Artificial intelligence for clinical decision support in neurology

Prediction begins with diagnosis: Estimating seizure recurrence risk in the First Seizure Clinic

OBJECTIVES

To assess the feasibility of using a seizure recurrence prediction tool in a First Seizure Clinic, considering (1) the accuracy of initial clinical diagnoses and (2) performance of automated computational models in predicting seizure recurrence after first unprovoked seizure (FUS).

METHODS

To assess diagnostic accuracy, we analysed all sustained and revised diagnoses in patients seen at a First Seizure Clinic over 5 years with 6+ months follow-up ('accuracy cohort', n = 487). To estimate prediction of 12-month seizure recurrence after FUS, we used a logistic regression of clinical factors on a multicentre FUS cohort ('prediction cohort', n = 181), and compared performance to a recently published seizure recurrence model.

RESULTS

Initial diagnosis was sustained over 6+ months follow-up in 69% of patients in the 'accuracy cohort'. Misdiagnosis occurred in 5%, and determination of unclassified diagnosis in 9%. Progression to epilepsy occurred in 17%, either following FUS or initial acute symptomatic seizure. Within the 'prediction cohort' with FUS, 12-month seizure recurrence rate was 41% (95% CI [33.8%, 48.5%]). Nocturnal seizure, focal seizure semiology and developmental disability were predictive factors. Our model yielded an Area under the Receiver Operating Characteristic curve (AUC) of 0.60 (95% CI [0.59, 0.64]).

CONCLUSIONS

High clinical accuracy can be achieved at the initial visit to a First Seizure Clinic. This shows that diagnosis will not limit the application of seizure recurrence prediction tools in this context. However, based on the modest performance of currently available seizure recurrence prediction tools using clinical factors, we conclude that data beyond clinical factors alone will be needed to improve predictive performance.

Identifying Facilitators and Barriers to Implementation of AI-Assisted Clinical Decision Support in an Electronic Health Record System

Recent advancements in computing have led to the development of artificial intelligence (AI) enabled healthcare technologies. AI-assisted clinical decision support (CDS) integrated into electronic health records (EHR) was demonstrated to have a significant potential to improve clinical care. With the rapid proliferation of AI-assisted CDS, came the realization that a lack of careful consideration of socio-technical issues surrounding the implementation and maintenance of these tools can result in unanticipated consequences, missed opportunities, and suboptimal uptake of these potentially useful technologies. The 48-h Discharge Prediction Tool (48DPT) is a new AI-assisted EHR CDS to facilitate discharge planning. This study aimed to methodologically assess the implementation of 48DPT and identify the barriers and facilitators of adoption and maintenance using the validated implementation science frameworks. The major dimensions of RE-AIM (Reach, Effectiveness, Adoption, Implementation, Maintenance) and the constructs of the Consolidated Framework for Implementation Research (CFIR) frameworks have been used to analyze interviews of 24 key stakeholders using 48DPT. The systematic assessment of the 48DPT implementation allowed us to describe facilitators and barriers to implementation such as lack of awareness, lack of accuracy and trust, limited accessibility, and transparency. Based on our evaluation, the factors that are crucial for the successful implementation of AI-assisted EHR CDS were identified. Future implementation efforts of AI-assisted EHR CDS should engage the key clinical stakeholders in the AI tool development from the very inception of the project, support transparency and explainability of the AI models, provide ongoing education and onboarding of the clinical users, and obtain continuous input from clinical staff on the CDS performance.

Integrating AI in fighting advancing Alzheimer: diagnosis, prevention, treatment, monitoring, mechanisms, and clinical trials

The application of artificial intelligence (AI) in neurology is a growing field offering opportunities to improve accuracy of diagnosis and treatment of complicated neuronal disorders, plus fostering a deeper understanding of the aetiologies of these diseases through AI-based analyses of large omics data. The most common neurodegenerative disease, Alzheimer's disease (AD), is characterized by brain accumulation of specific pathological proteins, accompanied by cognitive impairment. In this review, we summarize the latest progress on the use of AI in different AD-related fields, such as analysis of neuroimaging data enabling early and accurate AD diagnosis; prediction of AD progression, identification of patients at higher risk and evaluation of new treatments; improvement of the evaluation of drug response using AI algorithms to analyze patient clinical and neuroimaging data; the development of personalized AD therapies; and the use of AI-based techniques to improve the quality of daily life of AD patients and their caregivers.

Computer-aided prognosis of tuberculous meningitis combining imaging and non-imaging data

Tuberculous meningitis (TBM) is the most lethal form of tuberculosis. Clinical features, such as coma, can predict death, but they are insufficient for the accurate prognosis of other outcomes, especially when impacted by co-morbidities such as HIV infection. Brain magnetic resonance imaging (MRI) characterises the extent and severity of disease and may enable more accurate prediction of complications and poor outcomes. We analysed clinical and brain MRI data from a prospective longitudinal study of 216 adults with TBM; 73 (34%) were HIV-positive, a factor highly correlated with mortality. We implemented an end-to-end framework to model clinical and imaging features to predict disease progression. Our model used state-of-the-art machine learning models for automatic imaging feature encoding, and time-series models for forecasting, to predict TBM progression. The proposed approach is designed to be robust to missing data via a novel tailored model optimisation framework. Our model achieved a 60% balanced accuracy in predicting the prognosis of TBM patients over the six different classes. HIV status did not alter the performance of the models. Furthermore, our approach identified brain morphological lesions caused by TBM in both HIV and non-HIV-infected, associating lesions to the disease staging with an overall accuracy of 96%. These results suggest that the lesions caused by TBM are analogous in both populations, regardless of the severity of the disease. Lastly, our models correctly identified changes in disease symptomatology and severity in 80% of the cases. Our approach is the first attempt at predicting the prognosis of TBM by combining imaging and clinical data, via a machine learning model. The approach has the potential to accurately predict disease progression and enable timely clinical intervention.

