Measuring instability in chronic human intracortical neural recordings towards stable, long-term brain-computer interfaces

Intracortical brain-computer interfaces (iBCIs) enable people with tetraplegia to gain intuitive cursor control from movement intentions. To translate to practical use, iBCIs should provide reliable performance for extended periods of time. However, performance begins to degrade as the relationship between kinematic intention and recorded neural activity shifts compared to when the decoder was initially trained. In addition to developing decoders to better handle long-term instability, identifying when to recalibrate will also optimize performance. We propose a method to measure instability in neural data without needing to label user intentions. Longitudinal data were analyzed from two BrainGate2 participants with tetraplegia as they used fixed decoders to control a computer cursor spanning 142 days and 28 days, respectively. We demonstrate a measure of instability that correlates with changes in closed-loop cursor performance solely based on the recorded neural activity (Pearson r = 0.93 and 0.72, respectively). This result suggests a strategy to infer online iBCI performance from neural data alone and to determine when recalibration should take place for practical long-term use.

Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication

Intracortical brain-computer interfaces (iBCIs) have shown promise for restoring rapid communication to people with neurological disorders such as amyotrophic lateral sclerosis (ALS). However, to maintain high performance over time, iBCIs typically need frequent recalibration to combat changes in the neural recordings that accrue over days. This requires iBCI users to stop using the iBCI and engage in supervised data collection, making the iBCI system hard to use. In this paper, we propose a method that enables self-recalibration of communication iBCIs without interrupting the user. Our method leverages large language models (LMs) to automatically correct errors in iBCI outputs. The self-recalibration process uses these corrected outputs ("pseudo-labels") to continually update the iBCI decoder online. Over a period of more than one year (403 days), we evaluated our Continual Online Recalibration with Pseudo-labels (CORP) framework with one clinical trial participant. CORP achieved a stable decoding accuracy of 93.84% in an online handwriting iBCI task, significantly outperforming other baseline methods. Notably, this is the longest-running iBCI stability demonstration involving a human participant. Our results provide the first evidence for long-term stabilization of a plug-and-play, high-performance communication iBCI, addressing a major barrier for the clinical translation of iBCIs.

Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication

Intracortical brain-computer interfaces (iBCIs) have shown promise for restoring rapid communication to people with neurological disorders such as amyotrophic lateral sclerosis (ALS). However, to maintain high performance over time, iBCIs typically need frequent recalibration to combat changes in the neural recordings that accrue over days. This requires iBCI users to stop using the iBCI and engage in supervised data collection, making the iBCI system hard to use. In this paper, we propose a method that enables self-recalibration of communication iBCIs without interrupting the user. Our method leverages large language models (LMs) to automatically correct errors in iBCI outputs. The self-recalibration process uses these corrected outputs ("pseudo-labels") to continually update the iBCI decoder online. Over a period of more than one year (403 days), we evaluated our Continual Online Recalibration with Pseudo-labels (CORP) framework with one clinical trial participant. CORP achieved a stable decoding accuracy of 93.84% in an online handwriting iBCI task, significantly outperforming other baseline methods. Notably, this is the longest-running iBCI stability demonstration involving a human participant. Our results provide the first evidence for long-term stabilization of a plug-and-play, high-performance communication iBCI, addressing a major barrier for the clinical translation of iBCIs.

