Optimizing warfarin dosing for patients with atrial fibrillation using machine learning

While novel oral anticoagulants are increasingly used to reduce risk of stroke in patients with atrial fibrillation, vitamin K antagonists such as warfarin continue to be used extensively for stroke prevention across the world. While effective in reducing the risk of strokes, the complex pharmacodynamics of warfarin make it difficult to use clinically, with many patients experiencing under- and/or over- anticoagulation. In this study we employed a novel implementation of deep reinforcement learning to provide clinical decision support to optimize time in therapeutic International Normalized Ratio (INR) range. We used a novel semi-Markov decision process formulation of the Batch-Constrained deep Q-learning algorithm to develop a reinforcement learning model to dynamically recommend optimal warfarin dosing to achieve INR of 2.0-3.0 for patients with atrial fibrillation. The model was developed using data from 22,502 patients in the warfarin treated groups of the pivotal randomized clinical trials of edoxaban (ENGAGE AF-TIMI 48), apixaban (ARISTOTLE) and rivaroxaban (ROCKET AF). The model was externally validated on data from 5730 warfarin-treated patients in a fourth trial of dabigatran (RE-LY) using multilevel regression models to estimate the relationship between center-level algorithm consistent dosing, time in therapeutic INR range (TTR), and a composite clinical outcome of stroke, systemic embolism or major hemorrhage. External validation showed a positive association between center-level algorithm-consistent dosing and TTR (R2 = 0.56). Each 10% increase in algorithm-consistent dosing at the center level independently predicted a 6.78% improvement in TTR (95% CI 6.29, 7.28; p < 0.001) and a 11% decrease in the composite clinical outcome (HR 0.89; 95% CI 0.81, 1.00; p = 0.015). These results were comparable to those of a rules-based clinical algorithm used for benchmarking, for which each 10% increase in algorithm-consistent dosing independently predicted a 6.10% increase in TTR (95% CI 5.67, 6.54, p < 0.001) and a 10% decrease in the composite outcome (HR 0.90; 95% CI 0.83, 0.98, p = 0.018). Our findings suggest that a deep reinforcement learning algorithm can optimize time in therapeutic range for patients taking warfarin. A digital clinical decision support system to promote algorithm-consistent warfarin dosing could optimize time in therapeutic range and improve clinical outcomes in atrial fibrillation globally.

Machine Learning Approaches for the Image-Based Identification of Surgical Wound Infections: Scoping Review

BACKGROUND

Surgical site infections (SSIs) occur frequently and impact patients and health care systems. Remote surveillance of surgical wounds is currently limited by the need for manual assessment by clinicians. Machine learning (ML)-based methods have recently been used to address various aspects of the postoperative wound healing process and may be used to improve the scalability and cost-effectiveness of remote surgical wound assessment.

OBJECTIVE

The objective of this review was to provide an overview of the ML methods that have been used to identify surgical wound infections from images.

METHODS

We conducted a scoping review of ML approaches for visual detection of SSIs following the JBI (Joanna Briggs Institute) methodology. Reports of participants in any postoperative context focusing on identification of surgical wound infections were included. Studies that did not address SSI identification, surgical wounds, or did not use image or video data were excluded. We searched MEDLINE, Embase, CINAHL, CENTRAL, Web of Science Core Collection, IEEE Xplore, Compendex, and arXiv for relevant studies in November 2022. The records retrieved were double screened for eligibility. A data extraction tool was used to chart the relevant data, which was described narratively and presented using tables. Employment of TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) guidelines was evaluated and PROBAST (Prediction Model Risk of Bias Assessment Tool) was used to assess risk of bias (RoB).

RESULTS

In total, 10 of the 715 unique records screened met the eligibility criteria. In these studies, the clinical contexts and surgical procedures were diverse. All papers developed diagnostic models, though none performed external validation. Both traditional ML and deep learning methods were used to identify SSIs from mostly color images, and the volume of images used ranged from under 50 to thousands. Further, 10 TRIPOD items were reported in at least 4 studies, though 15 items were reported in fewer than 4 studies. PROBAST assessment led to 9 studies being identified as having an overall high RoB, with 1 study having overall unclear RoB.

CONCLUSIONS

Research on the image-based identification of surgical wound infections using ML remains novel, and there is a need for standardized reporting. Limitations related to variability in image capture, model building, and data sources should be addressed in the future.