The Application of Large Language Models for Radiologic Decision Making

BACKGROUND AND PURPOSE

Large language models (LLMs) have seen explosive growth, but their potential role in medical applications remains underexplored. Our study investigates the capability of LLMs to predict the most appropriate imaging study for specific clinical presentations in various subspecialty areas in radiology.

METHODS AND MATERIALS

Chat Generative Pretrained Transformer (ChatGPT), by OpenAI and Glass AI by Glass Health were tested on 1,075 clinical scenarios from 11 ACR expert panels to determine the most appropriate imaging study, benchmarked against the ACR Appropriateness Criteria. Two responses per clinical presentation were generated and averaged for the final clinical presentation score. Clinical presentation scores for each topic area were averaged as its final score. The average of the topic scores within a panel determined the final score of each panel. LLM responses were on a scale of 0 to 3. Partial scores were given for nonspecific answers. Pearson correlation coefficient (R-value) was calculated for each panel to determine a context-specific performance.

RESULTS

Glass AI scored significantly higher than ChatGPT (2.32 ± 0.67 versus 2.08 ± 0.74, P = .002). Both LLMs performed the best in the Polytrauma, Breast, and Vascular panels, and performed the worst in the Neurologic, Musculoskeletal, and Cardiac panels. Glass AI outperformed ChatGPT in 10 of 11 panels, except Obstetrics and Gynecology. Maximum agreement was in the Pediatrics, Neurologic, and Thoracic panels, and the most disagreement occurred in the Vascular, Breast, and Urologic panels.

CONCLUSION

LLMs can be used to predict imaging studies, with Glass AI's superior performance indicating the benefits of extra medical-text training. This supports the potential of LLMs in radiologic decision making.

Evaluations of artificial intelligence and machine learning algorithms in neurodiagnostics

This article evaluates the ethical implications of utilizing artificial intelligence (AI) algorithms in neurological diagnostic examinations. Applications of AI technology have been utilized to aid in the determination of pharmacological dosages of gadolinium for brain lesion detection, localization of seizure foci, and the characterization of large vessel occlusion in ischemic stroke patients. Multiple subtypes of AI/machine learning (ML) algorithms are analyzed, as AI-assisted neurology utilizes supervised, unsupervised, artificial neural network (ANN), and deep neural network (DNN) learning models. As ANN and DNN analyses can be applied to data with an unknown clinical diagnosis, these algorithms are evaluated according to Bayesian statistical analyses. Bayesian neural network analyses are incorporated, as these algorithms indicate that the predictive accuracy and model performance are dependent upon accurate configurations of the model's hyperparameters and neural inputs. Thus, mathematical evaluations of AI algorithms are comprehensively explored to examine their clinical utility, as underperformance of AI/ML models may have deleterious consequences that affect patient outcomes due to misdiagnosis and false-negative test results.

Preliminary Evidence of the Use of Generative AI in Health Care Clinical Services: Systematic Narrative Review

BACKGROUND

Generative artificial intelligence tools and applications (GenAI) are being increasingly used in health care. Physicians, specialists, and other providers have started primarily using GenAI as an aid or tool to gather knowledge, provide information, train, or generate suggestive dialogue between physicians and patients or between physicians and patients' families or friends. However, unless the use of GenAI is oriented to be helpful in clinical service encounters that can improve the accuracy of diagnosis, treatment, and patient outcomes, the expected potential will not be achieved. As adoption continues, it is essential to validate the effectiveness of the infusion of GenAI as an intelligent technology in service encounters to understand the gap in actual clinical service use of GenAI.

OBJECTIVE

This study synthesizes preliminary evidence on how GenAI assists, guides, and automates clinical service rendering and encounters in health care The review scope was limited to articles published in peer-reviewed medical journals.

METHODS

We screened and selected 0.38% (161/42,459) of articles published between January 1, 2020, and May 31, 2023, identified from PubMed. We followed the protocols outlined in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to select highly relevant studies with at least 1 element on clinical use, evaluation, and validation to provide evidence of GenAI use in clinical services. The articles were classified based on their relevance to clinical service functions or activities using the descriptive and analytical information presented in the articles.

RESULTS

Of 161 articles, 141 (87.6%) reported using GenAI to assist services through knowledge access, collation, and filtering. GenAI was used for disease detection (19/161, 11.8%), diagnosis (14/161, 8.7%), and screening processes (12/161, 7.5%) in the areas of radiology (17/161, 10.6%), cardiology (12/161, 7.5%), gastrointestinal medicine (4/161, 2.5%), and diabetes (6/161, 3.7%). The literature synthesis in this study suggests that GenAI is mainly used for diagnostic processes, improvement of diagnosis accuracy, and screening and diagnostic purposes using knowledge access. Although this solves the problem of knowledge access and may improve diagnostic accuracy, it is oriented toward higher value creation in health care.

