Deep Learning for Automatically Visual Evoked Potential Classification During Surgical Decompression of Sellar Region Tumors

Reliability of intraoperative visual evoked potentials (iVEPs) in monitoring visual function during endoscopic transsphenoidal surgery

OBJECTIVE

To refine a reliable and reproducible intraoperative visual evoked potentials (iVEPs) monitoring protocol during endoscopic transsphenoidal surgery. To assess the reliability of baseline iVEPs in predicting preoperative visual status and perioperative iVEP variation in predicting postoperative visual outcome.

METHODS

Sixty-four patients harboring tumors of the pituitary region were included. All patients underwent endoscopic endonasal approach (EEA) with iVEPs monitoring, using a totally intravenous anesthetic protocol. Ophthalmological evaluation included visual acuity and visual field studies.

RESULTS

Preoperatively, visual acuity was reduced in 86% and visual field in 76.5% of cases. Baseline iVEPs amplitude was significantly correlated with preoperative visual acuity and visual field (p = 0.001 and p = 0.0004, respectively), confirming the reliability of the neurophysiological/anesthetic protocol implemented. Importantly, perioperatively the variation in iVEPs amplitude was significantly correlated with the changes in visual acuity (p < 0.0001) and visual field (p = 0.0013). ROC analysis confirmed that iVEPs are an accurate predictor of perioperiative visual acuity improvement, with a 100% positive predictive value in patients with preoperative vision loss.

CONCLUSIONS

iVEPs during EEA is highly reliable in describing preoperative visual function and can accurately predict postoperative vision improvement.

SIGNIFICANCE

iVEPs represent a promising resource for carrying out a more effective and safe endoscopic transsphenoidal surgery.

Evolution of flash visual evoked potentials to monitor visual pathway integrity during tumor resection: illustrative cases and literature review

Flash visual evoked potentials (fVEPs) provide a means to interrogate visual system functioning intraoperatively during tumor resection in which the optic pathway is at risk for injury. Due to technical limitations, fVEPs have remained underutilized in the armamentarium of intraoperative neurophysiological monitoring (IONM) techniques. Here we review the evolution of fVEPs as an IONM technique with emphasis on the enabling technological and intraoperative improvements. A combined approach with electroretinography (ERG) has enhanced feasibility of fVEP neuromonitoring as a practical application to increase safety and reduce error during tumor resection near the prechiasmal optic pathway. The major advance has been towards differentiating true cases of damage from false findings. We use two illustrative neurosurgical cases in which fVEPs were monitored with and without ERG to discuss limitations and demonstrate how ERG data can clarify false-positive findings in the operating room. Standardization measures have focused on uniformity of photostimulation parameters for fVEP recordings between neurosurgical groups.

The classification of flash visual evoked potential based on deep learning

BACKGROUND

Visual electrophysiology is an objective visual function examination widely used in clinical work and medical identification that can objectively evaluate visual function and locate lesions according to waveform changes. However, in visual electrophysiological examinations, the flash visual evoked potential (FVEP) varies greatly among individuals, resulting in different waveforms in different normal subjects. Moreover, most of the FVEP wave labelling is performed automatically by a machine, and manually corrected by professional clinical technicians. These labels may have biases due to the individual variations in subjects, incomplete clinical examination data, different professional skills, personal habits and other factors. Through the retrospective study of big data, an artificial intelligence algorithm is used to maintain high generalization abilities in complex situations and improve the accuracy of prescreening.

METHODS

A novel multi-input neural network based on convolution and confidence branching (MCAC-Net) for retinitis pigmentosa RP recognition and out-of-distribution detection is proposed. The MCAC-Net with global and local feature extraction is designed for the FVEP signal that has different local and global information, and a confidence branch is added for out-of-distribution sample detection. For the proposed manual features,a new input layer is added.

RESULTS

The model is verified by a clinically collected FVEP dataset, and an accuracy of 90.7% is achieved in the classification task and 93.3% in the out-of-distribution detection task.