A versatile, fast and unbiased method for estimation of gene-by-environment interaction effects on biobank-scale datasets

Identification of gene-by-environment interactions (GxE) is crucial to understand the interplay of environmental effects on complex traits. However, current methods evaluating GxE on biobank-scale datasets have limitations. We introduce MonsterLM, a multiple linear regression method that does not rely on model specification and provides unbiased estimates of variance explained by GxE. We demonstrate robustness of MonsterLM through comprehensive genome-wide simulations using real genetic data from 325,989 individuals. We estimate GxE using waist-to-hip-ratio, smoking, and exercise as the environmental variables on 13 outcomes (N = 297,529-325,989) in the UK Biobank. GxE variance is significant for 8 environment-outcome pairs, ranging from 0.009 - 0.071. The majority of GxE variance involves SNPs without strong marginal or interaction associations. We observe modest improvements in polygenic score prediction when incorporating GxE. Our results imply a significant contribution of GxE to complex trait variance and we show MonsterLM to be well-purposed to handle this with biobank-scale data.

Machine learning for detecting centre-level irregularities in randomized controlled trials: A pilot study

Centralized statistical monitoring is sometimes employed as an alternative to onsite monitoring for randomized control trials. Current central monitoring methods have limitations, in that they are relatively resource intensive and do not necessarily generalize to studies where an irregularity pattern has not been observed before. Machine learning has been effective in detecting irregularities in industries such as finance and manufacturing, but to date none have been applied to clinical trials. We conducted a pilot study for the use of machine learning to identify center-level irregularities in data from multicenter clinical trials. We employed unsupervised machine learning methods, which do not rely on labelled data, and therefore allow for the automated discovery of previously unseen irregularity patterns while maintaining flexibility when applied to new data with different structures. This pilot study employs unsupervised machine learning to compute distance matrices between centres, which we used to produce centre-level continuous features. We then used a one-class support vector machine to learn the underlying distribution of each data set to identify data that was substantially different from these distributions. We evaluated our approach against current automatable centralized monitoring methods on two trials with known irregularities. While current approaches performed well on one trial (AUROC 0.752 for monitoring vs. 0.584 for machine learning), our techniques performed substantially better on the other (AUROC 0.140 for monitoring vs 0.728 for machine learning). The results of this pilot study suggest both the feasibility and the potential value of a machine learning-based approach to irregularity detection in RCTs.

Optimizing Health Coaching for Patients With Type 2 Diabetes Using Machine Learning: Model Development and Validation Study

Background

Health coaching is an emerging intervention that has been shown to improve clinical and patient-relevant outcomes for type 2 diabetes. Advances in artificial intelligence may provide an avenue for developing a more personalized, adaptive, and cost-effective approach to diabetes health coaching.

Objective

We aim to apply Q-learning, a widely used reinforcement learning algorithm, to a diabetes health-coaching data set to develop a model for recommending an optimal coaching intervention at each decision point that is tailored to a patient’s accumulated history.

Methods

In this pilot study, we fit a two-stage reinforcement learning model on 177 patients from the intervention arm of a community-based randomized controlled trial conducted in Canada. The policy produced by the reinforcement learning model can recommend a coaching intervention at each decision point that is tailored to a patient’s accumulated history and is expected to maximize the composite clinical outcome of hemoglobin A1c reduction and quality of life improvement (normalized to [ ​0, 1 ​], with a higher score being better). Our data, models, and source code are publicly available.

Results

Among the 177 patients, the coaching intervention recommended by our policy mirrored the observed diabetes health coach’s interventions in 17.5% (n=31) of the patients in stage 1 and 14.1% (n=25) of the patients in stage 2. Where there was agreement in both stages, the average cumulative composite outcome (0.839, 95% CI 0.460-1.220) was better than those for whom the optimal policy agreed with the diabetes health coach in only one stage (0.791, 95% CI 0.747-0.836) or differed in both stages (0.755, 95% CI 0.728-0.781). Additionally, the average cumulative composite outcome predicted for the policy’s recommendations was significantly better than that of the observed diabetes health coach’s recommendations (tn-1=10.040; P<.001).

Conclusions

Applying reinforcement learning to diabetes health coaching could allow for both the automation of health coaching and an improvement in health outcomes produced by this type of intervention.

Opening the black box: the promise and limitations of explainable machine learning in cardiology

Many clinicians remain wary of machine learning due to long-standing concerns about "black box" models. "Black box" is shorthand for models that are sufficiently complex that they are not straightforwardly interpretable to humans. Lack of interpretability in predictive models can undermine trust in those models, especially in health care where so many decisions are literally life and death. There has recently been an explosion of research in the field of explainable machine learning aimed at addressing these concerns. The promise of explainable machine learning is considerable, but it is important for cardiologists who may encounter these techniques in clinical decision support tools or novel research papers to have a critical understanding of both their strengths and their limitations. This paper reviews key concepts and techniques in the field of explainable machine learning as they apply to cardiology. Key concepts reviewed include interpretability versus explainability and global versus local explanations. Techniques demonstrated include permutation importance, surrogate decision trees, local interpretable model-agnostic explanations, and partial dependence plots. We discuss several limitations with explainability techniques, focusing on the how the nature of explanations as approximations may omit important information about how black box models work and why they make certain predictions. We conclude by proposing a rule of thumb about when it is appropriate to use black box models with explanations, rather than interpretable models.

