RESUMEN
BACKGROUND: Theoretically, artificial intelligence can provide an accurate automatic solution to measure right ventricular (RV) ejection fraction (RVEF) from cardiovascular magnetic resonance (CMR) images, despite the complex RV geometry. However, in our recent study, commercially available deep learning (DL) algorithms for RVEF quantification performed poorly in some patients. The current study was designed to test the hypothesis that quantification of RV function could be improved in these patients by using more diverse CMR datasets in addition to domain-specific quantitative performance evaluation metrics during the cross-validation phase of DL algorithm development. METHODS: We identified 100 patients from our prior study who had the largest differences between manually measured and automated RVEF values. Automated RVEF measurements were performed using the original version of the algorithm (DL1), an updated version (DL2) developed from a dataset that included a wider range of RV pathology and validated using multiple domain-specific quantitative performance evaluation metrics, and conventional methodology performed by a core laboratory (CORE). Each of the DL-RVEF approaches was compared against CORE-RVEF reference values using linear regression and Bland-Altman analyses. Additionally, RVEF values were classified into 3 categories: ≤ 35%, 35-50%, and ≥ 50%. Agreement between RVEF classifications made by the DL approaches and the CORE measurements was tested. RESULTS: CORE-RVEF and DL-RVEFs were obtained in all patients (feasibility of 100%). DL2-RVEF correlated with CORE-RVEF better than DL1-RVEF (r = 0.87 vs. r = 0.42), with narrower limits of agreement. As a result, DL2 algorithm also showed increasing accuracy from 0.53 to 0.80 for categorizing RV function. CONCLUSIONS: The use of a new DL algorithm cross-validated on a dataset with a wide range of RV pathology using multiple domain-specific metrics resulted in a considerable improvement in the accuracy of automated RVEF measurements. This improvement was demonstrated in patients whose images were the most challenging and resulted in the largest RVEF errors. These findings underscore the critical importance of this strategy in the development of DL approaches for automated CMR measurements.
Asunto(s)
Inteligencia Artificial , Disfunción Ventricular Derecha , Ventrículos Cardíacos/diagnóstico por imagen , Humanos , Imagen por Resonancia Magnética , Imagen por Resonancia Cinemagnética/métodos , Valor Predictivo de las Pruebas , Reproducibilidad de los Resultados , Volumen Sistólico , Disfunción Ventricular Derecha/diagnóstico por imagen , Función Ventricular DerechaRESUMEN
BACKGROUND AND OBJECTIVES: Neural tube defects (NTDs) are an important cause of global morbidity worldwide. Well-planned global neurosurgery and public health efforts can aid vulnerable communities, but there is a need to elucidate the global burden of NTDs and identify regions without available data to better target interventions. METHODS: A scoping review to quantify worldwide NTD prevalence using the PubMed/Medline and birth defects surveillance registries was conducted. Data published after January 1, 1990, encompassing prevalence values of at least the 2 most prevalent NTDs-spina bifida and encephalocele-were abstracted. Average NTD prevalence rates were aggregated by World Health Organization (WHO) region and World Bank classification, and differences were determined using the analysis of variance test. Differences in availability of nationally representative data by WHO region and World Bank classification were determined using χ2 tests. RESULTS: This review captured 140 studies from a total of 93 of 194 WHO member countries. The percentage of countries within a geographic region with available NTD prevalence data was highest in the Eastern Mediterranean (EMR) (85.7%) and lowest in Africa (AFR) (31.3%). The NTD prevalence range was 0.9-269.6 per 10 000 births. Statistically significant differences in reported NTD prevalence rates existed by WHO Region (P = .00027) and World Bank income level of study country (P = .00193). Forty countries (43%) had conducted national-level studies assessing NTD prevalence. There was a statistically significant difference in the availability of nationally representative prevalence data depending on the WHO region (P = .0081) and World Bank classification of study country (P = .0017). CONCLUSION: There is a gap in availability of NTD prevalence data worldwide, with many WHO member states lacking national-level NTD prevalence estimates. These findings highlight the need for greater NTD surveillance efforts to identify the countries with the greatest need for targeted global intervention.