CONCLUSIONS

GenAI informs rather than assisting or automating clinical service functions in health care. There is potential in clinical service, but it has yet to be actualized for GenAI. More clinical service-level evidence that GenAI is used to streamline some functions or provides more automated help than only information retrieval is needed. To transform health care as purported, more studies related to GenAI applications must automate and guide human-performed services and keep up with the optimism that forward-thinking health care organizations will take advantage of GenAI.

Combined Edge Loss UNet for Optimized Segmentation in Total Knee Arthroplasty Preoperative Planning

Bone segmentation and 3D reconstruction are crucial for total knee arthroplasty (TKA) surgical planning with Personalized Surgical Instruments (PSIs). Traditional semi-automatic approaches are time-consuming and operator-dependent, although they provide reliable outcomes. Moreover, the recent expansion of artificial intelligence (AI) tools towards various medical domains is transforming modern healthcare. Accordingly, this study introduces an automated AI-based pipeline to replace the current operator-based tibia and femur 3D reconstruction procedure enhancing TKA preoperative planning. Leveraging an 822 CT image dataset, a novel patch-based method and an improved segmentation label generation algorithm were coupled to a Combined Edge Loss UNet (CEL-UNet), a novel CNN architecture featuring an additional decoding branch to boost the bone boundary segmentation. Root Mean Squared Errors and Hausdorff distances compared the predicted surfaces to the reference bones showing median and interquartile values of 0.26 (0.19-0.36) mm and 0.24 (0.18-0.32) mm, and of 1.06 (0.73-2.15) mm and 1.43 (0.82-2.86) mm for the tibia and femur, respectively, outperforming previous results of our group, state-of-the-art, and UNet models. A feasibility analysis for a PSI-based surgical plan revealed sub-millimetric distance errors and sub-angular alignment uncertainties in the PSI contact areas and the two cutting planes. Finally, operational environment testing underscored the pipeline's efficiency. More than half of the processed cases complied with the PSI prototyping requirements, reducing the overall time from 35 min to 13.1 s, while the remaining ones underwent a manual refinement step to achieve such PSI requirements, performing the procedure four to eleven times faster than the manufacturer standards. To conclude, this research advocates the need for real-world applicability and optimization of AI solutions in orthopedic surgical practice.

Performance of Large Language Models on a Neurology Board-Style Examination

Importance

Recent advancements in large language models (LLMs) have shown potential in a wide array of applications, including health care. While LLMs showed heterogeneous results across specialized medical board examinations, the performance of these models in neurology board examinations remains unexplored.

Objective

To assess the performance of LLMs on neurology board-style examinations.

Design, Setting, and Participants

This cross-sectional study was conducted between May 17 and May 31, 2023. The evaluation utilized a question bank approved by the American Board of Psychiatry and Neurology and was validated with a small question cohort by the European Board for Neurology. All questions were categorized into lower-order (recall, understanding) and higher-order (apply, analyze, synthesize) questions based on the Bloom taxonomy for learning and assessment. Performance by LLM ChatGPT versions 3.5 (LLM 1) and 4 (LLM 2) was assessed in relation to overall scores, question type, and topics, along with the confidence level and reproducibility of answers.

Main Outcomes and Measures

Overall percentage scores of 2 LLMs.

Results

LLM 2 significantly outperformed LLM 1 by correctly answering 1662 of 1956 questions (85.0%) vs 1306 questions (66.8%) for LLM 1. Notably, LLM 2's performance was greater than the mean human score of 73.8%, effectively achieving near-passing and passing grades in the neurology board examination. LLM 2 outperformed human users in behavioral, cognitive, and psychological-related questions and demonstrated superior performance to LLM 1 in 6 categories. Both LLMs performed better on lower-order than higher-order questions, with LLM 2 excelling in both lower-order and higher-order questions. Both models consistently used confident language, even when providing incorrect answers. Reproducible answers of both LLMs were associated with a higher percentage of correct answers than inconsistent answers.

Conclusions and Relevance

Despite the absence of neurology-specific training, LLM 2 demonstrated commendable performance, whereas LLM 1 performed slightly below the human average. While higher-order cognitive tasks were more challenging for both models, LLM 2's results were equivalent to passing grades in specialized neurology examinations. These findings suggest that LLMs could have significant applications in clinical neurology and health care with further refinements.

Assessment of ChatGPT's performance on neurology written board examination questions

Background and objectives

ChatGPT has shown promise in healthcare. To assess the utility of this novel tool in healthcare education, we evaluated ChatGPT's performance in answering neurology board exam questions.

Methods

Neurology board-style examination questions were accessed from BoardVitals, a commercial neurology question bank. ChatGPT was provided a full question prompt and multiple answer choices. First attempts and additional attempts up to three tries were given to ChatGPT to select the correct answer. A total of 560 questions (14 blocks of 40 questions) were used, although any image-based questions were disregarded due to ChatGPT's inability to process visual input. The artificial intelligence (AI) answers were then compared with human user data provided by the question bank to gauge its performance.

Results

Out of 509 eligible questions over 14 question blocks, ChatGPT correctly answered 335 questions (65.8%) on the first attempt/iteration and 383 (75.3%) over three attempts/iterations, scoring at approximately the 26th and 50th percentiles, respectively. The highest performing subjects were pain (100%), epilepsy & seizures (85%) and genetic (82%) while the lowest performing subjects were imaging/diagnostic studies (27%), critical care (41%) and cranial nerves (48%).