A high-performance speech neuroprosthesis

Speech brain-computer interfaces (BCIs) have the potential to restore rapid communication to people with paralysis by decoding neural activity evoked by attempted speech into text1,2 or sound3,4. Early demonstrations, although promising, have not yet achieved accuracies sufficiently high for communication of unconstrained sentences from a large vocabulary1-7. Here we demonstrate a speech-to-text BCI that records spiking activity from intracortical microelectrode arrays. Enabled by these high-resolution recordings, our study participant-who can no longer speak intelligibly owing to amyotrophic lateral sclerosis-achieved a 9.1% word error rate on a 50-word vocabulary (2.7 times fewer errors than the previous state-of-the-art speech BCI2) and a 23.8% word error rate on a 125,000-word vocabulary (the first successful demonstration, to our knowledge, of large-vocabulary decoding). Our participant's attempted speech was decoded  at 62 words per minute, which is 3.4 times as fast as the previous record8 and begins to approach the speed of natural conversation (160 words per minute9). Finally, we highlight two aspects of the neural code for speech that are encouraging for speech BCIs: spatially intermixed tuning to speech articulators that makes accurate decoding possible from only a small region of cortex, and a detailed articulatory representation of phonemes that persists years after paralysis. These results show a feasible path forward for restoring rapid communication to people with paralysis who can no longer speak.

Long-term unsupervised recalibration of cursor BCIs

Intracortical brain-computer interfaces (iBCIs) require frequent recalibration to maintain robust performance due to changes in neural activity that accumulate over time. Compensating for this nonstationarity would enable consistently high performance without the need for supervised recalibration periods, where users cannot engage in free use of their device. Here we introduce a hidden Markov model (HMM) to infer what targets users are moving toward during iBCI use. We then retrain the system using these inferred targets, enabling unsupervised adaptation to changing neural activity. Our approach outperforms the state of the art in large-scale, closed-loop simulations over two months and in closed-loop with a human iBCI user over one month. Leveraging an offline dataset spanning five years of iBCI recordings, we further show how recently proposed data distribution-matching approaches to recalibration fail over long time scales; only target-inference methods appear capable of enabling long-term unsupervised recalibration. Our results demonstrate how task structure can be used to bootstrap a noisy decoder into a highly-performant one, thereby overcoming one of the major barriers to clinically translating BCIs.

Decoding spoken English from intracortical electrode arrays in dorsal precentral gyrus

OBJECTIVE

To evaluate the potential of intracortical electrode array signals for brain-computer interfaces (BCIs) to restore lost speech, we measured the performance of decoders trained to discriminate a comprehensive basis set of 39 English phonemes and to synthesize speech sounds via a neural pattern matching method. We decoded neural correlates of spoken-out-loud words in the 'hand knob' area of precentral gyrus, a step toward the eventual goal of decoding attempted speech from ventral speech areas in patients who are unable to speak.

APPROACH

Neural and audio data were recorded while two BrainGate2 pilot clinical trial participants, each with two chronically-implanted 96-electrode arrays, spoke 420 different words that broadly sampled English phonemes. Phoneme onsets were identified from audio recordings, and their identities were then classified from neural features consisting of each electrode's binned action potential counts or high-frequency local field potential power. Speech synthesis was performed using the 'Brain-to-Speech' pattern matching method. We also examined two potential confounds specific to decoding overt speech: acoustic contamination of neural signals and systematic differences in labeling different phonemes' onset times.

MAIN RESULTS

A linear decoder achieved up to 29.3% classification accuracy (chance = 6%) across 39 phonemes, while an RNN classifier achieved 33.9% accuracy. Parameter sweeps indicated that performance did not saturate when adding more electrodes or more training data, and that accuracy improved when utilizing time-varying structure in the data. Microphonic contamination and phoneme onset differences modestly increased decoding accuracy, but could be mitigated by acoustic artifact subtraction and using a neural speech onset marker, respectively. Speech synthesis achieved r = 0.523 correlation between true and reconstructed audio.

SIGNIFICANCE

The ability to decode speech using intracortical electrode array signals from a nontraditional speech area suggests that placing electrode arrays in ventral speech areas is a promising direction for speech BCIs.

Neural ensemble dynamics in dorsal motor cortex during speech in people with paralysis

Speaking is a sensorimotor behavior whose neural basis is difficult to study with single neuron resolution due to the scarcity of human intracortical measurements. We used electrode arrays to record from the motor cortex 'hand knob' in two people with tetraplegia, an area not previously implicated in speech. Neurons modulated during speaking and during non-speaking movements of the tongue, lips, and jaw. This challenges whether the conventional model of a 'motor homunculus' division by major body regions extends to the single-neuron scale. Spoken words and syllables could be decoded from single trials, demonstrating the potential of intracortical recordings for brain-computer interfaces to restore speech. Two neural population dynamics features previously reported for arm movements were also present during speaking: a component that was mostly invariant across initiating different words, followed by rotatory dynamics during speaking. This suggests that common neural dynamical motifs may underlie movement of arm and speech articulators.