RESUMEN
OBJECTIVE: Augmented reality (AR) is an emerging technology that may accelerate skill acquisition and improve accuracy of thoracolumbar pedicle screw placements. We aimed to quantify the relative assistance of AR compared with freehand (FH) pedicle screw accuracy across different surgical experience levels. METHODS: A spine fellowship-trained and board-certified attending neurosurgeon, postgraduate year 4 neurosurgery resident, and second-year medical student placed 32 FH and 32 AR-assisted thoracolumbar pedicle screws in 3 cadavers. A cableless, voice-activated AR system was paired with a headset. Accuracy was assessed using χ2 analysis and the Gertzbein-Robbins scale. Angular error, distance error, and time per pedicle screw were collected and compared. RESULTS: The attending neurosurgeon had 91.6% (11/12) clinically acceptable (Gertzbein-Robbins scale A or B) insertion in both FH and AR groups; the resident neurosurgeon had 100% (9/9) FH and AR in both cases; the medical student had 72.3% (8/11) FH accuracy and 81.8% (9/11) AR accuracy. The medical student displayed significantly lower ideal (Gertzbein-Robbins scale A) FH accuracy compared with the resident neurosurgeon (P = 0.017) and attending neurosurgeon (P = 0.005), but no difference when using AR. FH screw placement was faster by both the attending neurosurgeon (median 46 seconds vs. 94.5 seconds, P = 0.0047) and the neurosurgery resident neurosurgeon (median 144 seconds vs. 140 seconds, P = 0.05). Total clinically acceptable AR and FH accuracy was 90.6% (29/32) and 87.5% (28/32), respectively (P = 0.69). CONCLUSIONS: AR screw placement allowed an inexperienced medical student to double their accuracy in 1 training session. With subsequent iterations, this promising technology could serve as an important tool for surgical training.
Asunto(s)
Realidad Aumentada , Tornillos Pediculares , Procedimientos Quirúrgicos Robotizados , Fusión Vertebral , Cirugía Asistida por Computador , Humanos , Procedimientos Neuroquirúrgicos , Vértebras Lumbares/cirugíaRESUMEN
OBJECTIVE: Race plays a salient role in access to surgical care. However, few investigations have assessed the impact of race within surgical populations after care has been delivered. The objective of this study was to employ an exact matching protocol to a homogenous population of spine surgery patients in order to isolate the relationships between race and short-term postoperative outcomes. METHODS: In total, 4263 consecutive patients who underwent single-level, posterior-only lumbar fusion at a single multihospital academic medical center were retrospectively enrolled. Of these patients, 3406 patients self-identified as White and 857 patients self-identified as non-White. Outcomes were initially compared across all patients via logistic regression. Subsequently, White patients and non-White patients were exactly matched on the basis of key demographic and health characteristics (1520 matched patients). Outcome disparities were evaluated between the exact-matched cohorts. Primary outcomes were readmissions, emergency department (ED) visits, reoperations, mortality, intraoperative complications, and discharge disposition. RESULTS: Before matching, non-White patients were less likely to be discharged home and more likely to be readmitted, evaluated in the ED, and undergo reoperation. After matching, non-White patients experienced higher rates of nonhome discharge, readmissions, and ED visits. Non-White patients did not have more surgical complications either before or after matching. CONCLUSIONS: Between otherwise similar cohorts of spinal fusion cases, non-White patients experienced unfavorable discharge disposition and higher risk of multiple adverse postoperative outcomes. However, these findings were not accounted for by differences in surgical complications, suggesting that structural factors underlie the observed disparities.