Discussion

This study found that ChatGPT performed similarly to its human counterparts. The accuracy of the AI increased with multiple attempts and performance fell within the expected range of neurology resident learners. This study demonstrates ChatGPT's potential in processing specialised medical information. Future studies would better define the scope to which AI would be able to integrate into medical decision making.

An Explainable and Robust Deep Learning Approach for Automated Electroencephalography-based Schizophrenia Diagnosis

Schizophrenia (SZ) is a neuropsychiatric disorder that affects millions globally. Current diagnosis of SZ is symptom-based, which poses difficulty due to the variability of symptoms across patients. To this end, many recent studies have developed deep learning methods for automated diagnosis of SZ, especially using raw EEG, which provides high temporal precision. For such methods to be productionized, they must be both explainable and robust. Explainable models are essential to identify biomarkers of SZ, and robust models are critical to learn generalizable patterns, especially amidst changes in the implementation environment. One common example is channel loss during EEG recording, which could be detrimental to classifier performance. In this study, we developed a novel channel dropout (CD) approach to increase the robustness of explainable deep learning models trained on EEG data for SZ diagnosis to channel loss. We developed a baseline convolutional neural network (CNN) architecture and implement our approach as a CD layer added to the baseline (CNN-CD). We then applied two explainability approaches to both models for insight into learned spatial and spectral features and show that the application of CD decreases model sensitivity to channel loss. The CNN and CNN-CD achieved accuracies of 81.9% and 80.9% on testing data, respectively. Furthermore, our models heavily prioritized the parietal electrodes and the α-band, which is supported by existing literature. It is our hope that this study motivates the further development of explainable and robust models and bridges the transition from research to application in a clinical decision support role.

Deep learning for automated detection of generalized paroxysmal fast activity in Lennox-Gastaut syndrome

OBJECTIVES

Generalized paroxysmal fast activity (GPFA) is a key electroencephalographic (EEG) feature of Lennox-Gastaut Syndrome (LGS). Automated analysis of scalp EEG has been successful in detecting more typical abnormalities. Automatic detection of GPFA has been more challenging, due to its variability from patient to patient and similarity to normal brain rhythms. In this work, a deep learning model is investigated for detection of GPFA events and estimating their overall burden from scalp EEG.

METHODS

Data from 10 patients recorded during four ambulatory EEG monitoring sessions are used to generate and validate the model. All patients had confirmed LGS and were recruited into a trial for thalamic deep-brain stimulation therapy (ESTEL Trial).

RESULTS

The correlation coefficient between manual and model estimates of event counts was r2 = 0.87, and for total burden was r2 = 0.91. The average GPFA detection sensitivity was 0.876, with an average false-positive rate of 3.35 per minute. There was no significant difference found between patients with early or delayed deep brain stimulation (DBS) treatment, or those with active vagal nerve stimulation (VNS).

CONCLUSIONS

Overall, the deep learning model was able to accurately detect GPFA and provide accurate estimates of the overall GPFA burden and electrographic event counts, albeit with a high false-positive rate.

SIGNIFICANCE

Automated GPFA detection may enable automated calculation of EEG biomarkers of burden of disease in LGS.

Enhancing the Evidence with Algorithms: How Artificial Intelligence Is Transforming Forensic Medicine

BACKGROUND

The integration of artificial intelligence (AI) into various fields has ushered in a new era of multidisciplinary progress. Defined as the ability of a system to interpret external data, learn from it, and adapt to specific tasks, AI is poised to revolutionize the world. In forensic medicine and pathology, algorithms play a crucial role in data analysis, pattern recognition, anomaly identification, and decision making. This review explores the diverse applications of AI in forensic medicine, encompassing fields such as forensic identification, ballistics, traumatic injuries, postmortem interval estimation, forensic toxicology, and more.

RESULTS

A thorough review of 113 articles revealed a subset of 32 papers directly relevant to the research, covering a wide range of applications. These included forensic identification, ballistics and additional factors of shooting, traumatic injuries, post-mortem interval estimation, forensic toxicology, sexual assaults/rape, crime scene reconstruction, virtual autopsy, and medical act quality evaluation. The studies demonstrated the feasibility and advantages of employing AI technology in various facets of forensic medicine and pathology.

CONCLUSIONS

The integration of AI in forensic medicine and pathology offers promising prospects for improving accuracy and efficiency in medico-legal practices. From forensic identification to post-mortem interval estimation, AI algorithms have shown the potential to reduce human subjectivity, mitigate errors, and provide cost-effective solutions. While challenges surrounding ethical considerations, data security, and algorithmic correctness persist, continued research and technological advancements hold the key to realizing the full potential of AI in forensic applications. As the field of AI continues to evolve, it is poised to play an increasingly pivotal role in the future of forensic medicine and pathology.

Advancing Patient Care: How Artificial Intelligence Is Transforming Healthcare

Artificial Intelligence (AI) has emerged as a transformative technology with immense potential in the field of medicine. By leveraging machine learning and deep learning, AI can assist in diagnosis, treatment selection, and patient monitoring, enabling more accurate and efficient healthcare delivery. The widespread implementation of AI in healthcare has the role to revolutionize patients' outcomes and transform the way healthcare is practiced, leading to improved accessibility, affordability, and quality of care. This article explores the diverse applications and reviews the current state of AI adoption in healthcare. It concludes by emphasizing the need for collaboration between physicians and technology experts to harness the full potential of AI.

