Representation of internal speech by single neurons in human supramarginal gyrus.

Speech brain-machine interfaces (BMIs) translate brain signals into words or audio outputs, enabling communication for people having lost their speech abilities due to diseases or injury. While important advances in vocalized, attempted and mimed speech decoding have been achieved, results for internal speech decoding are sparse and have yet to achieve high functionality. Notably, it is still unclear from which brain areas internal speech can be decoded. Here two participants with tetraplegia with implanted microelectrode arrays located in the supramarginal gyrus (SMG) and primary somatosensory cortex (S1) performed internal and vocalized speech of six words and two pseudowords. In both participants, we found significant neural representation of internal and vocalized speech, at the single neuron and population level in the SMG. From recorded population activity in the SMG, the internally spoken and vocalized words were significantly decodable. In an offline analysis, we achieved average decoding accuracies of 55% and 24% for each participant, respectively (chance level 12.5%), and during an online internal speech BMI task, we averaged 79% and 23% accuracy, respectively. Evidence of shared neural representations between internal speech, word reading and vocalized speech processes was found in participant 1. SMG represented words as well as pseudowords, providing evidence for phonetic encoding. Furthermore, our decoder achieved high classification with multiple internal speech strategies (auditory imagination/visual imagination). Activity in S1 was modulated by vocalized but not internal speech in both participants, suggesting no articulator movements of the vocal tract occurred during internal speech production. This work represents a proof-of-concept for a high-performance internal speech BMI.

Understanding the human conflict processing network: A review of the literature on direct neural recordings during performance of a modified stroop task.

The Stroop Task is a well-known neuropsychological task developed to investigate conflict processing in the human brain. Our group has utilized direct intracranial neural recordings in various brain regions during performance of a modified color-word Stroop Task to gain a mechanistic understanding of non-emotional human conflict processing. The purpose of this review article is to: 1) synthesize our own studies into a model of human conflict processing, 2) review the current literature on the Stroop Task and other conflict tasks to put our research in context, and 3) describe how these studies define a network in conflict processing. The figures presented are reprinted from our prior publications and key publications referenced in the manuscript. We summarize all studies to date that employ invasive intracranial recordings in humans during performance of conflict-inducing tasks. For our own studies, we analyzed local field potentials (LFPs) from patients with implanted stereotactic electroencephalography (SEEG) electrodes, and we observed intracortical oscillation patterns as well as intercortical temporal relationships in the hippocampus, amygdala, and orbitofrontal cortex (OFC) during the cue-processing phase of a modified Stroop Task. Our findings suggest that non-emotional human conflict processing involves modulation across multiple frequency bands within and between brain structures.

Theta-frequency medial septal nucleus deep brain stimulation increases neurovascular activity in MK-801-treated mice.

Introduction: Deep brain stimulation (DBS) has shown remarkable success treating neurological and psychiatric disorders including Parkinson's disease, essential tremor, dystonia, epilepsy, and obsessive-compulsive disorder. DBS is now being explored to improve cognitive and functional outcomes in other psychiatric conditions, such as those characterized by reduced N-methyl-D-aspartate (NMDA) function (i.e., schizophrenia). While DBS for movement disorders generally involves high-frequency (>100 Hz) stimulation, there is evidence that low-frequency stimulation may have beneficial and persisting effects when applied to cognitive brain networks. Methods: In this study, we utilize a novel technology, functional ultrasound imaging (fUSI), to characterize the cerebrovascular impact of medial septal nucleus (MSN) DBS under conditions of NMDA antagonism (pharmacologically using Dizocilpine [MK-801]) in anesthetized male mice. Results: Imaging from a sagittal plane across a variety of brain regions within and outside of the septohippocampal circuit, we find that MSN theta-frequency (7.7 Hz) DBS increases hippocampal cerebral blood volume (CBV) during and after stimulation. This effect was not present using standard high-frequency stimulation parameters [i.e., gamma (100 Hz)]. Discussion: These results indicate the MSN DBS increases circuit-specific hippocampal neurovascular activity in a frequency-dependent manner and does so in a way that continues beyond the period of electrical stimulation.

Neuromodulation of Eating Disorders: A Review of Underlying Neural Network Activity and Neuromodulatory Treatments.