Asunto(s)
Fusión Vertebral , Humanos , Fusión Vertebral/métodos , Masculino , Femenino , Persona de Mediana Edad , Estudios Retrospectivos , Resultado del Tratamiento , Disparidades en Atención de Salud/etnología , Readmisión del Paciente/estadística & datos numéricos , Anciano , Reoperación/estadística & datos numéricos , Vértebras Lumbares/cirugía , Adulto , Población Blanca , Complicaciones Posoperatorias/epidemiologíaRESUMEN
BACKGROUND AND OBJECTIVES: Race has implications for access to medical care. However, the impact of race, after access to care has been attained, remains poorly understood. The objective of this study was to isolate the relationship between race and short-term outcomes across patients undergoing a single, common neurosurgical procedure. METHODS: In this retrospective cohort study, 3988 consecutive patients undergoing single-level, posterior-only open lumbar fusion at a single, multihospital, academic medical center were enrolled over a 6-year period. Among them, 3406 patients self-identified as White, and 582 patients self-identified as Black. Outcome disparities between all White patients vs all Black patients were estimated using logistic regression. Subsequently, coarsened exact matching controlled for outcome-mitigating factors; White and Black patients were exact-matched 1:1 on key demographic and health characteristics (matched n = 1018). Primary outcomes included 30-day and 90-day hospital readmissions, emergency department (ED) visits, reoperations, mortality, discharge disposition, and intraoperative complication. RESULTS: Before matching, Black patients experienced increased rate of nonhome discharge, readmissions, ED visits, and reoperations (all P < .001). After exact matching, Black patients were less likely to be discharged to home (odds ratio [OR] 2.68, P < .001) and had higher risk of 30-day and 90-day readmissions (OR 2.24, P < .001; OR 1.91, P < .001; respectively) and ED visits (OR 1.79, P = .017; OR 2.09, P < .001). Black patients did not experience greater risk of intraoperative complication (unintentional durotomy). CONCLUSION: Between otherwise homogenous spinal fusion cohorts, Black patients experienced unfavorable short-term outcomes. These disparities were not explained by differences in intraoperative complications. Further investigation must characterize and mitigate institutional and societal factors that contribute to outcome disparities.
RESUMEN
BACKGROUND AND OBJECTIVES: Smartphone activity data recorded through high-fidelity accelerometry can provide accurate postoperative assessments of patient mobility. The "big data" available through smartphones allows for advanced analyses, yielding insight into patient well-being. This study compared rate of change in functional activity data between lumbar fusion (LF) and lumbar decompression (LD) patients to determine preoperative and postoperative course differences. METHODS: Twenty-three LF and 18 LD patients were retrospectively included. Activity data (steps per day) recorded in Apple Health, encompassing over 70 000 perioperative data points, was classified into 6 temporal epochs representing distinct functional states, including acute preoperative decline, immediate postoperative recovery, and postoperative decline. The daily rate of change of each patient's step counts was calculated for each perioperative epoch. RESULTS: Patients undergoing LF demonstrated steeper preoperative declines than LD patients based on the first derivative of step count data (P = .045). In the surgical recovery phase, LF patients had slower recoveries (P = .041), and LF patients experienced steeper postoperative secondary declines than LD patients did (P = .010). The rate of change of steps per day demonstrated varying perioperative trajectories that were not explained by differences in age, comorbidities, or levels operated. CONCLUSION: Patients undergoing LF and LD have distinct perioperative activity profiles characterized by the rate of change in the patient daily steps. Daily steps and their rate of change is thus a valuable metric in phenotyping patients and understanding their postsurgical outcomes. Prospective studies are needed to expand upon these data and establish causal links between preoperative patient mobility, patient characteristics, and postoperative functional outcomes.
RESUMEN
Vestibular schwannomas (VS) are the most common tumor of the skull base with available treatment options that carry a risk of iatrogenic injury to the facial nerve, which can significantly impact patients' quality of life. As facial nerve outcomes remain challenging to prognosticate, we endeavored to utilize machine learning to decipher predictive factors relevant to facial nerve outcomes following microsurgical resection of VS. A database of patient-, tumor- and surgery-specific features was constructed via retrospective chart review of 242 consecutive patients who underwent microsurgical resection of VS over a 7-year study period. This database was then used to train non-linear supervised machine learning classifiers to predict facial nerve preservation, defined as House-Brackmann (HB) I vs. facial nerve injury, defined as HB II-VI, as determined at 6-month outpatient follow-up. A random forest algorithm demonstrated 90.5% accuracy, 90% sensitivity and 90% specificity in facial nerve injury prognostication. A random variable (rv) was generated by randomly sampling a Gaussian distribution and used as a benchmark to compare the predictiveness of other features. This analysis revealed age, body mass index (BMI), case length and the tumor dimension representing tumor growth towards the brainstem as prognosticators of facial nerve injury. When validated via prospective assessment of facial nerve injury risk, this model demonstrated 84% accuracy. Here, we describe the development of a machine learning algorithm to predict the likelihood of facial nerve injury following microsurgical resection of VS. In addition to serving as a clinically applicable tool, this highlights the potential of machine learning to reveal non-linear relationships between variables which may have clinical value in prognostication of outcomes for high-risk surgical procedures.