Skeletal forelimb measurements and hoof spread in relation to asymmetry in the bilateral forelimb of horses

REASONS FOR PERFORMING STUDY

Research has highlighted a high frequency of skeletal asymmetries in horses. In addition, research into hoof asymmetries has shown that within a bilateral pair, the hoof with the smaller angle is often subjected to greater loading. There has been limited attention paid to understanding compensatory mechanisms for skeletal asymmetries in the horse; the dynamic structure of the hoof could potentially be acting in a compensatory capacity.

OBJECTIVES

To investigate the relationship between morphometry of forelimb segments and hoof spread and their incidence of asymmetry.

METHODS

Ten bilateral measurements of the hoof and forelimb were taken from 34 leisure horses. The relationship between hoof spread and forelimb segment measurements were analysed using a generalised linear model (GLM).

RESULTS

In relation to left hoof spread, the GLM identified significant negative relationships with left side measurements (third metacarpal length, elbow height), and significant positive relationships with right side measurements (fetlock height, third metacarpal length, elbow height). In relation to right hoof spread, the GLM identified significant negative relationship with left elbow height, and significant positive relationships with right side measurements (fetlock height, point of shoulder). The difference between the number of horses larger to the left or to the right was found to be significant for point of shoulder height (chi2 = 4.8, P < 0.05), and highly significant for heel height (chi2 = 953, P < 0.01) and the third metacarpal length (chi2 = 7.26, P < 0.01).

CONCLUSIONS AND CLINICAL RELEVANCE

The study demonstrated considerable asymmetry in left-right morphometry of the equine limb. The fact that measurements of hoof spread were significantly associated with limb segment measurements could possibly indicate that an interaction exists. Any asymmetry in hoof spread measurements may suggest unequal loading of the limbs, which in turn may contribute to injuries and reduced performance.

T cell lymphoma in the lumbar spine of a domestic ferret (Mustela putorius furo )

A 22-month-old castrated male ferret developed acute pelvic limb paresis. Radiographs and computed tomography revealed a soft tissue mass with associated bony lysis of L5, and ultrasound-guided fine needle aspirates suggested that it was a lymphoma. Treatment with prednisone at immunosuppressive doses did not produce any detectable improvement in the ferret's clinical signs and it became moribund less than two weeks after they developed. A postmortem biopsy confirmed the presence of a lymphoma which had invaded the vertebral bone. No viruses were detected by cell culture, or electron microscopy.

Secondary nutritional hyperparathyroidism associated with vitamin D deficiency in two domestic skunks (Mephitis mephitis)

Two privately owned domestic skunks (Mephitis mephitis) developed clinical signs of hyperparathyroidism. Survey radiographs, complete blood counts and biochemical profiles, including the concentrations of ionised calcium, parathyroid hormone and 25-(OH)-vitamin D, established that they were deficient in vitamin D and had secondary nutritional hyperparathyroidism. They both responded to treatment, as well as to changes in their diet, and levels of exercise and exposure to sunlight.

Validation of a novel high-sensitivity radioimmunoassay procedure for measurement of total thyroxine concentration in psittacine birds and snakes

OBJECTIVE

To validate a novel high-sensitivity radioimmunoassay (RIA) procedure developed to accurately measure the relatively low serum total thyroxine (T4) concentrations of birds and reptiles and to establish initial reference ranges forT4 concentration in selected species of psittacine birds and snakes.

ANIMALS

56 healthy nonmolting adult psittacine birds representing 6 species and 42 captive snakes representing 4 species.