May Artificial Intelligence Influence Future Pediatric Research?-The Case of ChatGPT

BACKGROUND

In recent months, there has been growing interest in the potential of artificial intelligence (AI) to revolutionize various aspects of medicine, including research, education, and clinical practice. ChatGPT represents a leading AI language model, with possible unpredictable effects on the quality of future medical research, including clinical decision-making, medical education, drug development, and better research outcomes.

AIM AND METHODS

In this interview with ChatGPT, we explore the potential impact of AI on future pediatric research. Our discussion covers a range of topics, including the potential positive effects of AI, such as improved clinical decision-making, enhanced medical education, faster drug development, and better research outcomes. We also examine potential negative effects, such as bias and fairness concerns, safety and security issues, overreliance on technology, and ethical considerations.

CONCLUSIONS

While AI continues to advance, it is crucial to remain vigilant about the possible risks and limitations of these technologies and to consider the implications of these technologies and their use in the medical field. The development of AI language models represents a significant advancement in the field of artificial intelligence and has the potential to revolutionize daily clinical practice in every branch of medicine, both surgical and clinical. Ethical and social implications must also be considered to ensure that these technologies are used in a responsible and beneficial manner.

Visual deep learning of unprocessed neuroimaging characterises dementia subtypes and generalises across non-stereotypic samples

BACKGROUND

Dementia's diagnostic protocols are mostly based on standardised neuroimaging data collected in the Global North from homogeneous samples. In other non-stereotypical samples (participants with diverse admixture, genetics, demographics, MRI signals, or cultural origins), classifications of disease are difficult due to demographic and region-specific sample heterogeneities, lower quality scanners, and non-harmonised pipelines.

METHODS

We implemented a fully automatic computer-vision classifier using deep learning neural networks. A DenseNet was applied on raw (unpreprocessed) data from 3000 participants (behavioural variant frontotemporal dementia-bvFTD, Alzheimer's disease-AD, and healthy controls; both male and female as self-reported by participants). We tested our results in demographically matched and unmatched samples to discard possible biases and performed multiple out-of-sample validations.

FINDINGS

Robust classification results across all groups were achieved from standardised 3T neuroimaging data from the Global North, which also generalised to standardised 3T neuroimaging data from Latin America. Moreover, DenseNet also generalised to non-standardised, routine 1.5T clinical images from Latin America. These generalisations were robust in samples with heterogenous MRI recordings and were not confounded by demographics (i.e., were robust in both matched and unmatched samples, and when incorporating demographic variables in a multifeatured model). Model interpretability analysis using occlusion sensitivity evidenced core pathophysiological regions for each disease (mainly the hippocampus in AD, and the insula in bvFTD) demonstrating biological specificity and plausibility.

INTERPRETATION

The generalisable approach outlined here could be used in the future to aid clinician decision-making in diverse samples.

FUNDING

The specific funding of this article is provided in the acknowledgements section.

Experience of waiting for seizure freedom and perception of machine learning technologies to support treatment decision: A qualitative study in adults with recent onset epilepsy

PURPOSE

With no reliable surrogate biomarkers for treatment response, people with epilepsy currently await the passage of time to determine whether prescribed treatments are effective. Few studies have examined the issues faced by people with epilepsy during this waiting period. We aim to explore the experiences of people with recently diagnosed epilepsy as they wait to achieve seizure freedom.

METHODS

We purposively sampled adults of working age who had been diagnosed and treated for epilepsy for less than four years. Semi-structured interviews were undertaken between July and September 2021. A thematic analysis using a framework approach was performed.

RESULTS

We recruited 15 patients. Results revealed four main themes: 1) Impact on mental health, as people with newly diagnosed epilepsy described waiting for seizure freedom as a time of vulnerability, uncertainty, and confusion. 2) Participants described their life as "on hold", prior to achieving effective seizure control 3) Difficulty navigating health systems to find and understand information about epilepsy, tests, and medications, and to find the 'right' health professional to address their needs. 4) Technology systems that support clinician decision making with selecting effective medications early after diagnosis were cautiously welcomed by participants.

CONCLUSION

Interventions are needed to reduce the negative impacts experienced by people who are newly diagnosed with epilepsy while waiting for effective seizure control. Technology systems that support clinician decision making were acceptable, as people with epilepsy sought accessible and effective solutions to restore a sense of control in their lives.

Neurology education in the era of artificial intelligence

PURPOSE OF REVIEW

The practice of neurology is undergoing a paradigm shift because of advances in the field of data science, artificial intelligence, and machine learning. To ensure a smooth transition, physicians must have the knowledge and competence to apply these technologies in clinical practice. In this review, we describe physician perception and preparedness, as well as current state for clinical applications of artificial intelligence and machine learning in neurology.

RECENT FINDINGS

Digital health including artificial intelligence-based/machine learning-based technology has made significant inroads into various aspects of healthcare including neurological care. Surveys of physicians and healthcare stakeholders suggests an overall positive perception about the benefits of artificial intelligence/machine learning in clinical practice. This positive perception is tempered by concerns for lack of knowledge and limited opportunities to build competence in artificial intelligence/machine learning technology. Literature about neurologist's perception and preparedness towards artificial intelligence/machine learning-based technology is scant. There are very few opportunities for physicians particularly neurologists to learn about artificial intelligence/machine learning-based technology.