CONCLUSION

We built a deep learning-based FVEP classification algorithm that promises to be an excellent tool for screening RP diseases by using FVEP signals.

Clinical features and genetic spectrum of NMNAT1-associated retinal degeneration

OBJECTIVES

To systematically analyse the NMNAT1 variant spectrum and frequency, the associated phenotypic characteristics, and potential genotype-phenotype correlations based on our data and literature review.

METHODS

Biallelic potential pathogenic variants (PPV) in NMNAT1 were collected from our in-house exome sequencing data. Whole-genome sequencing was conducted subsequently for patients with only one heterozygous PPV detected in NMNAT1. The clinical data were reviewed and evaluated in detail. Furthermore, the literature was reviewed for reports of NMNAT1 variants and their associated phenotypes.

RESULTS

Eleven NMNAT1 variants, including two novel variants, were detected in 8 families from our cohort. All of the 9 available patients showed generalized tapetoretinal dystrophy at an early age (88.9% in the first decade), and disciform macular atrophy was identified in six patients from five unrelated families. Among a total of 125 patients from 8 families of our cohort and 91 families reported by the available literature, 92.9% patients showed onset of disease in the first year after birth, and 89.0% patients showed visual acuity of 0.05 or lower. All of the 39 patients with fundus photos available presented disciform macular atrophy with generalized tapetoretinal dystrophy. Most (54/80, 67.5%) of causative NMNAT1 variants were missense. The most frequent variants in Caucasian and Asian population are p.E257K and p.R237C, respectively.

CONCLUSIONS

Early-onset age, disciform macular atrophy with generalized tapetoretinal dystrophy, and poor visual acuity are the typical features of NMNAT1-associated retinal degeneration. Different variant hot spots of NMNAT1 were observed in different populations.

The future of artificial intelligence in neurosurgery: A narrative review

Background:

Artificial intelligence (AI) and machine learning (ML) algorithms are on the tremendous rise for being incorporated into the field of neurosurgery. AI and ML algorithms are different from other technological advances as giving the capability for the computer to learn, reason, and problem-solving skills that a human inherits. This review summarizes the current use of AI in neurosurgery, the challenges that need to be addressed, and what the future holds.

Methods:

A literature review was carried out with a focus on the use of AI in the field of neurosurgery and its future implication in neurosurgical research.

Results:

The online literature on the use of AI in the field of neurosurgery shows the diversity of topics in terms of its current and future implications. The main areas that are being studied are diagnostic, outcomes, and treatment models.

Conclusion:

Wonders of AI in the field of medicine and neurosurgery hold true, yet there are a lot of challenges that need to be addressed before its implications can be seen in the field of neurosurgery from patient privacy, to access to high-quality data and overreliance on surgeons on AI. The future of AI in neurosurgery is pointed toward a patient-centric approach, managing clinical tasks, and helping in diagnosing and preoperative assessment of the patients.

Explainable deep learning algorithm for distinguishing incomplete Kawasaki disease by coronary artery lesions on echocardiographic imaging

BACKGROUND AND OBJECTIVE

Incomplete Kawasaki disease (KD) has often been misdiagnosed due to a lack of the clinical manifestations of classic KD. However, it is associated with a markedly higher prevalence of coronary artery lesions. Identifying coronary artery lesions by echocardiography is important for the timely diagnosis of and favorable outcomes in KD. Moreover, similar to KD, coronavirus disease 2019, currently causing a worldwide pandemic, also manifests with fever; therefore, it is crucial at this moment that KD should be distinguished clearly among the febrile diseases in children. In this study, we aimed to validate a deep learning algorithm for classification of KD and other acute febrile diseases.

METHODS

We obtained coronary artery images by echocardiography of children (n = 138 for KD; n = 65 for pneumonia). We trained six deep learning networks (VGG19, Xception, ResNet50, ResNext50, SE-ResNet50, and SE-ResNext50) using the collected data.