Machine Learning-Based Early Warning Systems for Clinical Deterioration: Systematic Scoping Review

Background

Timely identification of patients at a high risk of clinical deterioration is key to prioritizing care, allocating resources effectively, and preventing adverse outcomes. Vital signs–based, aggregate-weighted early warning systems are commonly used to predict the risk of outcomes related to cardiorespiratory instability and sepsis, which are strong predictors of poor outcomes and mortality. Machine learning models, which can incorporate trends and capture relationships among parameters that aggregate-weighted models cannot, have recently been showing promising results.

Objective

This study aimed to identify, summarize, and evaluate the available research, current state of utility, and challenges with machine learning–based early warning systems using vital signs to predict the risk of physiological deterioration in acutely ill patients, across acute and ambulatory care settings.

Methods

PubMed, CINAHL, Cochrane Library, Web of Science, Embase, and Google Scholar were searched for peer-reviewed, original studies with keywords related to “vital signs,” “clinical deterioration,” and “machine learning.” Included studies used patient vital signs along with demographics and described a machine learning model for predicting an outcome in acute and ambulatory care settings. Data were extracted following PRISMA, TRIPOD, and Cochrane Collaboration guidelines.

Results

We identified 24 peer-reviewed studies from 417 articles for inclusion; 23 studies were retrospective, while 1 was prospective in nature. Care settings included general wards, intensive care units, emergency departments, step-down units, medical assessment units, postanesthetic wards, and home care. Machine learning models including logistic regression, tree-based methods, kernel-based methods, and neural networks were most commonly used to predict the risk of deterioration. The area under the curve for models ranged from 0.57 to 0.97.

Conclusions

In studies that compared performance, reported results suggest that machine learning–based early warning systems can achieve greater accuracy than aggregate-weighted early warning systems but several areas for further research were identified. While these models have the potential to provide clinical decision support, there is a need for standardized outcome measures to allow for rigorous evaluation of performance across models. Further research needs to address the interpretability of model outputs by clinicians, clinical efficacy of these systems through prospective study design, and their potential impact in different clinical settings.

The emerging role of Toll-like receptor 4 in myocardial inflammation

Toll-like receptors (TLRs) are a family of pattern recognition receptors involved in cardiovascular diseases. Notably, numerous studies have demonstrated that TLR4 activates the expression of several of pro-inflammatory cytokine genes that play pivotal roles in myocardial inflammation, particularly myocarditis, myocardial infarction, ischemia-reperfusion injury, and heart failure. In addition, TLR4 is an emerging target for anti-inflammatory therapies. Given the significance of TLR4, it would be useful to summarize the current literature on the molecular mechanisms and roles of TLR4 in myocardial inflammation. Thus, in this review, we first introduce the basic knowledge of the TLR4 gene and describe the activation and signaling pathways of TLR4 in myocardial inflammation. Moreover, we highlight the recent progress of research on the involvement of TLR4 in myocardial inflammation. The information reviewed here may be useful to further experimental research and to increase the potential of TLR4 as a therapeutic target.

Presynaptic noradrenergic regulation of glutamate inputs to hypothalamic magnocellular neurones

Glutamate and norepinephrine transmitter systems play critical roles in the synaptic control of hypothalamic magnocellular neurones. We recently reported on a norepinephrine-sensitive glutamate circuit within the paraventricular nucleus (PVN) that projects to magnocellular neurones. Here, we present evidence for norepinephrine regulation of glutamate release in the PVN and supraoptic nucleus (SON) via actions on presynaptic terminals. Whole-cell synaptic currents were recorded in magnocellular neurones of the SON and PVN in an acute slice preparation. Bath application of norepinephrine (100 microm) caused a robust, reversible increase in the frequency of spontaneous glutamatergic excitatory postsynaptic currents in 100% of SON neurones (246%) and in 88% of PVN magnocellular neurones (259%). The norepinephrine-induced increase in glutamate release was mediated by activation of both presynaptic alpha1 receptors and alpha2 receptors, but the alpha1-receptor component was the predominant component of the response. The presynaptic actions of norepinephrine were predominantly, although not completely, resistant to blockade of Na-dependent spikes, implicating a presynaptic terminal locus of action. Interestingly, the spike-dependent component of the response was greater in PVN than in SON magnocellular neurones. This robust presynaptic facilitation of glutamate release by norepinephrine, combined with the known excitatory postsynaptic actions of norepinephrine, activational effects on local glutamate circuits, and inhibitory effects on gamma-aminobutyric acid release, indicate a strong excitatory role of norepinephrine in the regulation of oxytocin and vasopressin release during physiological stimulation.

