Search

This paper examines a kind of explainable AI, centered around what we term pro-hoc explanations, that is a form of support that consists of offering alternative explanations (one for each possible outcome) instead of a specific post-hoc explanation following specific advice. Specifically, our support mechanism utilizes explanations by examples, featuring analogous cases for each category in a binary setting. Pro-hoc explanations are an instance of what we called frictional AI, a general class of decision support aimed at achieving a useful compromise between the increase of decision effectiveness and the mitigation of cognitive risks, such as over-reliance, automation bias and deskilling. To illustrate an instance of frictional AI, we conducted an empirical user study to investigate its impact on the task of radiological detection of vertebral fractures in x-rays. Our study engaged 16 orthopedists in a 'human-first, second-opinion' interaction protocol. In this protocol, clinicians first made initial assessments of the x-rays without AI assistance and then provided their final diagnosis after considering the pro-hoc explanations. Our findings indicate that physicians, particularly those with less experience, perceived pro-hoc XAI support as significantly beneficial, even though it did not notably enhance their diagnostic accuracy. However, their increased confidence in final diagnoses suggests a positive overall impact. Given the promisingly high effect size observed, our results advocate for further research into pro-hoc explanations specifically, and into the broader concept of frictional AI., Competing Interests: Declaration of competing interest None, (Copyright © 2024 The Authors. Published by Elsevier B.V. All rights reserved.)

This paper proposes a user study aimed at evaluating the impact of Class Activation Maps (CAMs) as an eXplainable AI (XAI) method in a radiological diagnostic task, the detection of thoracolumbar (TL) fractures from vertebral X-rays. In particular, we focus on two oft-neglected features of CAMs, that is granularity and coloring, in terms of what features, lower-level vs higher-level, should the maps highlight and adopting which coloring scheme, to bring better impact to the decision-making process, both in terms of diagnostic accuracy (that is effectiveness) and of user-centered dimensions, such as perceived confidence and utility (that is satisfaction), depending on case complexity, AI accuracy, and user expertise. Our findings show that lower-level features CAMs, which highlight more focused anatomical landmarks, are associated with higher diagnostic accuracy than higher-level features CAMs, particularly among experienced physicians. Moreover, despite the intuitive appeal of semantic CAMs, traditionally colored CAMs consistently yielded higher diagnostic accuracy across all groups. Our results challenge some prevalent assumptions in the XAI field and emphasize the importance of adopting an evidence-based and human-centered approach to design and evaluate AI- and XAI-assisted diagnostic tools. To this aim, the paper also proposes a hierarchy of evidence framework to help designers and practitioners choose the XAI solutions that optimize performance and satisfaction on the basis of the strongest evidence available or to focus on the gaps in the literature that need to be filled to move from opinionated and eminence-based research to one more based on empirical evidence and end-user work and preferences., Competing Interests: Declaration of competing interest None Declared., (Copyright © 2024 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Purpose: Scoliosis is a cause of loading imbalance between the lower limbs, which can result in BMD differences between the two femurs. We investigated the discrepancy in BMD values assessed by quantitative computed tomography (QCT) between femurs in patients with and without scoliosis, also assessing if this difference can be related to spine convexity., Methods: Abdominal CT examinations were retrospectively reviewed. An ''asynchronous'' calibration of CT images was performed to obtain BMD values from QCT. Scoliosis was evaluated on the antero-posterior CT localizer to calculate the Cobb angle. Differences between aBMD and vBMD of femurs were assessed in both scoliotic and non-scoliotic subjects., Results: Final study cohort consisted of 263 subjects, 225 of them without scoliosis (85.6%) and 38 with scoliosis (14.4%). No significant differences were found in the general population without scoliosis, except for vBMD at the neck. Comparison of femurs in scoliotic patients showed statistically significant differences at neck aBMD -0.028 g/cm 2 , p = 0.004), total femur aBMD (--0.032 g/cm 2 , p = 0.008) and total femur vBMD (--8.9 mg/cm 3 , p = 0.011), with lower BMD values on the convexity side. In 10 cases (26%) a change in the final T-score diagnosis was observed., Conclusion: QCT analysis demonstrated a difference in both areal and volumetric BMD between the two femurs of scoliotic patients, in relation to the side of the scoliotic curve. If these data will be confirmed by larger studies, bilateral femoral DXA acquisition may be proposed for these patients., (© 2023. The Author(s).)