Asunto(s)
Traumatismos del Nervio Facial , Aprendizaje Automático , Microcirugia , Neuroma Acústico , Humanos , Neuroma Acústico/cirugía , Masculino , Femenino , Persona de Mediana Edad , Microcirugia/efectos adversos , Microcirugia/métodos , Pronóstico , Traumatismos del Nervio Facial/etiología , Estudios Retrospectivos , Adulto , Anciano , AlgoritmosRESUMEN
OBJECTIVE: Atlantooccipital dislocation (AOD) is a highly unstable and often neurologically devastating injury to the craniocervical junction that typically results from high-energy trauma. Management of these devastating injuries is complex, with prognostication difficult due to high rates of concomitant intracranial and systemic injuries. This report highlights advances in management of AOD and appropriate implementation of operative adjuncts including neuronavigation and the use of intraoperative neuromonitoring. METHODS: All patients with AOD presenting to a high-volume, level 1 trauma center between January 2015 and August 2021 were retrospectively identified through a prospectively maintained database of patients presenting with traumatic spine injuries. Medical records, including imaging reports, clinical documentation, and intraoperative neurophysiological reports were reviewed. RESULTS: A total of 11 patients were identified with patterns of injury consistent with AOD. Fifty-five percent of patients survived until discharge. 73% of patients underwent surgery for stabilization. All 4 patients with preoperative neurologic deficits who underwent surgery had monitorable transcranial motor evoked potentials and somatosensory evoked potentials. Two experienced significant motor recovery postoperatively, and 2 did not survive to discharge. Blunt cerebrovascular injuries were identified in 73% of patients. CONCLUSION: AOD is encountered with increasing frequency. The identification and management of this specific injury is complicated by the volume and severity of associated injuries, especially concomitant traumatic brain injury. Timely recognition is critical and the use of surgical adjuncts including intraoperative neurophysiologic monitoring and surgical navigation can increase the safety and success of these procedures while also providing prognostic information on potential for motor recovery.
Asunto(s)
Monitorización Neurofisiológica Intraoperatoria , Luxaciones Articulares , Humanos , Estudios Retrospectivos , Centros Traumatológicos , Potenciales Evocados Motores/fisiología , Potenciales Evocados Somatosensoriales/fisiología , Monitorización Neurofisiológica Intraoperatoria/métodosRESUMEN
Anterior lumbar interbody fusion (ALIF), traditionally performed supine, allows for significant restoration of lumbar lordosis, disc height, and foraminal height in degenerative spine diseases; however, an iatrogenic injury to the viscera and the great vessels can have devastating consequences. Although lateral lumbar interbody fusion (LLIF) is an acceptable and minimally invasive alternative at the L5-S1 level, this approach is suboptimal because of a narrow surgical corridor limited by the iliac crest, common iliac artery and vein, and psoas. Furthermore, combining supine L5-S1 ALIF and lateral decubitus (LD) LLIF requires time-consuming patient repositioning.1,2 To maximize the advantages of both procedures in patients with disease spanning the lumbosacral junction, ALIF and LLIF can be performed in a single stage with the patient remaining in an LD position throughout. To improve the efficiency of this single-position procedure, a fluoroscopy-based instrument tracking system (TrackX Technology Inc., Hillsborough, North Carolina, USA) was used to navigate surgical tools during the procedure. We show this technique in a 43-year-old patient with medically intractable back and leg pain secondary to multi-level degenerative lumbar spondylosis. The patient consented to this procedure; all participants consented to publication of their images. This tracking system allowed for accurate and precise virtual projections of surgical instruments, thereby facilitating the identification of midline and proper trajectories to perform discectomy and implant placement, reducing the amount of intraoperative fluoroscopy use, and eliminating intraoperative computed tomography. To our knowledge, this is the first operative video showing a fluoroscopy-based instrument tracking system used in a combined single-position LD-ALIF and LD-LLIF.