Influence of sextant prostate needle biopsy or surgery on the detection and harvest of intact circulating prostate cancer cells

PURPOSE

The feasibility of harvesting intact, circulating prostate cancer cells from the blood of men with advanced prostate cancer has previously been demonstrated. We studied the influence of sextant prostate needle biopsy and radical prostatectomy on harvesting intact circulating prostate cancer cells.

MATERIALS AND METHODS

Via standard venipuncture 20 c.c. blood were obtained preoperatively, and 30 minutes and 3 days postoperatively from 23 men with clinically localized prostate cancer undergoing surgery. Similarly, blood was obtained before and after routine prostate biopsy from 13 men for an elevated prostate specific antigen level and/or abnormal digital rectal examination. The blood cells were removed via density centrifugation and magnetic cell sorting. The remaining prostate epithelial cells were characterized by indirect fluorescent immunocytochemical staining and fluorescent in situ hybridization using deoxyribonucleic acid probes.

RESULTS

Sextant biopsy of the prostate induced circulating cells in 3 of 13 men (23%), only 1 of whom demonstrated cells with aneuploidy (Gleason score 3+4 = 7). Circulating cells were detected preoperatively, 30 minutes or 3 days postoperatively in 35% of radical prostatectomy cases. Of the patients 13% had detectable circulating cells 30 minutes postoperatively only and 9% had cells harvested on postoperative day 3. Persistence of circulating prostate cancer cells was noted in 13% of men on postoperative day 3. Serum prostate specific antigen level and pathological stage did not appear to be related to harvested cell number.

CONCLUSIONS

Prostate cancer cells can be harvested from men with clinically localized disease undergoing sextant needle biopsy or radical prostatectomy. Routine prostate biopsy and surgery may influence the number of measurable circulating cells in the short term but the clinical significance and long-term prevalence of detectable circulating cells are unknown. Further studies are needed to evaluate the clinical usefulness of this assay for detecting, staging and monitoring prostate cancer.

A novel herpesvirus amplicon system for in vivo gene delivery

For gene therapy approaches to succeed, improved vector systems are needed that combine a large carrying capacity with high transduction efficiency in vivo. Towards this goal, we have developed a novel herpes simplex virus (HSV) amplicon vector, pHE, which contains an HSV-1 replication origin (ori S) and packaging sequence that permit vector replication and packaging into HSV-1 capsids. The vector also contains the Epstein-Barr virus (EBV) unique latent replication origin (ori P) sequence and a modified EBNA-1 gene to allow the vector to be maintained as an episome in transfected E5 helper cells. This system allows for efficient packaging of high-titer vector since the E5 cells are first selected for the presence of the pHE vector before helper virus infection. The infectious pHE vector has efficient transgene expression in a variety of human cell lines in vitro. Stereotactic injection of pHE vector supernatant into the rat brain resulted in high, localized reporter gene expression. Finally, the pHE vector could carry a stable 21 kb DNA payload into HSV virions. This pHE vector system should have a broad range of gene transfer applications.

Relaxin-induced expression of Fos in the forebrain of the late pregnant rat

Relaxin, administered parenterally, has been shown to increase the release of oxytocin (OT) into the circulation and increase the firing rate of OTergic neurons. The objective of the present study was to determine if relaxin administration can result in the expression of a transcription factor, suggesting that it alters transcriptional activity within OTergic neurons at the level of the hypothalamus. Primigravid rats were ovariectomized and a jugular cannula was inserted on day 11 of gestation (g11). Pregnancy was maintained by implanting 17 beta-estradiol and progesterone caplets subcutaneously at the time of ovariectomy. At gl9, rats were challenged with intravenous relaxin or isotonic saline and the brains were removed for study. Immunohistochemistry was performed on coronal brain sections, utilizing Fos as a marker of cellular activation. In the group receiving relaxin, Fos-like immunoreactivity (Fos-IR) was abundant only in the supraoptic (SON) and paraventricular nuclei (PVN) of the hypothalamus as well as in the subfornical organ (SFO). In contrast, Fos-IR in the group given isotonic saline was lacking in these three brain regions. A double label study using antibodies against Fos and OT demonstrated that a majority of the Fos-labeled cells in the hypothalamus were OTergic. Because Fos can act as a transcription factor, we interpret these data to indicate that transcription within OTergic cells is altered following relaxin administration, with abundant Fos-IR being limited to the SON and PVN of the hypothalamus and the SFO during late pregnancy in the rat.