Lower instrumented vertebra (LIV) selection is critical to avoid complications like adding-on. This study aims to determine the usefulness of the traction test (TR) in selecting the LIV during surgery for adolescent idiopathic scoliosis (AIS). We analyzed 42 AIS patients with Lenke 1 curves who had preoperative, postoperative, and at least 12-month follow-up X-rays, as well as preoperative side bending (SB) and TR radiograms. Neutral vertebra (NV), stable vertebra (SV), lower instrumented vertebra (LIV), and Cobb angles were identified and compared on all radiographic images. In 23 cases, the TR resulted in SV proximalization compared to the preoperative X-rays, while in 8 cases, SV-TR was more distal. This distalization occurred in 50% of Lenke 1C curves, where a greater correction of the distal curve was found. NV-TR was proximal to NV-preop in 9 cases, while NV-SB was proximal in 22 cases. LIV was proximal to SV-TR in 8 cases, while it was proximal to SV-preop in 22. One patient with LIV proximal to SV-TR developed adding-on. In conclusion, the TR is crucial in AIS preoperative planning as it provides information distinct from that of standard X-rays and SB: (1) it better assesses gravitational stability than rotational stability; and (2) choosing LIV equal to or proximal to SV-TR may prevent adding-on, except in 'flexible' Lenke 1C curves where LIV should be equal or distal to SV-preop.

Purpose: To determine diagnostic performance of MRI radiomics-based machine learning for classification of deep-seated lipoma and atypical lipomatous tumor (ALT) of the extremities., Material and Methods: This retrospective study was performed at three tertiary sarcoma centers and included 150 patients with surgically treated and histology-proven lesions. The training-validation cohort consisted of 114 patients from centers 1 and 2 (n = 64 lipoma, n = 50 ALT). The external test cohort consisted of 36 patients from center 3 (n = 24 lipoma, n = 12 ALT). 3D segmentation was manually performed on T1- and T2-weighted MRI. After extraction and selection of radiomic features, three machine learning classifiers were trained and validated using nested fivefold cross-validation. The best-performing classifier according to previous analysis was evaluated and compared to an experienced musculoskeletal radiologist in the external test cohort., Results: Eight features passed feature selection and were incorporated into the machine learning models. After training and validation (74% ROC-AUC), the best-performing classifier (Random Forest) showed 92% sensitivity and 33% specificity in the external test cohort with no statistical difference compared to the radiologist (p = 0.474)., Conclusion: MRI radiomics-based machine learning may classify deep-seated lipoma and ALT of the extremities with high sensitivity and negative predictive value, thus potentially serving as a non-invasive screening tool to reduce unnecessary referral to tertiary tumor centers., (© 2023. The Author(s).)

Introduction: Mayer-Rokitansky-Küster-Hauser syndrome (MRKH) syndrome can be classified into two types: type I (isolated) without extragenital abnormalities; type II (associated) with the presence of extragenital dimorphisms. Skeletal abnormalities are the second most frequent extragenital manifestations., Research Question: Association between MRKH and congenital scoliosis has been described; on the contrary, hyperkyphosis is very rare and sparsely described in the medical literature. Here we report our experience in the management of thoracolumbar hyperkyphosis in 16-year-old patient diagnosed with MRKH syndrome with an acute neurological impairment due to T11-T12 disc herniation., Material and Methods: Clinical and radiological images of the case were retrieved from the medical notes, operative records and imaging system., Results: Posterior surgical correction was proposed to treat the severe spinal deformity; however, surgery was delayed because of SARS-CoV2 pandemic outbreak. During the pandemic, the patient had a major clinical and radiological deterioration with development of paraparesis. Complete clinical resolution of the paraparesis and restoration of balance was achieved with a two stage surgical approach, with a first anterior stage followed by a delayed posterior approach aimed at deformity correction., Discussion: Congenital kyphosis are rare deformities that can progress rapidly leading to severe neurological deficits and worsening of the deformity. When patient has neurological deficit the surgical strategy to address the neurological problem first and plan the more complex and demanding corrective surgery remains a valid strategy that must be consider., Conclusion: This is the first reported case of hyperkyphosis in Mayer-Rokitansky-Küster-Hauser syndrome (MRKH) syndrome surgically treated., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2023 The Authors. Published by Elsevier B.V. on behalf of EUROSPINE, the Spine Society of Europe, EANS, the European Association of Neurosurgical Societies.)