Asunto(s)
Lordosis , Fusión Vertebral , Humanos , Adulto , Vértebras Lumbares/diagnóstico por imagen , Vértebras Lumbares/cirugía , Fluoroscopía , Región Lumbosacra/cirugía , Lordosis/cirugía , Discectomía , Fusión Vertebral/métodosRESUMEN
BACKGROUND: Lumbar lateral interbody fusion (LLIF) is traditionally performed in 2 stages: placing the interbody cage in the lateral decubitus position, then placing the percutaneous pedicle screw in the prone position. Performing interbody fusion and posterior fixation simultaneously could improve operative efficiency and clinical outcomes associated with longer operative times. We describe the operative steps and report clinical and radiographic outcomes associated with a simultaneous anterior and posterior approach (SAPA) for LLIF. METHODS: Patients who underwent SAPA LLIF performed by a single surgeon over 1 year were retrospectively reviewed. Demographic, clinical, and radiographic data were analyzed, an operative guideline was created, and a learning curve was constructed using operative times. RESULTS: SAPA LLIF was performed in 11 patients. Three patients experienced transient postoperative femoral nerve plexopathy with symptoms of ipsilateral hip flexion weakness and/or anterior thigh numbness; there were no other complications in the cohort. Radiographically, patients achieved significant increases in disc height (8.3 mm vs. 13.5 mm, P = 0.002) and foraminal height (20.2 mm vs. 25.3 mm, P = 0.0001). Patients showed significant improvements in Oswestry Disability Index (52 vs. 27.8, P = 0.002) and Patient-Reported Outcome Measurement Information System Physical Function (32.6 vs. 39, P = 0.048) and Pain Interference (64.9 vs. 59.6, P = 0.001) at 3 months. A downward trend in operative time was observed for 1-level SAPA LLIF. CONCLUSIONS: SAPA LLIF is a safe approach for LLIF that results in favorable clinical outcomes. This technique can potentially improve operative efficiency further along the course of a surgeon's learning curve.
Asunto(s)
Tornillos Pediculares , Procedimientos Quirúrgicos Robotizados , Fusión Vertebral , Humanos , Estudios Retrospectivos , Región Lumbosacra/cirugía , Fusión Vertebral/métodos , Vértebras Lumbares/diagnóstico por imagen , Vértebras Lumbares/cirugíaRESUMEN
BACKGROUND: Augmented reality (AR) and virtual reality (VR) implementation in spinal surgery has expanded rapidly over the past decade. This systematic review summarizes the use of AR/VR technology in surgical education, preoperative planning, and intraoperative guidance. METHODS: A search query for AR/VR technology in spine surgery was conducted through PubMed, Embase, and Scopus. After exclusions, 48 studies were included. Included studies were then grouped into relevant subsections. Categorization into subsections yielded 12 surgical training studies, 5 preoperative planning, 24 intraoperative usage, and 10 radiation exposure. RESULTS: VR-assisted training significantly reduced penetration rates or increased accuracy rates compared to lecture-based groups in 5 studies. Preoperative VR planning significantly influenced surgical recommendations and reduced radiation exposure, operating time, and estimated blood loss. For 3 patient studies, AR-assisted pedicle screw placement accuracy ranged from 95.77% to 100% using the Gertzbein grading scale. Head-mounted display was the most common interface used intraoperatively followed by AR microscope and projector. AR/VR also had applications in tumor resection, vertebroplasty, bone biopsy, and rod bending. Four studies reported significantly reduced radiation exposure in AR group compared to fluoroscopy group. CONCLUSIONS: AR/VR technologies have the potential to usher in a paradigm shift in spine surgery. However, the current evidence indicates there is still a need for 1) defined quality and technical requirements for AR/VR devices, 2) more intraoperative studies that explore usage outside of pedicle screw placement, and 3) technological advancements to overcome registration errors via the development of an automatic registration method.
Asunto(s)
Realidad Aumentada , Tornillos Pediculares , Cirugía Asistida por Computador , Realidad Virtual , Humanos , Cirugía Asistida por Computador/métodos , Procedimientos NeuroquirúrgicosRESUMEN
OBJECTIVE: Lumbar interbody fusion (LIF) techniques have seen impressive innovation in recent years, leading to an expansion of the LIF lexicon. This study systematically analyzes LIF nomenclature in contemporary literature and proposes a standardized classification system for reporting LIF terminology. METHODS: A search query was conducted through the PubMed database using "lumbar fusion OR lumbar interbody fusion." A total of 1455 articles were identified, and 605 references to LIF were recorded. Following a systematic review of the terminology, we developed a LIF reporting guidelines that capture the existing LIF nomenclature while avoiding redundant or ambiguous terminology. RESULTS: The most referenced anatomical approaches were transforaminal (43.0%), followed by posterior (25.0%), lateral (19.7%), and anterior (10.9%). Overall, there were 72 unique ways to describe LIF. Unique prefixes were recorded by approach (posterior: 26; lateral: 13; anterior: 3). Forty unique prefixes/suffixes overlapped in their usage. "MI" (14.4%), "MIS" (38.1%), and "MISS" (0.6%) all referenced a minimally invasive approach. "O" (12.5%), "CO" (1.3%), and "TO" (1.3%) all described open techniques. "Endo" (0.6%), "Endoscopic-assisted" (1.3%), and "PE" (1.9%) all referenced endoscopic-assisted procedures. CONCLUSIONS: The current LIF nomenclature contains many unique LIF terms that were found to be inconsistently defined, redundant, or ambiguous. We propose the standardization of a 4-part naming system which highlights the crucial parts of LIF: (1) intraoperative repositioning, (2) patient position, (3) anatomical approach, and (4) orientation of the surgical corridor to the psoas muscles.