The spatiotemporal organization of auditory, visual, and auditory-visual evoked potentials in rat cortex

Four placements of an 8 x 8 channel microelectrode array were used to map auditory, visual, and combined auditory-visual evoked potentials (AEP, VEP, AVEP) from a total of 256 electrode sites over a 7 x 7 mm2 area including most of somatosensory, auditory, and visual cortex in the right hemisphere of the rat. The unimodal AEP and VEP consisted of an archetypal response sequence representing a systematic spatial and temporal activation of primary and secondary sensory cortex. Spatiotemporal analysis of these waveforms indicated that they could be decomposed into a small number of spatial and temporal components; components that are related to patterns of specific and non-specific thalamocortical projections connecting the auditory and visual nuclei of the thalamus with primary and secondary auditory and visual cortex. These data suggest that the AEP and VEP complex are the cortical reflection of asynchronous activation of parallel thalamocortical projection systems. The areal distribution of the AEP and VEP also overlapped, primarily in secondary auditory and visual cortex, indicating that these regions contain populations of cells responding to either modality. Polymodal auditory-visual stimulation resulted in unique activation of two isolated populations of neurons positioned in secondary auditory and secondary visual cortex which were revealed by difference waveforms, computed by subtracting the sum of the AEP and VEP from the AVEP complex. Retrograde labeling of the polymodal zones indicated that they receive parallel thalamocortical projections primarily from non-specific auditory and visual thalamic nuclei including the medial and dorsal divisions of the medial geniculate nucleus (MGm and MGd), the suprageniculate nucleus (SGN), and the lateral posterior nucleus (LP). The polymodal zone in visual cortex also receives specific projections from the dorsal division of the lateral geniculate nucleus (LGd). These data conform to a general model of thalamocortical organization in which specific thalamic nuclei with a high degree of modality specificity make restricted projections to primary sensory cortex and parts of secondary sensory cortex, and association thalamic nuclei with a high degree of sensory convergence make more divergent cortical projections. Primary and secondary sensory cortex, as well as distinct zones of polysensory cortex appear to be activated in tandem via parallel thalamocortical projections. Thus, the cerebral cortex must have simultaneous access to both unimodal and polymodal sensory information.

The anatomic organization of evoked potentials in rat parietal cortex: electrically evoked commissural responses

1. Two 8 x 8 channel microelectrode arrays were positioned over 3.5 x 3.5 mm2 areas in homologous regions of right and left parietal cortex of four rats. Potentials were evoked by delivering epicortical electrical stimulation to each electrode on one hemisphere while mapping the commissural response from the contralateral array. Spatial distributions of the electrically evoked potential (EECP) complex were compared directly with cytochrome oxidase-stained sections of the recorded region. 2. Electrode sites most capable of eliciting a commissural EECP were arranged along a diagonal band extending medially from the rostral to caudal region of each electrode array, approximating the pattern of dysgranular cortex separating primary auditory (Te1) from primary somatosensory (Par1) cortex. Electrode sites in the rostromedial and caudolateral region were ineffectual in eliciting an EECP in either hemisphere. Stimulation sites within secondary visual cortex (Oc2L) also produced strong responses. Only weak responses were elicited from stimulation of Te1 and no EECP could be evoked when stimulating within Par1. 3. When an EECP in the maximally sensitive diagonal region was elicited, its spatial distribution was typically asymmetrical throughout the recording array; the response was largest along a diagonal region also extending medially from the rostral to caudal area of each electrode array. Thus the pattern of EECP in each hemisphere closely matched the pattern of electrically excitable regions in the contralateral hemisphere. 4. The EECP was usually heterogeneous. EECP distributions within the strongly responding diagonal area often formed two regions of maximum amplitude separated by a less active zone. Although responses in Te1 were significantly weaker than those in the adjacent dysgranular cortex, they also revealed a heterogeneous spatial distribution with multiple closely spaced maxima. Only responses in Oc2L appeared consistently homogeneous, with a single maximum representing the EECP. 5. These results provide functional evidence supporting a model of parietal cortex in which there are two basic types of recipient regions, densely granular regions, which are the termination sites of specific thalamocortical fibers, and dysgranular or agranular regions, which receive both ipsilateral and contralateral projections. The functional parceling of rodent parietal cortex on the basis of the spatial and temporal distribution of the epicortical evoked potential complex may be superimposed onto the anatomic parceling into granular and dysgranular zones. Implications for stages of sensory information processing are discussed.

A horseradish peroxidase study of parallel thalamocortical projections responsible for the generation of mid-latency auditory-evoked potentials

Mid-latency auditory-evoked potentials (MAEP) were recorded from the parietotemporal region of the rat using a high spatial resolution epicortical multielectrode array. Horseradish peroxidase was injected into regions of primary and secondary auditory cortex which generate spatially and temporally distinct components of the MAEP complex to retrogradely label their thalamocortical projections. These data provide anatomical evidence for three parallel thalamocortical projection systems, originating in the ventral, dorsal and medial subdivisions of the medial geniculate nucleus, which may be responsible for the asynchronous activation of three distinct subpopulations of cortical neurons giving rise to components of the MAEP complex. Specific and non-specific characteristics of the thalamocortical projections are discussed.

