at the osteotomy site. Revisions and complication rate decreased while success rate improved during the 2-yr follow-up period.

Design and Testing of 2 Novel Scores That Predict Global Sagittal Alignment Utilizing Cervical or Lumbar Plain Radiographs

Neurosurgery 82:163–171, 2018

Global sagittal deformity is an established cause of disability. However, measurements of sagittal alignment are often ignored when patients present with symptoms localizing to the cervical or lumbar spine.

OBJECTIVE: To develop scoring scales to predict the risk of sagittal malalignment in patients with only cervical or lumbar spine radiographs.

METHODS: A retrospective review of a prospectively maintained multicenter adult spinal deformity database was performed. Primary outcome (sagittal malalignment) was defined as a C7 plumbline ≥ 50 mm. Two multivariate logistic regressions were performed using patient characteristics and measurements derived from cervical or lumbar radiographs as covariates. Point scores were assigned to age, body mass index (BMI), and lumbar lordosis or T1 slope by rounding their ß coefficients to the nearest integer.

RESULTS: Nine hundred seventy-nine patients were included, with 652 randomly assigned to the derivation cohort (used to build the score) and 327 comprising the validation set. Final cervical score for the primary outcome included BMI ≥ 25 (1 point), age ≥ 55 yr (2 points), and T1 slope ≥ 27º (2 points). Final lumbar score for the primary outcome included BMI≥25 (1 point), age≥55 yr (1 point), and lumbar lordosis ≥45º (–1 points). High scores for both the cervical and lumbar spine presented with high specificity and positive likelihood ratios of sagittal malalignment.

CONCLUSION: We developed scoring scales to predict global sagittal malalignment utilizing clinical covariates and cervical or lumbar radiographs. Patients with high scores may prompt imaging with long-cassette plain films to evaluate for global sagittal imbalance.

The Impact of Standing Regional Cervical Sagittal Alignment on Outcomes in Posterior Cervical Fusion Surgery

Neurosurgery 71:662–669, 2012

Positive spinal regional and global sagittal malalignment has been repeatedly shown to correlate with pain and disability in thoracolumbar fusion.

OBJECTIVE: To evaluate the relationship between regional cervical sagittal alignment and postoperative outcomes for patients receiving multilevel cervical posterior fusion.

METHODS: From 2006 to 2010, 113 patients received multilevel posterior cervical fusion for cervical stenosis, myelopathy, and kyphosis. Radiographic measurements made at intermediate follow-up included the following: (1) C1-C2 lordosis, (2) C2-C7 lordosis, (3) C2-C7 sagittal vertical axis (C2-C7 SVA; distance between C2 plumb line and C7), (4) center of gravity of head SVA (CGH-C7 SVA), and (5) C1-C7 SVA. Health-related quality-of-life measures included neck disability index (NDI), visual analog pain scale, and SF-36 physical component scores. Pearson product-moment correlation coefficients were calculated between pairs of radiographic measures and health-related quality-oflife scores.

RESULTS: Both C2-C7 SVA and CGH-C7 SVA negatively correlated with SF-36 physical component scores (r = -0.43, P <.001 and r = -0.36, P = .005, respectively). C2-C7 SVA positively correlated with NDI scores (r = 0.20, P = .036). C2-C7 SVA positively correlated with C1-C2 lordosis (r = 0.33, P = .001). For significant correlations between C2-C7 SVA and NDI scores, regression models predicted a threshold C2-C7 SVA value of approximately 40 mm, beyond which correlations were most significant.

CONCLUSION: Our findings demonstrate that, similar to the thoracolumbar spine, the severity of disability increases with positive sagittal malalignment following surgical reconstruction.

Short-term Progressive Spinal Deformity Following Laminoplasty Versus Laminectomy for Resection of Intradural Spinal Tumors: Analysis of 238 Patients

Neurosurgery 66:1005-1012, 2010 DOI: 10.1227/01.NEU.0000367721.73220.C9

Gross total resection of intradural spinal tumors can be achieved in the majority of cases with preservation of long-term neurological function. However, postoperative progressive spinal deformity complicates outcome in a subset of patients after surgery. We set out to determine whether the use of laminoplasty (LP) vs laminectomy (LM) has reduced the incidence of subsequent spinal deformity following intradural tumor resection at our institution.

METHODS:We retrospectively reviewed the records of 238 consecutive patients undergoing resection of intradural tumor at a single institution. The incidence of subsequent progressive kyphosis or scoliosis, perioperative morbidity, and neurological outcome were compared between the LP and LM cohorts.

RESULTS: One hundred eighty patients underwent LM and 58 underwent LP. Patients were 46 ± 19 years old with median modified McCormick score of 2. Tumors were intramedullary in 102 (43%) and extramedullary in 102 (43%). All baseline clinical, radiographic, and operative variables were similar between the LP and LM cohorts. LP was associated with a decreased mean length of hospitalization (5 vs 7 days; P = .002) and trend of decreased incisional cerebrospinal fluid leak (3% vs 9%; P = .14). Following LP vs LM, 5 (9%) vs 21 (12%) patients developed progressive deformity (P = .728) a mean of 14 months after surgery. The incidence of progressive deformity was also similar between LP vs LM in pediatric patients < 18 years of age (43% vs 36%), with preoperative scoliosis or loss of cervical/lumbar lordosis (28% vs 22%), or with intramedullary tumors (11% vs 11%).

CONCLUSION: LP for the resection of intradural spinal tumors was not associated with a decreased incidence of short-term progressive spinal deformity or improved neurological function. However, LP may be associated with a reduction in incisional cerebrospinal fluid leak. Longer-term follow-up is warranted to definitively assess the long-term effect of LP and the risk of deformity over time.