RESULTS

SE-ResNext50 showed the best performance in terms of accuracy, specificity, and precision in the classification. SE-ResNext50 offered a precision of 81.12%, a sensitivity of 84.06%, and a specificity of 58.46%.

CONCLUSIONS

The results of our study suggested that deep learning algorithms have similar performance to an experienced cardiologist in detecting coronary artery lesions to facilitate the diagnosis of KD.

Machine learning and clinical neurophysiology

Clinical neurophysiology constructs a wealth of dynamic information pertaining to the integrity and function of both central and peripheral nervous systems. As with many technological fields, there has been an explosion of data in neurophysiology over recent years, and this requires considerable analysis by experts. Computational algorithms and especially advances in machine learning (ML) have the ability to assist with this task and potentially reveal hidden insights. In this update article, we will provide a brief overview where such technology is being applied in clinical neurophysiology and possible future directions.

Clinical exome sequencing for inherited retinal degenerations at a tertiary care center

Inherited retinal degenerations are clinically and genetically heterogeneous diseases characterized by progressive deterioration of vision. This study aimed at assessing the diagnostic yield of exome sequencing (ES) for an unselected cohort of individuals with hereditary retinal disorders. It is a retrospective study of 357 unrelated affected individuals, diagnosed with retinal disorders who underwent clinical ES. Variants from ES were filtered, prioritized, and classified using the ACMG recommendations. Clinical diagnosis of the individuals included rod-cone dystrophy (60%), macular dystrophy (20%), cone-rod dystrophy (9%), cone dystrophy (4%) and other phenotypes (7%). Majority of the cases (74%) were singletons and 6% were trios. A confirmed molecular diagnosis was obtained in 24% of cases. In 6% of cases, two pathogenic variants were identified with phase unknown, bringing the potential molecular diagnostic rate to ~ 30%. Including the variants of uncertain significance (VUS), potentially significant findings were reported in 57% of cases. Among cases with a confirmed molecular diagnosis, variants in EYS, ABCA4, USH2A, KIZ, CERKL, DHDDS, PROM1, NR2E3, CNGB1, ABCC6, PRPH2, RHO, PRPF31, PRPF8, SNRNP200, RP1, CHM, RPGR were identified in more than one affected individual. Our results support the utility of clinical ES in the diagnosis of genetically heterogeneous retinal disorders.

Machine Learning Application of Transcranial Motor-Evoked Potential to Predict Positive Functional Outcomes of Patients

Intraoperative neuromonitoring (IONM) has been used to help monitor the integrity of the nervous system during spine surgery. Transcranial motor-evoked potential (TcMEP) has been used lately for lower lumbar surgery to prevent nerve root injuries and also to predict positive functional outcomes of patients. There were a number of studies that proved that the TcMEP signal's improvement is significant towards positive functional outcomes of patients. In this paper, we explored the possibilities of using a machine learning approach to TcMEP signal to predict positive functional outcomes of patients. With 55 patients who underwent various types of lumbar surgeries, the data were divided into 70 : 30 and 80 : 20 ratios for training and testing of the machine learning models. The highest sensitivity and specificity were achieved by Fine KNN of 80 : 20 ratio with 87.5% and 33.33%, respectively. In the meantime, we also tested the existing improvement criteria presented in the literature, and 50% of TcMEP improvement criteria achieved 83.33% sensitivity and 75% specificity. But the rigidness of this threshold method proved unreliable in this study when different datasets were used as the sensitivity and specificity dropped. The proposed method by using machine learning has more room to advance with a larger dataset and various signals' features to choose from.

Artificial Intelligence Meets Neuro-Ophthalmology

ABSTRACT

Recent advances in artificial intelligence have provided ophthalmologists with fast, accurate, and automated means for diagnosing and treating ocular conditions, paving the way to a modern and scalable eye care system. Compared to other ophthalmic disciplines, neuro-ophthalmology has, until recently, not benefitted from significant advances in the area of artificial intelligence. In this narrative review, we summarize and discuss recent advancements utilizing artificial intelligence for the detection of structural and functional optic nerve head abnormalities, and ocular movement disorders in neuro-ophthalmology.