Polysensory evoked potentials in rat parietotemporal cortex: combined auditory and somatosensory responses

A 64 channel microelectrode array was used to map auditory evoked potentials (AEP), somatosensory evoked potentials (SEP) as well as combined auditory and somatosensory evoked potentials (ASEP) from a 7 x 7 mm2 area in rat parietotemporal neocortex. Cytochrome oxidase (CO) stained sections of layer IV were obtained in the same animals to provide anatomical information underlying epicortical field potentials. Epicortical responses evoked by click or vibrissa stimuli replicated earlier findings from our laboratory, and appeared as a family of waveforms centered on primary auditory (AI) or somatosensory (SI) cortical areas as determined from CO histology. Selective microinjections of HRP to AI and SI further confirmed their specific sensory relay nuclei in the thalamus. A small polysensory area between AI and SI, responded uniquely with an enhanced negative sharp wave to combined auditory and somatosensory stimuli. HRP retrograde labeling revealed that the thalamocortical projections to this area were from the posterior nuclear group (Po) and medial division of the medial geniculate (MGm). These data establish close relationships between epicortical AEP, SEP, and especially ASEP and corresponding cortical structures and thalamocortical projections. The neurogenesis of unimodal and polysensory evoked potentials is discussed in terms of specific and non-specific systems.

Binaural vs. monaural auditory evoked potentials in rat neocortex

High spatial resolution epicortical recording techniques and numerical modeling were used to investigate laterality effects on the middle latency auditory evoked potential (MAEP) complex. Our data confirm previous reports that auditory stimulus laterality has a consistent effect on the amplitude, timing, and spatial distribution of the MAEP complex. The earliest temporal components (P1a, P1b and N1) show the greatest sensitivity, and are absent during ipsilateral stimulation. The later positive slow wave (P2) is present at the same amplitude during all stimulation conditions. Generation of the P2 appears to be independent of prior activation of areas 36 and 41 reflected in the early components, suggesting its generation by a more diffuse thalamocortical pathway, possibly from the medial division of the medial geniculate. Serial vs. parallel activation of rodent auditory cortex is discussed in the context of laterality-sensitive MAEP components.

Anatomic organization of evoked potentials in rat parietotemporal cortex: somatosensory and auditory responses

1. Two 8 x 8-channel microelectrode arrays were used to map epicortical field potentials from a 3.5 x 3.5-mm2 area in homologous regions of right and left parietotemporal cortex of four rats. Potentials were evoked with bilaterally presented click stimuli and with bilateral tactile stimulation of the 25 major vibrissae. The spatial distribution of temporal components of the somatosensory evoked potential (SEP) and auditory evoked potential (AEP) complex were compared directly with cytochrome oxidase-stained sections of the recorded region. 2. Epicortical responses in both hemispheres to bilateral vibrissal stimuli consisted of a biphasic sharp wave (P1a-N1) constrained to the vibrissa/barrel granular region of primary somatosensory cortex (SmI). A slightly later sharp positive wave (P1b) was localized to secondary somatosensory cortex (SmII) and to perigranular cortex medial to the vibrissa/barrel field. The SEP complex ended with a biphasic slow wave (P2-N2). The P2 was centered on SmI and spread to dysgranular lateral cortex, caudal to but excluding SmII. The N2 was centered on SmII and spread to dysgranular cortex caudal to but excluding SmI. 3. The anatomic organization of the AEP in many ways approximated that of the SEP in the same animals. The timing and morphology of the AEP were nearly identical to the SEP. The AEP consisted of a P1a-N1 sharp wave constrained to the estimated region of primary auditory cortex (AI) in the lateral parietotemporal region, a later P1b localized to secondary auditory cortex (AII), and subsequent slow waves (P2 and N2) that were centered on AI and AII, respectively, and spread to dysgranular regions overlapping the distributions of the P2 and N2 of the SEP complex. 4. These data suggest that the basic neural generators for the SEP and AEP in parietotemporal cortex are quite similar, and provide evidence for the functional anatomy of each temporal component of the sensory evoked potential complex. It is concluded that the early fast waves of the SEP and AEP are modality specific and may represent the parallel activation of primary and secondary sensory cortex through established parallel afferent projections from lateral and medial thalamic nuclei. The later slow waves of the SEP and AEP appear to selectively involve primary and secondary sensory cortex but are more widely distributed, possibly reflecting a less modality-specific level of information processing in dysgranular cortex.