Background: Vertebral arthrodesis for degenerative pathology of the lumbar spine still remains burdened by clinical problems with significant negative results. The introduction of the sagittal balance assessment with the evaluation of the meaning of pelvic parameters and spinopelvic (PI-LL) mismatch offered new evaluation criteria for this widespread pathology, but there is a lack of consistent evidence on long-term outcome., Methods: The authors performed an extensive systematic review of literature, with the aim to identify all potentially relevant studies about the role and usefulness of the restoration or the assessment of Sagittal balance in lumbar degenerative disease. They present the study protocol RELApSE (NCT05448092 ID) and discuss the rationale through a comprehensive literature review., Results: From the 237 papers on this topic, a total of 176 articles were selected in this review. The analysis of these literature data shows sparse and variable evidence. There are no observations or guidelines about the value of lordosis restoration or PI-LL mismatch. Most of the works in the literature are retrospective, monocentric, based on small populations, and often address the topic evaluation partially., Conclusions: The RELApSE study is based on the possibility of comparing a heterogeneous population by pathology and different surgical technical options on some homogeneous clinical and anatomic-radiological measures aiming to understanding the value that global lumbar and segmental lordosis, distribution of lordosis, pelvic tilt, and PI-LL mismatch may have on clinical outcome in lumbar degenerative pathology and on the occurrence of adjacent segment disease., (© 2023 The Authors.)

Purpose: Hybrid constructs with sublaminar bands have recently regained popularity as an alternative to all-screw construct for correction of adolescent idiopathic scoliosis (AIS). The aim of this study is to evaluate the ability of hybrid constructs with sublaminar bands to achieve a tridimensional correction of the scoliotic deformity. Our hypothesis is that hybrid construct with sublaminar bands are able to achieve a substantial derotation of the apical vertebrae, while preserving the thoracic kyphosis., Methods: A prospective evaluation of 50 consecutive cases (41 F, 9 M, mean age 14.7 ± 2 years) of AIS correction with hybrid construct was performed. In all cases, sublaminar bands were used at the apex of the main curve on concave side. All patients underwent pre and postoperative X-rays with EOS System, with full 3D reconstruction. Spinopelvic parameters and axial rotation of the vertebrae were measured pre and postoperatively., Results: 2.7 ± 0.9 mean sublaminar bands were used per patient. Mean correction of deformity was 50 ± 9.5%. on the coronal plane. The mean axial rotation of the apical vertebra went from 18° ± 11.5° preoperatively to 9.4° ± 7.2° postoperatively (p < 0.001) with a mean derotation of 47.7%. Thoracic kyphosis went from 32.1° ± 18° preoperatively to 37.3° ± 13.1° postoperatively (p < 0.05). No intraoperative complications due to sublaminar bands were recorded., Conclusions: Hybrid construct with sublaminar band have been showed to be safe and effective in deformity correction and in maintaining or restoring thoracic kyphosis. This study showed that with sublaminar bands applied at the curve apex a substantial derotation of the apical vertebrae can be achieved., (© 2022. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.)