Artificial intelligence-enhanced intraoperative neurosurgical workflow: state of the art and future perspectives

BACKGROUND

Artificial Intelligence (AI) and Machine Learning (ML) augment decision-making processes and productivity by supporting surgeons over a range of clinical activities: from diagnosis and preoperative planning to intraoperative surgical assistance. We reviewed the literature to identify current AI platforms applied to neurosurgical perioperative and intraoperative settings and describe their role in multiple subspecialties.

METHODS

A systematic review of the literature was conducted following the PRISMA guidelines. PubMed, EMBASE, and Scopus databases were searched from inception to December 31, 2020. Original articles were included if they: presented AI platforms implemented in perioperative, intraoperative settings and reported ML models' performance metrics. Due to the heterogeneity in neurosurgical applications, a qualitative synthesis was deemed appropriate. The risk of bias and applicability of predicted outcomes were assessed using the PROBAST tool.

RESULTS

41 articles were included. All studies evaluated a supervised learning algorithm. A total of 10 ML models were described; the most frequent were neural networks (n = 15) and tree-based models (n = 13). Overall, the risk of bias was medium-high, but applicability was considered positive for all studies. Articles were grouped into 4 categories according to the subspecialty of interest: neuro-oncology, spine, functional and other. For each category, different prediction tasks were identified.

CONCLUSIONS

In this review, we summarize the state-of-art applications of AI for the intraoperative augmentation of neurosurgical workflows across multiple subspecialties. ML models may boost surgical team performances by reducing human errors and providing patient-tailored surgical plans, but further and higher-quality studies need to be conducted.

Intraoperative monitoring of visual evoked potentials in patients undergoing transsphenoidal surgery for pituitary adenoma: a systematic review

BACKGROUND

Transsphenoidal surgery is the gold standard for pituitary adenoma resection. Although rare, a serious complication of surgery is worsened vision post-operatively.

OBJECTIVE

To determine whether, in patients undergoing transsphenoidal surgery for pituitary adenoma, intraoperative monitoring of visual evoked potentials (VEP) is a safe, reproducible, and effective technological adjunct in predicting postoperative visual function.

METHODS

The PubMed and OVID platforms were searched between January 1993 and December 2020 to identify publications that (1) featured patients undergoing transsphenoidal surgery for pituitary adenoma, (2) used intraoperative optic nerve monitoring with VEP and (3) reported on safety or effectiveness. Reference lists were cross-checked and expert opinion sought to identify further publications.

RESULTS

Eleven studies were included comprising ten case series and one prospective cohort study. All employed techniques to improve reliability. No safety issues were reported. The only comparative study included described a statistically significant improvement in post-operative visual field testing when VEP monitoring was used. The remaining case-series varied in conclusion. In nine studies, surgical manipulation was halted in the event of a VEP amplitude decrease suggesting a widespread consensus that this is a warning sign of injury to the anterior optic apparatus.

CONCLUSIONS

Despite limited and low-quality published evidence regarding intra-operative VEP monitoring, our review suggests that it is a safe, reproducible, and increasingly effective technique of predicting postoperative visual deficits. Further studies specific to transsphenoidal surgery are required to determine its utility in protecting visual function in the resection of complex pituitary tumours.

Pathogenic variants and associated phenotypic spectrum of TSPAN12 based on data from a large cohort

PURPOSE

The pathogenic variants in TSPAN12 could lead to familial exudative vitreoretinopathy (FEVR), which has high clinical variability. This study aims to assess the pathogenicity of TSPAN12 variants and their phenotypic spectrum based on exome sequencing from 7092 probands with different eye conditions.

METHODS

The variants in TSPAN12 were selected from exome sequencing data of samples from 7092 probands with different forms of eye conditions. Potentially pathogenic variants were evaluated through the annotation of types, locations, population frequencies, and in silico predictions of variants from in-house data, gnomAD, and published literature. The clinical features of patients with potentially pathogenic variants in TSPAN12 were assessed.