Topographical analysis of epileptiform potentials in rat somatosensory cortex: the interictal to ictal transition

Large quantities of penicillin were applied to the face and forelimb region of rat somatosensory cortex, producing an epileptic focus with both electrographic and behavioral signs of seizures that regularly repeated over a period of several minutes. Epicortical potentials were recorded simultaneously from a 64 channel micro-electrode array (8 x 8 platinum electrodes) with inter-electrode distances of 0.5 mm, covering a 3.5 x 3.5 mm2 area centered on the penicillin injection site. Cluster analysis was used to classify successive epileptiform discharges into interictal, transitional, and ictal groups. Principal components analysis (PCA) was used to extract fundamental waveforms producing the spike complex in each group, and to estimate the locations and spatial extent of neuronal populations participating in epileptiform discharge. During all states of epileptic excitability, it was possible to account for over 90% of the variance in the epicortical potential waveforms using a model with only two spatially overlapping populations of cells. The location and spatial extent of the populations remained unchanged by the transition to seizures; the interictal and ictal states were distinguished only by changes in the timing and amplitude of potentials in the two putative neuronal populations. The present model, using only two stationary neuronal populations to reproduce all spatiotemporal patterns in the neocortical epileptogenic focus, is compared to models proposed by others in which epileptic discharge is thought to propagate sequentially through adjacent cortex. It is concluded that the initiation, maintenance, and termination of seizures in neocortex relies on mechanisms that are not necessarily reflected in changes in spatiotemporal interactions among epicortically recorded cell groups within the focus. These mechanisms may be distinguished from those responsible for the spread of seizures within neocortex.

The functional anatomy of middle-latency auditory evoked potentials: thalamocortical connections

1. An 8 x 8-channel microelectrode array was used to map epicortical field potentials from a 4.375 x 4.375-mm2 area in the right parietotemporal neocortex of four rats. Potentials were evoked with bilaterally presented click stimuli and with electrical stimulation of the ventral and dorsal divisions of the medial geniculate body. 2. Epicortical responses to click stimuli replicated earlier findings. The responses consisted of a positive-negative biphasic waveform (P1a and N1) in the region of primary auditory cortex (area 41) and a positive monophasic waveform (P1b) in the region of secondary auditory cortex (area 36). Two potential patterns, one at the latency of the N1 and the other at the latency of the P1b, were used to represent activation of cells within areas 41 and 36. A linear combination of these patterns was sufficient to explain from 90 to 94% of the variance of the evoked potential complex at all latencies. 3. In the same animals, epicortical responses to electrical stimulation of the ventral and dorsal divisions of the medial geniculate body were also localized to areas 41 and 36, respectively. A linear combination of potential patterns from these separate stimulation conditions was sufficient to explain from 80 to 93% of the variance of the original click-evoked potential complex at all latencies. 4. These data provide functional evidence for anatomically defined topographical thalamocortical projections to primary and secondary auditory cortex. They suggest that short-latency cortical evoked potentials (10-60 ms poststimulus) are dominated by parallel thalamocortical activation of areas 41 and 36.

The functional anatomy of middle latency auditory evoked potentials

The neural origins of middle latency auditory evoked potentials (MAEP) were studied in rat cortex. MAEP were mapped from the cortical surface with a high spatial resolution electrode array. Spatiotemporal analysis, based on multivariate statistical methods, was then used to relate putative neural generators of the MAEP complex to established cytoarchitectural anatomy. These data indicate that the MAEP waveform reflects systematic asynchronous activation of both primary and secondary auditory cortex during the processing of simple click stimuli.

The electrophysiological basis of epileptiform magnetic fields in neocortex: spontaneous ictal phenomena

In a previous report (Barth, D.S. and Di, S., Brain Research, 530 (1990) 35-39), electrical measurements of epileptiform cellular currents produced by physiologically evoked interictal penicillin spikes in rat somatosensory cortex were directly compared to the extracranial magnetic fields these currents generate. The present study uses the same methodology to extend these observations to spontaneous interictal and ictal phenomena in rat auditory cortex, and provides a more realistic empirical foundation for physical models with which to interpret non-invasive neuromagnetic recordings of human focal seizures. These data indicate that seizure foci under 1 x 1 cm2 in cortical surface area are capable of producing magnetic fields that may be recorded at extracranial distances similar to those used in humans. Furthermore, physical models based on the dipole approximation appear to be appropriate for the interpretation of ictal magnetic field phenomena in neocortex.