Asunto(s)
Procedimientos Quirúrgicos Mínimamente Invasivos , Fusión Vertebral , Humanos , Procedimientos Quirúrgicos Mínimamente Invasivos/métodos , Región Lumbosacra/cirugía , Fusión Vertebral/métodos , Vértebras Lumbares/cirugíaRESUMEN
STUDY DESIGN: Prospective cohort study. SUMMARY OF BACKGROUND DATA: C-arm fluoroscopy and O-arm navigation are vital tools in modern spine surgeries, but their repeated usage can endanger spine surgeons. Although a surgeon's chest and abdomen are protected by lead aprons, the eyes and extremities generally receive less protection. OBJECTIVE: In this study, we compare differences in intraoperative radiation exposure across the protected and unprotected regions of a surgeon's body. METHODS: Sixty-five consecutive spine surgeries were performed by a single spine-focused neurosurgeon over 9 months. Radiation exposure to the primary surgeon was measured through dosimeters worn over the lead apron, under the lead apron, on surgical loupes, and as a ring on the dominant hand. Differences were assessed with rigorous statistical testing and radiation exposure per surgical case was extrapolated. RESULTS: During the study, the measured radiation exposure over the apron, 176 mrem, was significantly greater than that under the apron, 8 mrem (P = 0.0020), demonstrating a shielding protective effect. The surgeon's dominant hand was exposed to 329 mrem whereas the eyes were exposed to 152.5 mrem of radiation. Compared with the surgeon's protected abdominal area, the hands (P = 0.0002) and eyes (P = 0.0002) received significantly greater exposure. Calculated exposure per case was 2.8 mrem for the eyes and 5.1 mrem for the hands. It was determined that a spine-focused neurosurgeon operating 400 cases annually will incur a radiation exposure of 60,750 mrem to the hands and 33,900 mrem to the eyes over a 30-year career. CONCLUSIONS: Our study found that spine surgeons encounter significantly more radiation exposure to the eyes and the extremities compared with protected body regions. Lifetime exposure exceeds the annual limits set by the International Commission on Radiologic Protection for the extremities (50,000 mrem/y) and the eyes (15,000 mrem/y), calling for increased awareness about the dangerous levels of radiation exposure that a spine surgeon incurs over one's career.