RESULTS

A total of 45 variants in TSPAN12 with coding effects were detected based on the exome data from 7092 probands, among which 31 were classified as pathogenic variants including 15 novels. The 31 variants were identified in 34 probands with various initial diagnoses, including FEVR in 21 probands and diseases other than FEVR in the remaining 13 probands. Biallelic pathogenic variants were identified in one proband with initial diagnosis of high myopia.

CONCLUSION

Truncating variants and the missense variants that are predicted as deleterious are likely pathogenic variants of TSPAN12. Approximately 61.8% of patients with pathogenic variants in this gene had an initial diagnosis of FEVR, and the remaining 38.2% of patients had various initial diagnoses. These findings expand the understanding about variant evaluation of TSPAN12 and phenotypic spectrum of TSPAN12-associated FEVR.

Pathogenicity evaluation and the genotype-phenotype analysis of OPA1 variants

Autosomal dominant optic atrophy (ADOA) is an important cause of irreversible visual impairment in children and adolescents. About 60-90% of ADOA is caused by the pathogenic variants of OPA1 gene. By evaluating the pathogenicity of OPA1 variants and summarizing the relationship between the genotype and phenotype, this study aimed to provide a reference for clinical genetic test involving OPA1. Variants in OPA1 were selected from the exome sequencing results in 7092 cases of hereditary eye diseases and control groups from our in-house data. At the same time, the urine cells of some optic atrophy patients with OPA1 variants as well as their family members were collected and oxygen consumption rates (OCR) were measured in these cells to evaluate the pathogenicity of variants. As a result, 97 variants were detected, including 94 rare variants and 3 polymorphisms. And the 94 rare variants were classified into three groups: pathogenic (33), variants of uncertain significance (19), and likely benign (42). Our results indicated that the frameshift variants at the 3' terminus might be pathogenic, while the variants in exon 7 and intron 4 might be benign. The penetrance of the missense variants was higher than that of truncation variants. The OCR of cells with pathogenic OPA1 variants were significantly lower than those without pathogenic variants. In conclusion, some variants might be benign although predicted pathogenic in previous studies while some might have unknown pathogenesis. Measuring the OCR in urine cells could be used as a method to evaluate the pathogenicity of some OPA1 variants.

Pituitary Tumors in the Computational Era, Exploring Novel Approaches to Diagnosis, and Outcome Prediction with Machine Learning

BACKGROUND

Machine learning has emerged as a viable asset in the setting of pituitary surgery. In the past decade, the number of machine learning models developed to aid in the diagnosis of pituitary lesions and predict intraoperative and postoperative complications following transsphenoidal surgery has increased exponentially. As computational processing power continues to increase, big data sets continue to expand, and learning algorithms continue to surpass gold standard predictive tools, machine learning will serve to become an important component in improving patient care and outcomes.

METHODS

Relevant studies were identified based on a literature search in PubMed and MEDLINE databases, as well as from other sources including reference lists of published articles.

RESULTS

Radiomics and artificial neural networks comprise the majority of machine learning-based applications in pituitary surgery. Radiomics serves to quantify specific imaging features, which can then be used to noninvasively identify tumor characteristics and make definitive diagnoses, circumventing presurgical biopsy altogether. Neural networks can be adapted to predict intraoperative changes in visual evoked potentials or cerebral spinal fluid leak. In addition, these algorithms may be combined with others to predict tumor aggressiveness, gross total resection, recurrence and remission, and even total cost burden.

CONCLUSIONS

The field of machine learning is broad, with radiomics and artificial neural networks comprising 2 commonly used supervised learning methods in pituitary surgery. Given the large heterogeneity of pituitary and sellar lesions, the promise of machine learning lies in its ability to identify relationships and patterns that are otherwise hidden from standard statistical methods. While machine learning has great potential as a clinical adjunct during the surgical preplanning process and in predicting complications and outcomes, challenges moving forward include standardization and validation of these paradigms.