SUMMARY

Neurologists have not been surveyed about their perception and preparedness to adopt artificial intelligence/machine learning-based technology in clinical practice. We propose development of a practical artificial intelligence/machine learning curriculum to enhance neurologists' competence in these newer technologies.

Machine Learning Model to Predict Diagnosis of Mild Cognitive Impairment by Using Radiomic and Amyloid Brain PET

PURPOSE

The study aimed to develop a deep learning model for predicting amnestic mild cognitive impairment (aMCI) diagnosis using radiomic features and amyloid brain PET.

PATIENTS AND METHODS

Subjects (n = 328) from the Alzheimer's Disease Neuroimaging Initiative database and the EudraCT 2015-001184-39 trial (159 males, 169 females), with a mean age of 72 ± 7.4 years, underwent PET/CT with 18 F-florbetaben. The study cohort consisted of normal controls (n = 149) and subjects with aMCI (n = 179). Thirteen gray-level run-length matrix radiomic features and amyloid loads were extracted from 27 cortical brain areas. The least absolute shrinkage and selection operator regression was used to select features with the highest predictive value. A feed-forward neural multilayer network was trained, validated, and tested on 70%, 15%, and 15% of the sample, respectively. Accuracy, precision, F1-score, and area under the curve were used to assess model performance. SUV performance in predicting the diagnosis of aMCI was also assessed and compared with that obtained from the machine learning model.

RESULTS

The machine learning model achieved an area under the receiver operating characteristic curve of 90% (95% confidence interval, 89.4-90.4) on the test set, with 80% and 78% for accuracy and F1-score, respectively. The deep learning model outperformed SUV performance (area under the curve, 71%; 95% confidence interval, 69.7-71.4; 57% accuracy, 48% F1-score).

CONCLUSIONS

Using radiomic and amyloid PET load, the machine learning model identified MCI subjects with 84% specificity at 81% sensitivity. These findings show that a deep learning algorithm based on radiomic data and amyloid load obtained from brain PET images improves the prediction of MCI diagnosis compared with SUV alone.

Cardiovascular Diseases in the Digital Health Era: A Translational Approach from the Lab to the Clinic

Translational science has been introduced as the nexus among the scientific and the clinical field, which allows researchers to provide and demonstrate that the evidence-based research can connect the gaps present between basic and clinical levels. This type of research has played a major role in the field of cardiovascular diseases, where the main objective has been to identify and transfer potential treatments identified at preclinical stages into clinical practice. This transfer has been enhanced by the intromission of digital health solutions into both basic research and clinical scenarios. This review aimed to identify and summarize the most important translational advances in the last years in the cardiovascular field together with the potential challenges that still remain in basic research, clinical scenarios, and regulatory agencies.

Multimodal deep learning for Alzheimer's disease dementia assessment

Worldwide, there are nearly 10 million new cases of dementia annually, of which Alzheimer's disease (AD) is the most common. New measures are needed to improve the diagnosis of individuals with cognitive impairment due to various etiologies. Here, we report a deep learning framework that accomplishes multiple diagnostic steps in successive fashion to identify persons with normal cognition (NC), mild cognitive impairment (MCI), AD, and non-AD dementias (nADD). We demonstrate a range of models capable of accepting flexible combinations of routinely collected clinical information, including demographics, medical history, neuropsychological testing, neuroimaging, and functional assessments. We then show that these frameworks compare favorably with the diagnostic accuracy of practicing neurologists and neuroradiologists. Lastly, we apply interpretability methods in computer vision to show that disease-specific patterns detected by our models track distinct patterns of degenerative changes throughout the brain and correspond closely with the presence of neuropathological lesions on autopsy. Our work demonstrates methodologies for validating computational predictions with established standards of medical diagnosis.

Towards a guideline for evaluation metrics in medical image segmentation

In the last decade, research on artificial intelligence has seen rapid growth with deep learning models, especially in the field of medical image segmentation. Various studies demonstrated that these models have powerful prediction capabilities and achieved similar results as clinicians. However, recent studies revealed that the evaluation in image segmentation studies lacks reliable model performance assessment and showed statistical bias by incorrect metric implementation or usage. Thus, this work provides an overview and interpretation guide on the following metrics for medical image segmentation evaluation in binary as well as multi-class problems: Dice similarity coefficient, Jaccard, Sensitivity, Specificity, Rand index, ROC curves, Cohen's Kappa, and Hausdorff distance. Furthermore, common issues like class imbalance and statistical as well as interpretation biases in evaluation are discussed. As a summary, we propose a guideline for standardized medical image segmentation evaluation to improve evaluation quality, reproducibility, and comparability in the research field.

IDEAL approach to the evaluation of machine learning technology in epilepsy surgery: protocol for the MAST trial

Epilepsy and epilepsy surgery lend themselves well to the application of machine learning (ML) and artificial intelligence (AI) technologies. This is evidenced by the plethora of tools developed for applications such as seizure detection and analysis of imaging and electrophysiological data. However, few of these tools have been directly used to guide patient management. In recent years, the Idea, Development, Exploration, Assessment, Long-Term Follow-Up (IDEAL) collaboration has formalised stages for the evaluation of surgical innovation and medical devices, and, in many ways, this pragmatic framework is also applicable to ML/AI technology, balancing innovation and safety. In this protocol paper, we outline the preclinical (IDEAL stage 0) evaluation and the protocol for a prospective (IDEAL stage 1/2a) study to evaluate the utility of an ML lesion detection algorithm designed to detect focal cortical dysplasia from structural MRI, as an adjunct in the planning of stereoelectroencephalography trajectories in children undergoing intracranial evaluation for drug-resistant epilepsy.