Asunto(s)
Exposición a la Radiación , Cirujanos , Cirugía Asistida por Computador , Humanos , Estudios Prospectivos , Cuerpo Humano , Imagenología Tridimensional , Tomografía Computarizada por Rayos X , Fluoroscopía/efectos adversos , Fluoroscopía/métodosRESUMEN
OBJECTIVE: Patient-reported outcome measures (PROMs) are the gold standard for assessing postoperative outcomes in spine surgery. However, PROMs are also limited by the inherent subjectivity of self-reported qualitative data. Recent literature has highlighted the utility of patient mobility data streamed from smartphone accelerometers as an objective measure of functional outcomes and complement to traditional PROMs. Still, for activity-based data to supplement existing PROMs, they must be validated against current metrics. In this study, the authors assessed the relationships and concordance between longitudinal smartphone-based mobility data and PROMs. METHODS: Patients receiving laminectomy (n = 21) or fusion (n = 10) between 2017 and 2022 were retrospectively included. Activity data (steps-per-day count) recorded in the Apple Health mobile application over a 2-year perioperative window were extracted and subsequently normalized to allow for intersubject comparison. PROMS, including the visual analog scale (VAS), Patient Reported Outcome Measurement Information System Pain Interference (PROMIS-PI), Oswestry Disability Index (ODI), and EQ-5D, collected at the preoperative and 6-week postoperative visits were retrospectively extracted from the electronic medical record. Correlations between PROMs and patient mobility were assessed and compared between patients who did and those who did not achieve the established minimal clinically important difference (MCID) for each measure. RESULTS: A total of 31 patients receiving laminectomy (n = 21) or fusion (n = 10) were included. Change between preoperative and 6-week postoperative VAS and PROMIS-PI scores demonstrated moderate (r = -0.46) and strong (r = -0.74) inverse correlations, respectively, with changes in normalized steps-per-day count. In cohorts of patients who achieved PROMIS-PI MCID postoperatively, indicating subjective improvement in pain, there was a 0.784 standard deviation increase in normalized steps per day, representing a 56.5% improvement (p = 0.027). Patients who did achieve the MCID of improvement in either PROMIS-PI or VAS after surgery were more likely to experience an earlier sustained improvement in physical activity commensurate to or greater than their preoperative baseline (p = 2.98 × 10-18) than non-MCID patients. CONCLUSIONS: This study demonstrates a strong correlation between changes in mobility data extracted from patient smartphones and changes in PROMs following spine surgery. Further elucidating this relationship will allow for more robust supplementation of existing spine outcome measure tools with analyzed objective activity data.
Asunto(s)
Diferencia Mínima Clínicamente Importante , Teléfono Inteligente , Humanos , Estudios Retrospectivos , Limitación de la Movilidad , Medición de Resultados Informados por el Paciente , Encuestas y Cuestionarios , Dolor , Resultado del TratamientoRESUMEN
BACKGROUND: The development of accurate machine learning algorithms requires sufficient quantities of diverse data. This poses a challenge in health care because of the sensitive and siloed nature of biomedical information. Decentralized algorithms through federated learning (FL) avoid data aggregation by instead distributing algorithms to the data before centrally updating one global model. OBJECTIVE: To establish a multicenter collaboration and assess the feasibility of using FL to train machine learning models for intracranial hemorrhage (ICH) detection without sharing data between sites. METHODS: Five neurosurgery departments across the United States collaborated to establish a federated network and train a convolutional neural network to detect ICH on computed tomography scans. The global FL model was benchmarked against a standard, centrally trained model using a held-out data set and was compared against locally trained models using site data. RESULTS: A federated network of practicing neurosurgeon scientists was successfully initiated to train a model for predicting ICH. The FL model achieved an area under the ROC curve of 0.9487 (95% CI 0.9471-0.9503) when predicting all subtypes of ICH compared with a benchmark (non-FL) area under the ROC curve of 0.9753 (95% CI 0.9742-0.9764), although performance varied by subtype. The FL model consistently achieved top three performance when validated on any site's data, suggesting improved generalizability. A qualitative survey described the experience of participants in the federated network. CONCLUSION: This study demonstrates the feasibility of implementing a federated network for multi-institutional collaboration among clinicians and using FL to conduct machine learning research, thereby opening a new paradigm for neurosurgical collaboration.
Asunto(s)
Algoritmos , Benchmarking , Humanos , Hemorragias Intracraneales , Aprendizaje Automático , Redes Neurales de la ComputaciónRESUMEN
INTRODUCTION: Technological advancements are reshaping medical education, with digital tools becoming essential in all levels of training. Amidst this transformation, the study explores the potential of ChatGPT, an artificial intelligence model by OpenAI, in enhancing neurosurgical board education. The focus extends beyond technology adoption to its effective utilization, with ChatGPT's proficiency evaluated against practice questions from the Primary Neurosurgery Written Board Exam. METHODS: Using the Congress of Neurologic Surgeons (CNS) Self-Assessment Neurosurgery (SANS) Exam Board Review Prep questions, we conducted 3 rounds of analysis with ChatGPT. We developed a novel ChatGPT Neurosurgical Evaluation Matrix (CNEM) to assess the output quality, accuracy, concordance, and clarity of ChatGPT's answers. RESULTS: ChatGPT achieved spot-on accuracy for 66.7% of prompted questions, 59.4% of unprompted questions, and 63.9% of unprompted questions with a leading phrase. Stratified by topic, accuracy ranged from 50.0% (Vascular) to 78.8% (Neuropathology). In comparison to SANS explanations, ChatGPT output was considered better in 19.1% of questions, equal in 51.6%, and worse in 29.3%. Concordance analysis showed that 95.5% of unprompted ChatGPT outputs and 97.4% of unprompted outputs with a leading phrase were aligned. CONCLUSIONS: Our study evaluated the performance of ChatGPT in neurosurgical board education by assessing its accuracy, clarity, and concordance. The findings highlight the potential and challenges of integrating AI technologies like ChatGPT into medical and neurosurgical board education. Further research is needed to refine these tools and optimize their performance for enhanced medical education and patient care.