Eating disorders are a group of psychiatric conditions that involve pathological relationships between patients and food. The most prolific of these disorders are anorexia nervosa, bulimia nervosa, and binge eating disorder. The current standard of care involves psychotherapy, pharmacotherapy, and the management of comorbid conditions, with nutritional rehabilitation reserved for severe cases of anorexia nervosa. Unfortunately, many patients often fail to respond, leaving a concerning treatment gap between the current and requisite treatments for eating disorders. To better understand the neurobiology underlying these eating disorders, investigations have been undertaken to characterize the activity of various neural networks, primarily those activated during tasks of executive inhibition, reward processing, and self-reference. Various neuromodulatory techniques have been proposed to stimulate these networks with the goal of improving patients' BMI and mental health. The aim of this review is to compile a comprehensive summarization of the current literature regarding the underlying neural connectivity of anorexia nervosa, bulimia nervosa, and binge eating disorder as well as the numerous neuromodulatory modalities that have been investigated. Importantly, we aimed to summarize the most significant clinical trials to date as well as to provide an updated assessment of the role of deep brain stimulation, summarizing numerous recently published clinical studies that have greatly contributed to the literature. In this review, we found therapeutic evidence for transcranial magnetic stimulation and transcranial direct current stimulation in treating individuals suffering from anorexia nervosa, bulimia nervosa, and binge eating disorder. We also found significant evidence for the role of deep brain stimulation, particularly as an escalatory therapy option for the those who failed standard therapy. Finally, we hope to provide promising directions for future clinical investigations.

A compact and cost-effective laser-powered speckle visibility spectroscopy (SVS) device for measuring cerebral blood flow.

In the realm of cerebrovascular monitoring, primary metrics typically include blood pressure, which influences cerebral blood flow (CBF) and is contingent upon vessel radius. Measuring CBF non-invasively poses a persistent challenge, primarily attributed to the difficulty of accessing and obtaining signal from the brain. This study aims to introduce a compact speckle visibility spectroscopy (SVS) device designed for non-invasive CBF measurements, offering cost-effectiveness and scalability while tracking CBF with remarkable sensitivity and temporal resolution. The wearable hardware has a modular design approach consisting solely of a laser diode as the source and a meticulously selected board camera as the detector. They both can be easily placed on a subject's head to measure CBF with no additional optical elements. The SVS device can achieve a sampling rate of 80 Hz with minimal susceptibility to external disturbances. The device also achieves better SNR compared with traditional fiber-based SVS devices, capturing about 70 times more signal and showing superior stability and reproducibility. It is designed to be paired and distributed in multiple configurations around the head, and measure signals that exceed the quality of prior optical CBF measurement techniques. Given its cost-effectiveness, scalability, and simplicity, this laser-centric tool offers significant potential in advancing non-invasive cerebral monitoring technologies.

Developing a hippocampal neural prosthetic to facilitate human memory encoding and recall of stimulus features and categories.

Objective: Here, we demonstrate the first successful use of static neural stimulation patterns for specific information content. These static patterns were derived by a model that was applied to a subject's own hippocampal spatiotemporal neural codes for memory. Approach: We constructed a new model of processes by which the hippocampus encodes specific memory items via spatiotemporal firing of neural ensembles that underlie the successful encoding of targeted content into short-term memory. A memory decoding model (MDM) of hippocampal CA3 and CA1 neural firing was computed which derives a stimulation pattern for CA1 and CA3 neurons to be applied during the encoding (sample) phase of a delayed match-to-sample (DMS) human short-term memory task. Main results: MDM electrical stimulation delivered to the CA1 and CA3 locations in the hippocampus during the sample phase of DMS trials facilitated memory of images from the DMS task during a delayed recognition (DR) task that also included control images that were not from the DMS task. Across all subjects, the stimulated trials exhibited significant changes in performance in 22.4% of patient and category combinations. Changes in performance were a combination of both increased memory performance and decreased memory performance, with increases in performance occurring at almost 2 to 1 relative to decreases in performance. Across patients with impaired memory that received bilateral stimulation, significant changes in over 37.9% of patient and category combinations was seen with the changes in memory performance show a ratio of increased to decreased performance of over 4 to 1. Modification of memory performance was dependent on whether memory function was intact or impaired, and if stimulation was applied bilaterally or unilaterally, with nearly all increase in performance seen in subjects with impaired memory receiving bilateral stimulation. Significance: These results demonstrate that memory encoding in patients with impaired memory function can be facilitated for specific memory content, which offers a stimulation method for a future implantable neural prosthetic to improve human memory.