Study Design: Biomechanical finite-element study., Objective: To directly compare the biomechanical effects of two different techniques for sagittal plane correction of adult spine deformity based on the anterior longitudinal ligament (ALL) resection and use of hyperlordotic cages, namely, the anterior column realignment (ACR) in L3-4, and ALIF in L5-S1 in terms of primary stability and rod stresses using finite-element models., Methods: A finite-element model of the thoracolumbar spine was used to perform the analysis. Starting from this "intact" model, three further models were constructed through the insertion of spinal instrumentation, i.e., pedicle screws, rods and cages: 1) posterior instrumentation between T9 and S1 (referred to as "T9-S1"); 2) posterior instrumentation T9-S1 + Hyperlordotic (26°) ALIF cage in L5-S1 ("ALIF"); 3) posterior instrumentation T9-S1 + Hyperlordotic (30°) ACR cage in L3-4 ("ACR"). These models were studied by simulations applying, alternately, a pure moment of 7.5 Nm between the three planes of motion (flexion, extension, lateral bending, and bilateral axial rotation), uniformly distributed over the upper surface of the T9 thoracic vertebra. A total of 24 simulations were performed (6 per models)., Results: All models presented a significant reduced ROM when compared to the intact model; the ROM reduction was higher both at L3-4 in the ACR model and at L5-S1 in the ALIF model. At L3-4, the ACR model had, in all cases, the lowest maximum values of Von Mises stresses on the rods, especially in flexion-extension. At L4-5, the ALIF model had the lowest stresses during flexion-extension and axial rotation, while the ACR model had the lowest stresses during lateral bending. At L5-S1, the ALIF model had, in all cases, the lowest stresses on the rods., Conclusions: This finite-element study showed how both ACR at L3-4 and ALIF-ACR at L5-S1 are effective in restoring lumbar lordosis (LL), stabilizing the spine and reducing stress on posterior rods at the index level when compared to a simple fixation model. Interestingly, ALIF-ACR reduces rod stress even at L4-5 in flexion-extension and axial rotation, possibly due to a better distribution of LL, especially on the lower arch, while ACR reduces the stress at L4-5 in lateral bending, possibly thanks to the larger footprint of the cage that increases the area of contact with the lateral side of the endplates., (© 2022. The Author(s), under exclusive licence to Scoliosis Research Society.)

Background and Objectives : Commonly being the first step in trauma routine imaging, up to 67% fractures are missed on plain radiographs of the thoracolumbar (TL) spine. The aim of this study was to develop a deep learning model that detects traumatic fractures on sagittal radiographs of the TL spine. Identifying vertebral fractures in simple radiographic projections would have a significant clinical and financial impact, especially for low- and middle-income countries where computed tomography (CT) and magnetic resonance imaging (MRI) are not readily available and could help select patients that need second level imaging, thus improving the cost-effectiveness. Materials and Methods : Imaging studies (radiographs, CT, and/or MRI) of 151 patients were used. An expert group of three spinal surgeons reviewed all available images to confirm presence and type of fractures. In total, 630 single vertebra images were extracted from the sagittal radiographs of the 151 patients-302 exhibiting a vertebral body fracture, and 328 exhibiting no fracture. Following augmentation, these single vertebra images were used to train, validate, and comparatively test two deep learning convolutional neural network models, namely ResNet18 and VGG16. A heatmap analysis was then conducted to better understand the predictions of each model. Results : ResNet18 demonstrated a better performance, achieving higher sensitivity (91%), specificity (89%), and accuracy (88%) compared to VGG16 (90%, 83%, 86%). In 81% of the cases, the "warm zone" in the heatmaps correlated with the findings, suggestive of fracture within the vertebral body seen in the imaging studies. Vertebras T12 to L2 were the most frequently involved, accounting for 48% of the fractures. A4, A3, and A1 were the most frequent fracture types according to the AO Spine Classification. Conclusions : ResNet18 could accurately identify the traumatic vertebral fractures on the TL sagittal radiographs. In most cases, the model based its prediction on the same areas that human expert classifiers used to determine the presence of a fracture.