PROCEDURE

A solid-phase RIA designed to measure free T4 concentrations in dialysates of human serum samples was used without dialysis to evaluate total T4 concentration in treated samples obtained from birds and reptiles. Serum T4 binding components were removed to allow assay of undialyzed samples. Assay validation was assessed by determining recovery of expected amounts of T4 in treated samples that were serially diluted or to which T4 was added. Intra- and interassay coefficient of variation (CV) was determined.

RESULTS

Mean recovery of T4 added at 4 concentrations ranged from 84.9 to 115.0% and 95.8 to 119.4% in snakes and birds, respectively. Intra- and interassay CV was 3.8 and 11.3%, respectively. Serum total T4 concentrations for 5 species of birds ranged from 2.02 to 768 nmol/L but ranged from 3.17 to 142 nmol/L for blue-fronted Amazon parrots; concentrations ranged from 0.21 to 6.06 nmol/L for the 4 species of snakes.

CONCLUSIONS AND CLINICAL RELEVANCE

This new RIA method provides a commercially available, accurate, and sensitive method for measurement of the relatively low serum T4 concentrations of birds and snakes. Initial ranges for the species evaluated were established.

MR evaluation of early myelination patterns in normal and developmentally delayed infants

This study demonstrates the ability of MR imaging to show progression of myelination in 64 infants and young children (ages 4 days to 36 months). T2-weighted spin-echo pulse sequences, frequently used for routine screening of intracranial disease, were used. Gray-white matter differentiation was seen in all patients, and changes occurring with age were documented. Three distinct patterns were seen, and age ranges were established for each pattern in developmentally normal children: (1) infantile (birth-6 months); (2) isointense (8-12 months); and (3) early adult (10 months onward). There was a statistically significant difference between the age ranges of the normal and developmentally delayed children showing all three patterns. These data should be helpful for identifying and following sequentially both infants with clinically suspected developmental delay and those with dysmyelinating or demyelinating disease.

MR imaging of the nasopharynx and floor of the middle cranial fossa. Part I. Normal anatomy

The normal anatomy of the nasopharynx and floor of the middle cranial fossa was analyzed with magnetic resonance (MR) imaging. MR images from five healthy volunteers were correlated with whole-organ cryomicrotome sections from three cadavers. Anatomic connections exist between the paranasopharyngeal spaces and the surface structures of the skull base. These anatomic connections include the intimate relationship between the eustachian tube and the pharyngobasilar fascia, the attachment of the muscles of mastication and deglutition to the skull base, and vascular and nervous structures in the foramina. The inherent contrast between the soft tissues of the nasopharynx and related structures and the bone of the floor of the middle cranial fossa allowed excellent visualization of these anatomic connections.

MR imaging of the nasopharynx and floor of the middle cranial fossa. Part II. Malignant tumors

The intracranial extension of tumors of the nasopharynx and related spaces presents a difficult imaging problem. Unlike computed tomography (CT) scans, magnetic resonance (MR) images are not limited by beam-hardening artifacts from bone or dental amalgam. Forty-six patients with malignant tumors of the nasopharynx and related spaces affecting the skull base underwent MR imaging. MR images were obtained with a 0.3-T permanent-magnet imaging system in axial, sagittal, and coronal planes. MR findings were compared with clinical records, plain radiographs, CT scans, and pathologic correlates when available. MR imaging could demonstrate neoplastic invasion of the bone of the floor of the middle cranial fossa and the vital soft-tissue structures related to it as well as or better than CT. Tumor extension was viewed directly as a continuous mass or indirectly by marrow replacement or displacement of normal structures. Specific anatomic routes through which tumors extend from the nasopharynx to the middle cranial fossa were inferred from MR findings.

Partially thrombosed giant aneurysm simulating an arteriovenous malformation on MR imaging

A case of a partially thrombosed giant intracranial aneurysm with a magnetic resonance appearance simulating an arteriovenous malformation is presented. Causes for the similarity in signal pattern are discussed.