Using the Theory of Planned Behavior to determine COVID-19 vaccination intentions and behavior among international and domestic college students in the United States.

Vaccination is the most effective strategy for preventing infectious diseases such as COVID-19. College students are important targets for COVID-19 vaccines given this population's lower intentions to be vaccinated; however, limited research has focused on international college students' vaccination status. This study explored how psychosocial factors from the Theory of Planned Behavior (TPB; attitudes, perceived behavioral control, subjective norms, and behavioral intentions) related to students' receipt of the full course of COVID-19 vaccines and their plans to receive a booster. Students were recruited via Amazon mTurk and the Office of the Registrar at a U.S. state university. We used binary logistic regression to examine associations between students' psychosocial factors and full COVID-19 vaccination status. Hierarchical multiple regression was employed to evaluate relationships between these factors and students' intentions to receive a booster. The majority of students in our sample (81% of international students and 55% of domestic students) received the complete vaccination series. Attitudes were significantly associated with all students' full vaccination status, while perceived behavioral control was significantly associated with domestic students' status. Students' intentions to receive COVID-19 vaccines were significantly correlated with their intentions to receive a booster, with international students scoring higher on booster intentions. Among the combined college student population, attitudes, intentions to receive COVID-19 vaccines, and subjective norms were significantly related to students' intentions to receive a booster. Findings support the TPB's potential utility in evidence-based interventions to enhance college students' COVID-19 vaccination rates. Implications for stakeholders and future research directions are discussed.

Stereotactic Frame-Based Electrode Insertion: The Accuracy of Increasingly Oblique Insertion Angles.

AIM: To investigate the relationship between planned drill approach angle and angular deviation of the stereotactically placed intracranial electrode tips. MATERIAL AND METHODS: Stereotactic electrode implantation was performed in 13 patients with drug resistant epilepsy. A total of 136 electrodes were included in our analysis. Stereotactic targets were planned on pre-operative magnetic resonance imaging (MRI) scans and implantation was carried out using a Cosman-Roberts-Wells stereotactic frame with the Ad-Tech drill guide and electrodes. Post implant electrode angles in the axial, coronal, and sagittal planes were determined from post-operative computerized tomography (CT) scans and compared with planned angles using Bland-Altman plots and linear regression. RESULTS: Qualitative assessment of correlation plots between planned and actual angles demonstrated a linear relationship for axial, coronal, and sagittal planes, with no overt angular deflection for any magnitude of the planned angle. CONCLUSION: The accuracy of CRW frame-based electrode placement using the Ad-Tech drill guide and electrodes is not significantly affected by the magnitude of the planning angle. Based on our results, oblique electrode insertion is a safe and accurate procedure.

Primary somatosensory cortex organization for engineering artificial somatosensation.

Somatosensory deficits from stroke, spinal cord injury, or other neurologic damage can lead to a significant degree of functional impairment. The primary (SI) and secondary (SII) somatosensory cortices encode information in a medial to lateral organization. SI is generally organized topographically, with more discrete cortical representations of specific body regions. SII regions corresponding to anatomical areas are less discrete and may represent a more functional rather than topographic organization. Human somatosensory research continues to map cortical areas of sensory processing with efforts primarily focused on hand and upper extremity information in SI. However, research into SII and other body regions is lacking. In this review, we synthesize the current state of knowledge regarding the cortical organization of human somatosensation and discuss potential applications for brain computer interface. In addition to accurate individualized mapping of cortical somatosensation, further research is required to uncover the neurophysiological mechanisms of how somatosensory information is encoded in the cortex.

The flaw of averages: Bayes factors as posterior means of the likelihood ratio.

As an alternative to the Frequentist p-value, the Bayes factor (or ratio of marginal likelihoods) has been regarded as one of the primary tools for Bayesian hypothesis testing. In recent years, several researchers have begun to re-analyze results from prominent medical journals, as well as from trials for FDA-approved drugs, to show that Bayes factors often give divergent conclusions from those of p-values. In this paper, we investigate the claim that Bayes factors are straightforward to interpret as directly quantifying the relative strength of evidence. In particular, we show that for nested hypotheses with consistent priors, the Bayes factor for the null over the alternative hypothesis is the posterior mean of the likelihood ratio. By re-analyzing 39 results previously published in the New England Journal of Medicine, we demonstrate how the posterior distribution of the likelihood ratio can be computed and visualized, providing useful information beyond the posterior mean alone.