Asunto(s)
Neurocirugia , Humanos , Inteligencia Artificial , Escolaridad , Procedimientos Neuroquirúrgicos , LenguajeRESUMEN
BACKGROUND: Artificial intelligence is increasingly utilized to aid in the interpretation of cardiac magnetic resonance (CMR) studies. One of the first steps is the identification of the imaging plane depicted, which can be achieved by both deep learning (DL) and classical machine learning (ML) techniques without user input. We aimed to compare the accuracy of ML and DL for CMR view classification and to identify potential pitfalls during training and testing of the algorithms. METHODS: To train our DL and ML algorithms, we first established datasets by retrospectively selecting 200 CMR cases. The models were trained using two different cohorts (passively and actively curated) and applied data augmentation to enhance training. Once trained, the models were validated on an external dataset, consisting of 20 cases acquired at another center. We then compared accuracy metrics and applied class activation mapping (CAM) to visualize DL model performance. RESULTS: The DL and ML models trained with the passively-curated CMR cohort were 99.1% and 99.3% accurate on the validation set, respectively. However, when tested on the CMR cases with complex anatomy, both models performed poorly. After training and testing our models again on all 200 cases (active cohort), validation on the external dataset resulted in 95% and 90% accuracy, respectively. The CAM analysis depicted heat maps that demonstrated the importance of carefully curating the datasets to be used for training. CONCLUSIONS: Both DL and ML models can accurately classify CMR images, but DL outperformed ML when classifying images with complex heart anatomy.
Asunto(s)
Aprendizaje Profundo , Inteligencia Artificial , Humanos , Aprendizaje Automático , Imagen por Resonancia Magnética , Estudios RetrospectivosRESUMEN
OBJECTIVES: The aim of this study was to determine whether left ventricular ejection fraction (LVEF) and right ventricular ejection fraction (RVEF) and left ventricular mass (LVM) measurements made using 3 fully automated deep learning (DL) algorithms are accurate and interchangeable and can be used to classify ventricular function and risk-stratify patients as accurately as an expert. BACKGROUND: Artificial intelligence is increasingly used to assess cardiac function and LVM from cardiac magnetic resonance images. METHODS: Two hundred patients were identified from a registry of individuals who underwent vasodilator stress cardiac magnetic resonance. LVEF, LVM, and RVEF were determined using 3 fully automated commercial DL algorithms and by a clinical expert (CLIN) using conventional methodology. Additionally, LVEF values were classified according to clinically important ranges: <35%, 35% to 50%, and ≥50%. Both ejection fraction values and classifications made by the DL ejection fraction approaches were compared against CLIN ejection fraction reference. Receiver-operating characteristic curve analysis was performed to evaluate the ability of CLIN and each of the DL classifications to predict major adverse cardiovascular events. RESULTS: Excellent correlations were seen for each DL-LVEF compared with CLIN-LVEF (r = 0.83-0.93). Good correlations were present between DL-LVM and CLIN-LVM (r = 0.75-0.85). Modest correlations were observed between DL-RVEF and CLIN-RVEF (r = 0.59-0.68). A >10% error between CLIN and DL ejection fraction was present in 5% to 18% of cases for the left ventricle and 23% to 43% for the right ventricle. LVEF classification agreed with CLIN-LVEF classification in 86%, 80%, and 85% cases for the 3 DL-LVEF approaches. There were no differences among the 4 approaches in associations with major adverse cardiovascular events for LVEF, LVM, and RVEF. CONCLUSIONS: This study revealed good agreement between automated and expert-derived LVEF and similarly strong associations with outcomes, compared with an expert. However, the ability of these automated measurements to accurately classify left ventricular function for treatment decision remains limited. DL-LVM showed good agreement with CLIN-LVM. DL-RVEF approaches need further refinements.