Background: The most common conservative treatment for Adolescent Idiopathic Scoliosis (AIS) is bracing. However, several papers questioned the effectiveness of bracing for curves between 40° and 50° Cobb: the effectiveness in preventing curve progression could be as low as 35%. Seriate casting is considered a standard approach in early onset scoliosis; however, in the setting of AIS, cast treatment is seldom utilized, with only few studies reporting on its effectiveness. Aim of the study: The main aim of the study is to determine whether a seriate casting with Risser casts associated with bracing is more effective in preventing curve progression than bracing alone in curves larger than 40°. Furthermore, the secondary endpoints were: (1) is there a difference in effectiveness of casting between Thoracic (T) and Thoracolumbar/Lumbar (TL/L) curves? (2) Does the ‘in cast’ correction predicts the treatment outcome? (3) What is the effect on thoracic kyphosis of casting? Methods: This is a retrospective monocentric case−control study; through an Institutional Database search we identified all the patients treated at our institution between 1 January 2017 and 31 December 2020, with a diagnosis of AIS, Risser grade between 0 and 4 at the beginning of the treatment, at least one curve above 40° Cobb and treatment with either seriate Risser casting and bracing (Study Group, SG) or bracing alone (Control Group, CG). Standing full spine X-rays in AP and LL are obtained before and after the cast treatment; only AP standing full spine X-rays ‘in-cast’ are obtained for each cast made. Patients were stratified according to the curve behavior at the end of treatment (Risser 5): progression was defined as ≥6° increase in the curve magnitude or fusion needed; stabilization is defined as a change in curve by ±5°; and improvement was defined as ≥6° reduction in the curve. Results: For the final analysis, 55 compliant patients (12 M, 43 F, mean age 13.5 ± 1.6) were included in the SG and 27 (4 M, 23 F, mean age 13.6 ± 1.6) in the CG. Eight (14.5%) patients in the SG failed the conservative treatment while 14 (51.3%) failed in the CG. Consequently, the Relative Risk for progression in the Efficacy Analysis was 1.8 (95% CI 1, 3−2.6, p = 0.001), and the Number Needed to Treat was 2,4. No significant difference was found between the T and TL/L curves concerning the ‘progressive’ endpoint (z-score 0.263, p = 0.79). The mean percentage of ‘in cast’ curve reduction was 40.1 ± 15.2%; no significant correlation was found between the percentage of correction and the outcome (Spearman Correlation Coefficient 0.18). Finally, no significant differences between baseline and end of FU TK were found (32° ± 16.2 vs. 29.6 ± 15.8, p = ns). Discussion: Seriate Risser casting for AIS with larger curves (>40° Cobb) is effective in reducing curve progression when compared with full time bracing alone in treatment compliant patients. The treatment is equally effective in controlling T and TL/L curves; furthermore, a slight but non-significant decrease in TK was observed in patients treated with casting. This type of treatment should be considered for AIS patients who present with large curves to potentially reduce the percentage of surgical cases. Short Abstract: The aim of the study is to determine whether seriate Risser casting associated with bracing is more effective in preventing curve progression than bracing alone in curves larger than 40°. This is a retrospective monocentric case−control study; we identified all the patients treated at our institution with a diagnosis of AIS, Risser grade 0−4 at the beginning of the treatment, at least one curve above 40° Cobb (35° if treated with bracing alone) and treatment with either seriate Risser casting and bracing (Study Group, SG) or bracing alone (Control Group, CG). Fifty-five patients (12 M, 43 F, mean age 13.5 ± 1.6) were included in the SG and 30 (5 M, 25 F, mean age 13.9 ± 1.7) in the CG. Eight (14,5%) patients in the SG failed the conservative treatment while fifteen (50%) failed in the CG. Consequently, the Relative Risk for progression in the Efficacy Analysis was 1.8 (95% CI 1.3−2.6, p = 0.001), and the Number Needed to Treat was 2,4. Seriate Risser casting for AIS with larger curves (>40°) is effective in reducing curve progression when compared with full time bracing alone. This type of treatment should be considered for AIS patients who present with large curves.