Epistemic uncertainty in Bayesian predictive probabilities.

Bayesian predictive probabilities have become a ubiquitous tool for design and monitoring of clinical trials. The typical procedure is to average predictive probabilities over the prior or posterior distributions. In this paper, we highlight the limitations of relying solely on averaging, and propose the reporting of intervals or quantiles for the predictive probabilities. These intervals formalize the intuition that uncertainty decreases with more information. We present four different applications (Phase 1 dose escalation, early stopping for futility, sample size re-estimation, and assurance/probability of success) to demonstrate the practicality and generality of the proposed approach.

Civilian Ballistic Arthrotomies: Infection Rates and Operative Versus Nonoperative Management.

OBJECTIVES: The purpose of this study was to determine whether a significant difference existed in the rate of infection after ballistic traumatic arthrotomy managed operatively compared with those managed without surgery. DESIGN: Retrospective cohort study. SETTING: Academic Level I Trauma Center. PATIENT SELECTION CRITERIA: Patients with ballistic traumatic arthrotomies of the shoulder, elbow, wrist, hip, knee, or ankle who received operative or nonoperative management. OUTCOME MEASURES AND COMPARISONS: The rates of infection and septic arthritis in those who received operative or nonoperative management. RESULTS: One hundred ninety-five patients were studied. Eighty patients were treated nonoperatively (Non-Op group), 16 patients were treated with formal irrigation and debridement in the operating room (I&D group), and 99 patients were treated with formal I&D and open reduction and internal fixation (ORIF) (I&D + ORIF group). Patients in all 3 groups received local wound care and systemic antibiotics. No patients in the Non-Op or I&D group developed an infection. Six patients in the I&D + ORIF group developed extra-articular postoperative infections requiring additional interventions. CONCLUSIONS: The infection rate in the I&D + ORIF group was consistent with the infection rates reported in orthopaedic literature after fixation alone. In addition, none of the infections were cases of septic arthritis. This suggests that traumatic arthrotomy does not increase the risk for infection beyond what is expected after fixation alone. Importantly, the Non-Op group represented a series of 80 patients who were treated nonoperatively without developing an infection, indicating that I&D may not be necessary to prevent infection after ballistic arthrotomy. The results suggest that septic arthritis after civilian ballistic arthrotomy is a rare complication regardless of the choice of treatment. LEVEL OF EVIDENCE: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

Neurotropism of SARS-CoV-2: A Pathological Examination of Neurosurgical Specimens.

BACKGROUND AND OBJECTIVES: Neurological manifestations may occur in more than 80% of patients hospitalized with COVID-19 infection, including severe disruptions of the central nervous system (CNS), such as strokes, encephalitis, or seizures. Although the primary pathophysiological mechanism for the effects of COVID-19 in CNS remains unknown, evidence exists for both direct injury from neuroinvasion and indirect effects from disruptions in systemic inflammatory and coagulation pathways. In this study, we analyzed CNS tissue from living patients to better understand these processes. METHODS: With institutional review board approval and patient consent, samples that would be otherwise discarded from patients with active or recent (within 6 days of surgery) COVID-19 infection undergoing neurosurgical intervention were collected and tested for the presence of SARS-CoV-2 using immunohistochemistry, in situ hybridization, electron microscopy, and reverse transcription polymerase chain reaction. RESULTS: Five patients with perioperative mild-to-moderate COVID-19 infection met inclusion criteria (2 male, 3 female; mean age 38.8 ± 13.5 years). Neurosurgical diagnoses included a glioblastoma, a ruptured arteriovenous malformation, a ruptured posterior inferior cerebellar artery aneurysm, a middle cerebral artery occlusion, and a hemorrhagic pontine cavernous malformation. Samples analyzed included the frontal lobe cortex, olfactory nerve, arteriovenous malformation/temporal lobe parenchyma, middle cerebral artery, cerebellum, and cavernous malformation/brainstem parenchyma. Testing for the presence of SARS-CoV-2 was negative in all samples. CONCLUSION: The CNS is likely not a significant viral reservoir during mild-to-moderate COVID-19 infection, although direct neuroinvasion is not definitively excluded. Additional testing to help elucidate the relative contributions of direct and indirect pathways for CNS injury from COVID is warranted.

Decoding motor plans using a closed-loop ultrasonic brain-machine interface.