Multicenter Validation of Individual Preoperative Motor Outcome Prediction for Deep Brain Stimulation in Parkinson's Disease

BACKGROUND

Subthalamic nucleus deep brain stimulation (STN DBS) is an established therapy for Parkinson's disease (PD) patients suffering from motor response fluctuations despite optimal medical treatment, or severe dopaminergic side effects. Despite careful clinical selection and surgical procedures, some patients do not benefit from STN DBS. Preoperative prediction models are suggested to better predict individual motor response after STN DBS. We validate a preregistered model, DBS-PREDICT, in an external multicenter validation cohort.

METHODS

DBS-PREDICT considered eleven, solely preoperative, clinical characteristics and applied a logistic regression to differentiate between weak and strong motor responders. Weak motor response was defined as no clinically relevant improvement on the Unified Parkinson's Disease Rating Scale (UPDRS) II, III, or IV, 1 year after surgery, defined as, respectively, 3, 5, and 3 points or more. Lower UPDRS III and IV scores and higher age at disease onset contributed most to weak response predictions. Individual predictions were compared with actual clinical outcomes.

RESULTS

322 PD patients treated with STN DBS from 6 different centers were included. DBS-PREDICT differentiated between weak and strong motor responders with an area under the receiver operator curve of 0.76 and an accuracy up to 77%.

CONCLUSION

Proving generalizability and feasibility of preoperative STN DBS outcome prediction in an external multicenter cohort is an important step in creating clinical impact in DBS with data-driven tools. Future prospective studies are required to overcome several inherent practical and statistical limitations of including clinical decision support systems in DBS care.

Artificial intelligence in arthroplasty

Artificial intelligence (AI) is altering the world of medicine. Given the rapid advances in technology, computers are now able to learn and improve, imitating humanoid cognitive function. AI applications currently exist in various medical specialties, some of which are already in clinical use. This review presents the potential uses and limitations of AI in arthroplasty to provide a better understanding of the existing technology and future direction of this field.Recent literature demonstrates that the utilization of AI in the field of arthroplasty has the potential to improve patient care through better diagnosis, screening, planning, monitoring, and prediction. The implementation of AI technology will enable arthroplasty surgeons to provide patient-specific management in clinical decision making, preoperative health optimization, resource allocation, decision support, and early intervention. While this technology presents a variety of exciting opportunities, it also has several limitations and challenges that need to be overcome to ensure its safety and effectiveness.

Guidelines for Conducting Ethical Artificial Intelligence Research in Neurology: A Systematic Approach for Clinicians and Researchers

Pre-emptive recognition of the ethical implications of study design and algorithm choices in artificial intelligence (AI) research is an important but challenging process. AI applications have begun to transition from a promising future to clinical reality in neurology. As the clinical management of neurology is often concerned with discrete, often unpredictable, and highly consequential events linked to multimodal data streams over long timescales, forthcoming advances in AI have great potential to transform care for patients. However, critical ethical questions have been raised with implementation of the first AI applications in clinical practice. Clearly, AI will have far-reaching potential to promote, but also to endanger, ethical clinical practice. This article employs an anticipatory ethics approach to scrutinize how researchers in neurology can methodically identify ethical ramifications of design choices early in the research and development process, with a goal of pre-empting unintended consequences that may violate principles of ethical clinical care. First, we discuss the use of a systematic framework for researchers to identify ethical ramifications of various study design and algorithm choices. Second, using epilepsy as a paradigmatic example, anticipatory clinical scenarios that illustrate unintended ethical consequences are discussed, and failure points in each scenario evaluated. Third, we provide practical recommendations for understanding and addressing ethical ramifications early in methods development stages. Awareness of the ethical implications of study design and algorithm choices that may unintentionally enter AI is crucial to ensuring that incorporation of AI into neurology care leads to patient benefit rather than harm.

Critical Appraisal of a Machine Learning Paper: A Guide for the Neurologist

Machine learning (ML), a form of artificial intelligence (AI), is being increasingly employed in neurology. Reported performance metrics often match or exceed the efficiency of average clinicians. The neurologist is easily baffled by the underlying concepts and terminologies associated with ML studies. The superlative performance metrics of ML algorithms often hide the opaque nature of its inner workings. Questions regarding ML model's interpretability and reproducibility of its results in real-world scenarios, need emphasis. Given an abundance of time and information, the expert clinician should be able to deliver comparable predictions to ML models, a useful benchmark while evaluating its performance. Predictive performance metrics of ML models should not be confused with causal inference between its input and output. ML and clinical gestalt should compete in a randomized controlled trial before they can complement each other for screening, triaging, providing second opinions and modifying treatment.

Artificial Intelligence in Epilepsy

Background

The study of seizure patterns in electroencephalography (EEG) requires several years of intensive training. In addition, inadequate training and human error may lead to misinterpretation and incorrect diagnosis. Artificial intelligence (AI)-based automated seizure detection systems hold an exciting potential to create paradigms for proper diagnosis and interpretation. AI holds the promise to transform healthcare into a system where machines and humans can work together to provide an accurate, timely diagnosis, and treatment to the patients.