MR imaging of the lumbar spine: anatomic correlations and the effects of technical variations

A correlative anatomic study, a retrospective review of MR images performed in 35 patients, and a series of tests of the effectiveness of various MR scanning techniques were performed in order to improve comprehension of lumbar spine anatomy depicted on MR images, and thereby facilitate development of an optimal scanning protocol. Correlation of MR images with cryomicrotomed cadaver specimens enhanced understanding of the MR depiction of the intervertebral disks, ligamentum flavum, nerve roots, epidural fat, and epidural veins. Experiments were performed to assess the efficacy of a surface coil applied to the back, a solenoidal surface coil, a standard body coil, and an abdominal compression device in optimizing image quality. Experiments were also performed to determine the effect of alterations in the pulse sequence and variations of the phase-encoding axis. Based on these results, a protocol is proposed for routine imaging of the lumbar spine that yields high-resolution sagittal and oblique images and that does not require a surface coil. The recommended protocol employs heavily T1-weighted images with phase encoding along the z axis for sagittal images and along the x axis for axial images. This protocol yields multiple sagittal and oblique axial images through each of the lumbar disks, a larger field of view than obtained with surface coils, and a reduction of total imaging time to as little as 10 min.

Head and neck MR imaging in the pediatric patient

Magnetic resonance (MR) imaging studies of the head and neck (excluding the brain) were obtained in 49 children believed to have lesions of the head and neck. Seven children had normal images; in the remaining 42, lesions were divided into four categories: midline lesions, lesions of symmetric paired structures, facial lesions, and nasopharyngeal and oropharyngeal lesions. All entities were well delineated by MR imaging. The imaging planes and sequences chosen depended on the suspected abnormality. Midline lesions were best imaged in the sagittal plane, lesions of paired structures and the face in the axial or coronal planes, and nasopharyngeal and oropharyngeal lesions in the axial or sagittal planes. Intracranial extension of head and neck neoplasms was best evaluated in the coronal plane. Surface coils provided better resolution and were thus more useful in evaluating small superficial lesions; head or body coils were more useful in defining the extent of large lesions. T2-weighted images provided better differentiation between normal and tumor tissue in patients with head and neck neoplasms.

Magnetic resonance imaging of the brainstem and cranial nerves

Ten normal human volunteers and 44 patients with pathology of the brainstem or cranial nerves were scanned using a. 3 Tesla permanent MR imaging system. MR images were obtained of the cranial nerves and brainstem using various spin-echo pulse sequences and scanning planes. 4 mm thick sections with .75 mm pixels on a 256 display matrix were used whenever possible. The normal MR images were correlated with thin section cryodissection specimens of fresh human cadavers. Brainstem structures including major nuclei and tracts were then identified. The cranial nerves were followed through the subarachnoid cisterns and the base of the skull. Pathological involvement of the brainstem by tumors, infarcts, and demyelinating disease was well shown and correlated with clinical findings. Examples of optic glioma, fifth, eighth, and twelfth nerve schwannomas as well as other cranial nerve pathology were also demonstrated. Magnetic resonance produces excellent images of cranial nerves and brainstem with high contrast resolution. Unlike CT, there is no beam hardening artifact from bone. T1 weighted images maximize brainstem-CSF contrast and are useful for demonstrating the external anatomy of the brainstem and cranial nerves. The T2 weighted images show internal brainstem anatomy, CSF within neural foramina, and highlight many pathological conditions.

Opportunities for basic imaging research - diagnostic imaging research branch/national cancer institute perspective

The Diagnostic Imaging Research Branch of the National Cancer Institute, in collaboration with the Conjoint Committee on Diagnostic Radiology, is developing a long range plan for imaging research. A series of workshops is being held to accomplish this end. A summary of the recommendations from the first workshop are presented.

Cerebral CT findings in drug abuse: clinical and experimental observations

Computed tomographic changes compatible with cerebral atrophy have been observed clinically and in experimental drug abuse animals. Most of the patients were thought to be polydrug users, but amphetamines and alcohol were two of the more commonly misused drugs. In the experimental animals, the effects of amphetamine, secobarbital, and marijuana were studied. Brain damage was present in the animals that received intravenous amphetamine.