Brain-machine interfaces (BMIs) enable people living with chronic paralysis to control computers, robots and more with nothing but thought. Existing BMIs have trade-offs across invasiveness, performance, spatial coverage and spatiotemporal resolution. Functional ultrasound (fUS) neuroimaging is an emerging technology that balances these attributes and may complement existing BMI recording technologies. In this study, we use fUS to demonstrate a successful implementation of a closed-loop ultrasonic BMI. We streamed fUS data from the posterior parietal cortex of two rhesus macaque monkeys while they performed eye and hand movements. After training, the monkeys controlled up to eight movement directions using the BMI. We also developed a method for pretraining the BMI using data from previous sessions. This enabled immediate control on subsequent days, even those that occurred months apart, without requiring extensive recalibration. These findings establish the feasibility of ultrasonic BMIs, paving the way for a new class of less-invasive (epidural) interfaces that generalize across extended time periods and promise to restore function to people with neurological impairments.

A Pilot Study of the Effect of Transcutaneous Spinal Cord Stimulation on Micturition-Related Brain Activity and Lower Urinary Tract Symptoms After Stroke.

PURPOSE: Transcutaneous spinal cord stimulation (TSCS) is a novel neuromodulation modality developed to promote functional restoration in patients with neurological injury or disease. Previous pilot data suggest that lower urinary tract dysfunction (LUTD) due to stroke may be partially alleviated by TSCS. In this study, we examine the mechanism of this effect by evaluating bladder-related brain activity in patients before and after TSCS therapy and comparing it to healthy volunteers. MATERIALS AND METHODS: Patients who developed storage LUTD after a stroke and healthy volunteers without LUTD were recruited. Patients and healthy volunteers underwent simultaneous urodynamics and functional MRI. Patients then completed 24 biweekly sessions of TSCS and underwent another simultaneous urodynamics-functional MRI study. Clinical outcomes were assessed using validated questionnaires and voiding diary. RESULTS: Fifteen patients and 16 healthy volunteers completed the study. Following TSCS, patients exhibited increased blood-oxygen-level-dependent activity in areas including periaqueductal grey, the insula, the lateral prefrontal cortex, and motor cortex. Prior to TSCS therapy, healthy controls exhibited higher blood-oxygen-level-dependent activity in 17 regions, including multiple regions in the prefrontal cortex and basal ganglia. These differences were attenuated after TSCS with no frontal brain differences remaining between healthy volunteers and stroke participants who completed therapy. Neuroimaging changes were complemented by clinically significant improvements in questionnaire scores and voiding diary parameters. CONCLUSIONS: TSCS therapy modulated bladder-related brain activity, reducing differences between healthy volunteers and stroke patients with LUTD. These changes, alongside improved clinical outcomes, suggest TSCS as a promising approach for LUTD management.

Zinc deficiency impairs axonal regeneration and functional recovery after spinal cord injury by modulating macrophage polarization via NF-κB pathway.

Background: Spinal cord injury (SCI) is a devastating disease that results in permanent paralysis. Currently, there is no effective treatment for SCI, and it is important to identify factors that can provide therapeutic intervention during the course of the disease. Zinc, an essential trace element, has attracted attention as a regulator of inflammatory responses. In this study, we investigated the effect of zinc status on the SCI pathology and whether or not zinc could be a potential therapeutic target. Methods: We created experimental mouse models with three different serum zinc concentration by changing the zinc content of the diet. After inducing contusion injury to the spinal cord of three mouse models, we assessed inflammation, apoptosis, demyelination, axonal regeneration, and the number of nuclear translocations of NF-κB in macrophages by using qPCR and immunostaining. In addition, macrophages in the injured spinal cord of these mouse models were isolated by flow cytometry, and their intracellular zinc concentration level and gene expression were examined. Functional recovery was assessed using the open field motor score, a foot print analysis, and a grid walk test. Statistical analysis was performed using Wilcoxon rank-sum test and ANOVA with the Tukey-Kramer test. Results: In macrophages after SCI, zinc deficiency promoted nuclear translocation of NF-κB, polarization to pro-inflammatory like phenotype and expression of pro-inflammatory cytokines. The inflammatory response exacerbated by zinc deficiency led to worsening motor function by inducing more apoptosis of oligodendrocytes and demyelination and inhibiting axonal regeneration in the lesion site compared to the normal zinc condition. Furthermore, zinc supplementation after SCI attenuated these zinc-deficiency-induced series of responses and improved motor function. Conclusion: We demonstrated that zinc affected axonal regeneration and motor functional recovery after SCI by negatively regulating NF-κB activity and the subsequent inflammatory response in macrophages. Our findings suggest that zinc supplementation after SCI may be a novel therapeutic strategy for SCI.