Objective

This article presents a brief overview of research on the use of AI systems for pattern recognition in EEG for clinical diagnosis.

Material and Methods

The article begins with the need for understanding nonstationary signals such as EEG and simplifying their complexity for accurate pattern recognition in medical diagnosis. It also explains the core concepts of AI, machine learning (ML), and deep learning (DL) methods.

Results and Conclusions

In this present context of epilepsy diagnosis, AI may work in two ways; first by creating visual representations (e.g., color-coded paradigms), which allow persons with limited training to make a diagnosis. The second is by directly explaining a complete automated analysis, which of course requires more complex paradigms than the previous one. We also clarify that AI is not about replacing doctors and strongly emphasize the need for domain knowledge in building robust AI models that can work in real-time scenarios rendering good detection accuracy in a minimum amount of time.

Enhancing trust in AI through industry self-governance

Artificial intelligence (AI) is critical to harnessing value from exponentially growing health and healthcare data. Expectations are high for AI solutions to effectively address current health challenges. However, there have been prior periods of enthusiasm for AI followed by periods of disillusionment, reduced investments, and progress, known as "AI Winters." We are now at risk of another AI Winter in health/healthcare due to increasing publicity of AI solutions that are not representing touted breakthroughs, and thereby decreasing trust of users in AI. In this article, we first highlight recently published literature on AI risks and mitigation strategies that would be relevant for groups considering designing, implementing, and promoting self-governance. We then describe a process for how a diverse group of stakeholders could develop and define standards for promoting trust, as well as AI risk-mitigating practices through greater industry self-governance. We also describe how adherence to such standards could be verified, specifically through certification/accreditation. Self-governance could be encouraged by governments to complement existing regulatory schema or legislative efforts to mitigate AI risks. Greater adoption of industry self-governance could fill a critical gap to construct a more comprehensive approach to the governance of AI solutions than US legislation/regulations currently encompass. In this more comprehensive approach, AI developers, AI users, and government/legislators all have critical roles to play to advance practices that maintain trust in AI and prevent another AI Winter.

Perinatal health predictors using artificial intelligence: A review

Advances in public health and medical care have enabled better pregnancy and birth outcomes. The rates of perinatal health indicators such as maternal mortality and morbidity; fetal, neonatal, and infant mortality; low birthweight; and preterm birth have reduced over time. However, they are still a public health concern, and considerable disparities exist within and between countries. For perinatal researchers who are engaged in unraveling the tangled web of causation for maternal and child health outcomes and for clinicians involved in the care of pregnant women and infants, artificial intelligence offers novel approaches to prediction modeling, diagnosis, early detection, and monitoring in perinatal health. Machine learning, a commonly used artificial intelligence method, has been used to predict preterm birth, birthweight, preeclampsia, mortality, hypertensive disorders, and postpartum depression. Real-time electronic health recording and predictive modeling using artificial intelligence have found early success in fetal monitoring and monitoring of women with gestational diabetes especially in low-resource settings. Artificial intelligence-based methodologies have the potential to improve prenatal diagnosis of birth defects and outcomes in assisted reproductive technology too. In this scenario, we envision artificial intelligence for perinatal research to be based on three goals: (1) availability of population-representative, routine clinical data (rich multimodal data of large sample size) for perinatal research; (2) modification and application of current state-of-the-art artificial intelligence for prediction and classification in health care research to the field of perinatal health; and (3) development of methods for explaining the decision-making processes of artificial intelligence models for perinatal health indicators. Achieving these three goals via a multidisciplinary approach to the development of artificial intelligence tools will enable trust in these tools and advance research, clinical practice, and policies to ensure optimal perinatal health.

Teleneuropsychology in the time of COVID-19: The experience of The Australian Epilepsy Project

PURPOSE

Traditional neuropsychological testing carries elevated COVID-19 risk for both examinee and examiner. Here we describe how the pilot study of the Australian Epilepsy Project (AEP) has transitioned to tele-neuropsychology (teleNP), enabling continued safe operations during the pandemic.

METHODS

The AEP includes adults (age 18-60) with a first unprovoked seizure, new diagnosis of epilepsy or drug resistant focal epilepsy. Shortly after launching the study, COVID-related restrictions necessitated adaptation to teleNP, including delivery of verbal tasks via videoconference; visual stimulus delivery via document camera; use of web-hosted, computerised assessment; substitution of oral versions for written tests; online delivery of questionnaires; and discontinuation of telehealth incompatible tasks.

RESULTS

To date, we have completed 24 teleNP assessments: 18 remotely (participant in own home) and six on-site (participant using equipment at research facility). Five face-to-face assessments were conducted prior to the transition to teleNP. Eight of 408 tests administered via teleNP (1.9 %) have been invalidated, for a variety of reasons (technical, procedural, environmental). Data confirm typical patterns of epilepsy-related deficits (p < .05) affecting processing speed, executive function, language and memory. Questionnaire responses indicate elevated rates of patients at high risk of mood (34 %) and anxiety disorder (38 %).

CONCLUSION

Research teleNP assessments reveal a typical pattern of impairments in epilepsy. A range of issues must be considered when introducing teleNP, such as technical and administrative set up, test selection and delivery, and cohort suitability. TeleNP enables large-scale neuropsychological research during periods of social distancing (and beyond), and offers an opportunity to expand the reach and breadth of neuropsychological services.