Giant intracranial aneurysms simulating brain neoplasms on computed tomography

Giant intracranial aneurysms (GIA) may simulate brain neoplasms on computed tomography (CT) scans. We present three cases in which a GIA was mistakenly diagnosed as a brain neoplasm on the basis of CT findings. The correct diagnosis was made in each case by cerebral angiography. The possibility of GIA must be considered when a nonenhanced CT study shows a well circumscribed mass without edema at the base of the brain and contrast infusion reveals homogeneous or inhomogeneous enhancement.

Computed tomography in intracranial cysticercosis

The computed tomography (CT) scans and other radiographic studies of 23 patients with cerebral cysticercosis were reviewed. Parenchymal calcifications, the most common finding, are somewhat better seen on CT scans than on skull films. Ventricular cysts may be suspected by disproportionate enlargement of the involved ventricle. Cysts in subarachnoid cisterns are best detected by performing both CT and pneumoencephalography. Contrast enhancement was not seen with cisternal or ventricular cysts but may be with parenchymal cysticercus lesions.

Clinical comparison of three contrast agents used in cerebral angiography

A total of 139 cerebral angiograms were performed on patients assigned randomly to three groups. The contrast agents used for these groups included meglumine calcium metrizoate, meglumine diatrizoate and meglumine sodium diatrizoate. Pulse rate and blood pressure readings were taken preceding and following injections and pain and heat sensations were graded. There were no significant differences in incidence of nausea or in changes of blood pressure or pulse rate, or in severity of heat sensations. There was a lower incidence of severe injection pain with meglumine calcium metrizoate than with meglumine sodium diatrizoate.

Contrast enhancement by arterial perfusion during computerized tomography

Contrast enhancement of computerized tomography is usually performed by means of intravenous injection of iodinated contrast materials. The authors have utilized a technique of direct arterial infusion of iodinated contrast material supplying intracranial arteries during computerized axial tomography. A representative case is presented with comparison scans which were obtained without contrast enhancement and with conventional intravenous infusion of contrast. Marked improvement of the enhancement was clearly demonstrated in this case, with abnormalities readily defined that were not seen with the conventional intravenous enhancement procedure.

Catheter technique for encephalography

A catheter technique for encephalography utilizing a polyethylene catheter is described. The advantages are principally that it reduces the possibility of needle displacement during examination and thus the danger of extra-arachnoid injection of air.

Accuracy of scanning, angiography and encephalography in posterior fossa tumors

A material of 120 cases with posterior fossa tumors examined with scanning, vertebral angiography and encephalography was reviewed to determine the value of the information and accuracy these methods provide. Encephalography was the most accurate procedure (98% positive) compared to angiography (81%) and scanning (44%).

Neuroradiology of the sphenoidal region

The neuroradiological techniques employed in diagnosing a series of 54 lesions involving the superior orbital fissure and cavernous sinus (sphenoidal region) were reviewed in order to compare the value of sphenoidal tomography, pneumoencephalography, cerebral arteriography and sphenoidal venography. The patients included in the series exhibited clinical syndromes of sphenoidal region cranial nerve deficits. Confirmation of the sphenoidal region disorder was obtained in all cases. Of the four procedures, sphenoidal venography and basal tomography yielded the highest rate of positive studies.

Partial pneumoencephalography in following-up pituitary tumours

The `limited' pneumoencephalogram has been used with excellent success at UCLA for the continuing follow-up of pituitary tumours. It is most useful in following nonsecretory adenomas since tumour regrowth can occur in the absence of clinical signs and symptoms. Total serial pneumoencephalography has not been accepted previously for follow-up of pituitary tumours since there is a significant morbidity. The `limited' pneumoencephalogram of the diseased area drastically reduces the morbidity of the procedure so that the patients are willing to undergo serial studies on an outpatient basis.