Deep pathomics: A new image-based tool for predicting response to treatment in stage III non-small cell lung cancer.

Despite the advantages offered by personalized treatments, there is presently no way to predict response to chemoradiotherapy in patients with non-small cell lung cancer (NSCLC). In this exploratory study, we investigated the application of deep learning techniques to histological tissue slides (deep pathomics), with the aim of predicting the response to therapy in stage III NSCLC. We evaluated 35 digitalized tissue slides (biopsies or surgical specimens) obtained from patients with stage IIIA or IIIB NSCLC. Patients were classified as responders (12/35, 34.7%) or non-responders (23/35, 65.7%) based on the target volume reduction shown on weekly CT scans performed during chemoradiation treatment. Digital tissue slides were tested by five pre-trained convolutional neural networks (CNNs)-AlexNet, VGG, MobileNet, GoogLeNet, and ResNet-using a leave-two patient-out cross validation approach, and we evaluated the networks' performances. GoogLeNet was globally found to be the best CNN, correctly classifying 8/12 responders and 10/11 non-responders. Moreover, Deep-Pathomics was found to be highly specific (TNr: 90.1) and quite sensitive (TPr: 0.75). Our data showed that AI could surpass the capabilities of all presently available diagnostic systems, supplying additional information beyond that currently obtainable in clinical practice. The ability to predict a patient's response to treatment could guide the development of new and more effective therapeutic AI-based approaches and could therefore be considered an effective and innovative step forward in personalised medicine.

Power Budget of a Skull Unit in a Fully-Implantable Brain-Computer Interface: Bio-Heat Model.

The aim of this study is to estimate the maximum power consumption that guarantees the thermal safety of a skull unit (SU). The SU is part of a fully-implantable bi-directional brain computer-interface (BD-BCI) system that aims to restore walking and leg sensation to those with spinal cord injury (SCI). To estimate the SU power budget, we created a bio-heat model using the finite element method (FEM) implemented in COMSOL. To ensure that our predictions were robust against the natural variation of the model's parameters, we also performed a sensitivity analysis. Based on our simulations, we estimated that the SU can nominally consume up to 70 mW of power without raising the surrounding tissues' temperature above the thermal safety threshold of 1°C. When considering the natural variation of the model's parameters, we estimated that the power budget could range between 47 and 81 mW. This power budget should be sufficient to power the basic operations of the SU, including amplification, serialization and A/D conversion of the neural signals, as well as control of cortical stimulation. Determining the power budget is an important specification for the design of the SU and, in turn, the design of a fully-implantable BD-BCI system.

Selective prediction for extracting unstructured clinical data.

OBJECTIVE: While there are currently approaches to handle unstructured clinical data, such as manual abstraction and structured proxy variables, these methods may be time-consuming, not scalable, and imprecise. This article aims to determine whether selective prediction, which gives a model the option to abstain from generating a prediction, can improve the accuracy and efficiency of unstructured clinical data abstraction. MATERIALS AND METHODS: We trained selective classifiers (logistic regression, random forest, support vector machine) to extract 5 variables from clinical notes: depression (n = 1563), glioblastoma (GBM, n = 659), rectal adenocarcinoma (DRA, n = 601), and abdominoperineal resection (APR, n = 601) and low anterior resection (LAR, n = 601) of adenocarcinoma. We varied the cost of false positives (FP), false negatives (FN), and abstained notes and measured total misclassification cost. RESULTS: The depression selective classifiers abstained on anywhere from 0% to 97% of notes, and the change in total misclassification cost ranged from -58% to 9%. Selective classifiers abstained on 5%-43% of notes across the GBM and colorectal cancer models. The GBM selective classifier abstained on 43% of notes, which led to improvements in sensitivity (0.94 to 0.96), specificity (0.79 to 0.96), PPV (0.89 to 0.98), and NPV (0.88 to 0.91) when compared to a non-selective classifier and when compared to structured proxy variables. DISCUSSION: We showed that selective classifiers outperformed both non-selective classifiers and structured proxy variables for extracting data from unstructured clinical notes. CONCLUSION: Selective prediction should be considered when abstaining is preferable to making an incorrect prediction.