Introduction: A precise knowledge of the possible Adverse Events (AEs) related to spinal surgical procedures is crucial in clinical practice., Research Question: Purposes of this study are: to determine the prevalence and severity of perioperative AEs associated with pediatric and adult spine surgery in a high volume center; to estimate the impact of perioperative AEs on length of hospital stay (LOS)., Material and Methods: This is a prospective, observational, monocenter study, including 346 consecutive patients (294 adults and 52 pediatrics). The SAVES-V2 questionnaire was used to record AEs. The form was updated by the medical staff every time an adverse event was recorded during hospitalization., Results: 21,2% of pediatric patients and 20,7% of adults had at least 1 perioperative AEs. In adults, dural tear (3.1%) and neuropathic pain (4,8%), were the most frequent intraoperative and postoperative AE, respectively. In pediatric patients, neurologic deterioration was the most frequent postoperative AE. A diagnosis of deformity (p=0.01), an ASA grade equal or superior to 3 (p=0.023) and the procedure 'Posterior Spinal Fusion' (p=0.001) were associated with a higher frequency of AEs. AEs required prolonged LOS in 40 cases, 7 (70%) pediatric patients and 33 (65%) adults., Discussion and Conclusion: The overall prevalence of AEs is 20.8%, and, although the distribution is almost equal between adult and pediatric patients, their severity is related to age, being higher in pediatric patients. Deformities, deformity correction, revision surgery and AP surgery are the most impactful factors. AEs seriously affect hospitalization, with prolonged LOS (mean 6 days)., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2021 The Authors.)

Purpose: Sacropelvic fixation is frequently used in combination with thoracolumbar instrumentation for the correction of severe spinal deformities. The purpose of this study was to explore the effects of the triangular titanium implants on the iliac screw fixation. Our hypothesis was that the use of triangular titanium implants can increase the stability of the iliac screw fixation., Methods: Three T10-pelvis instrumented models were created: pedicle screws and rods in T10-S1, and bilateral iliac screws (IL); posterior fixation and bilateral iliac screws and triangular implants inserted bilaterally in a sacro-alar-iliac trajectory (IL-Tri-SAI); posterior fixation and bilateral iliac screws and two bilateral triangular titanium implants inserted in a lateral trajectory (IL-Tri-Lat). Outputs of these models, such as hardware stresses, were compared against a model with pedicle screws and rods in T10-S1 (PED)., Results: Sacropelvic fixation decreased the L5-S1 motion by 75-90%. The motion of the SIJ was reduced by 55-80% after iliac fixation; the addition of triangular titanium implants further reduced it. IL, IL-Tri-SAI and IL-Tri-Lat demonstrated lower S1 pedicle stresses with respect to PED. Triangular implants had a protective effect on the iliac screw stresses., Conclusion: Sacropelvic fixation decreased L5-S1 range of motion suggesting increased stability of the joint. The combination of triangular titanium implants and iliac screws reduced the residual flexibility of the sacroiliac joint, and resulted in a protective effect on the S1 pedicle screws and iliac screws themselves. Clinical studies may be performed to demonstrate applicability of these FEA results to patient outcomes., (© 2021. The Author(s).)

Few studies evaluated bacterial colonization of spinal implants from a "topographic" point of view. This lack of knowledge could hinder the development of more effective strategies in the prevention and treatment of postoperative spinal infections. The aim of this in vitro study was the analysis of the adhesion pattern of sessile cells on conventional spinal implants, to identify "hot spots" on implants where bacterial adhesion could be favored. Clinically relevant Staphylococcus aureus, Staphylococcus epidermidis, and Pseudomonas aeruginosa isolates were grown on commercially available end product spinal implants. To identify sessile cells attached to implant surfaces, confocal laser scan microscopy was used. Different areas from the spinal instrumentations (both Ti and CoCr) were selected for biofilm quantification. Bacterial biofilm was markedly increased in the cut of the rods, both Ti and CoCr, as the uneven surface deriving from the cut might foster cell adhesion. Though not statistically significant, a difference was observed between the rod and the area of the notch, possibly as a consequence of the smoothening effect deriving from the bending of the rod. Finally, the amount of biofilm produced on cobalt-chromium surfaces was always more significant than that formed on titanium surfaces. This study highlights how bacterial adhesion through biofilm formation is favored on the surfaces of higher irregularity and that staphylococci are able to increase sessile biomass on CoCr surfaces. These preliminary results show how surface modifications on the implants may play a key role in bacterial adhesion, opening an exciting field for future research., (© 2020 Orthopaedic Research Society. Published by Wiley Periodicals LLC.)