Spatiotemporal modeling of cerebral evoked magnetic fields to median nerve stimulation

We measured somatosensory evoked magnetic fields during median nerve stimulation in 6 normal subjects. We applied multiple dipole models to study the spatiotemporal structure of early somatosensory evoked magnetic fields (SEFs), as well as the number, 3-dimensional location and time activity of their underlying neuronal sources. Two dipole sources were necessary to model the first 40 msec of SEFs explaining 85% of the data variance. Source 1 was located deeper than source 2, showed primarily a tangential orientation, and accounted for a larger part of the variance; source 2 showed no consistent orientation across subjects. Both sources showed biphasic time activities corresponding to the previously described N20-P30 and P25-N35 components. Spatiotemporal modeling could identify sources which could not be modeled consistently above noise by single moving dipoles (P25 component), revealed small latency differences of the two sources in some subjects suggesting parallel activation of these sources, and allowed separation of sources overlapping considerably both in space and time. We conclude that spatiotemporal modeling of SEFs may be useful to study functional anatomy of human sensorimotor cortex non-invasively.

Topographic analysis of field potentials in rat vibrissa/barrel cortex

An 8 x 8 multichannel microelectrode array was used to simultaneously record epicortical field potentials, evoked by displacement of contralateral vibrissae, from a 4 x 4 mm2 area of vibrissa/barrel cortex in 4 rats. The epicortical responses began with early positive (P1) and negative (N1) sharp waves, followed by slower positive (P2) and negative (N2) waves. The potential complex systematically shifted location with vibrissa stimulated, in accordance with the known somatotopic anatomy of vibrissa/barrel cortex. Topographical distributions of potentials at the P1, N1, P2 and N2 peaks were approximately concentric, but had distinct spatial extents, suggesting that they were generated by different but overlapping neuronal subpopulations. We propose that the SEP in the vibrissa/barrel cortex is produced by both sequential and parallel processing of somatosensory information, and that all components of the epicortical SEP are generated only in primary somatosensory cortex of the rat. Applications and weaknesses of topographic analysis methods are discussed.

Laminar excitability cycles in neocortex

1. Laminar field potentials produced by paired electrocortical stimuli were recorded with a linear microelectrode array inserted perpendicular to the surface of rat somatosensory cortex. Current source-density (CSD) distributions of the direct cortical response (DCR) were computed from the potential profiles. Principal component analysis (PCA) was used to estimate the time course of evoked transmembrane currents of putative pyramidal cell populations in the supragranular and infragranular layers. 2. Both supra- and infragranular cells displayed an initial period after the conditioning stimulus in which test stimuli produced subnormal evoked response amplitudes. This was followed in both layers by a long period of supernormal then subnormal responses and a second period of supernormal responses. 3. The main laminar difference encountered was a general shortening of all phases of the excitability cycle in the supragranular cells. 4. Excitability cycles in the supra- and infragranular layers closely followed the morphology of average evoked responses to the conditioning stimulus alone. These results and physiological support to the validity of lamina-specific evoked response waveforms derived from combined CSD and PCA analysis of extracellular potential measurements. 5. The relationship between evoked potential amplitude changes and cortical excitability is discussed.

Three-dimensional analysis of auditory-evoked potentials in rat neocortex

1. A 8 X 8-channel microelectrode array was used to map epicortical field potentials evoked by bilaterally presented click stimuli from a 8 X 8-mm2 area in the right parietotemporal neocortex of four rats. In two rats, a 16-channel microelectrode array was also inserted into primary auditory cortex to record the laminar profile of auditory evoked potentials (AEP). 2. The epicortical responses began with a positive-negative fast wave followed by a positive-negative slow wave, similar to the previously reported P1, N1, P2, N2 complex. Topographical distributions of the potentials at the peak of each of these waves were distinct, suggesting that they were produced by separate but overlapping populations of cells. 3. Laminar recording revealed the asynchronous participation of supragranular and infragranular pyramidal cells in the generation of the evoked-response complex. The surface-recorded P1 was primarily produced by supragranular cells and the N1, by infragranular cells. The P2 and N2 were produced by temporally overlapping contributions from both cell groups. 4. We conclude that middle-latency components of the AEP complex are produced by both sequential and parallel activation of subpopulations of pyramidal cells in primary auditory cortex.

The electrophysiological basis of epileptiform magnetic fields in neocortex

Electrical measurements of epileptiform cellular currents in a penicillin model of focal epilepsy were directly compared to the extracranial magnetic fields these currents produce. Our data support the hypothesis that epileptiform magnetic fields result from intradendritic currents oriented perpendicular to the cortical surface. Furthermore, magnetic fields could be detected from epileptic foci smaller than 3 mm2. This work provides an empirical foundation for physical models with which to interpret noninvasive neuromagnetic recordings of epileptic discharge in human focal seizure disorders.

Laminar analysis of extracellular field potentials in rat vibrissa/barrel cortex

1. A 16-channel electrode array was used to record simultaneously extracellular laminar field potentials evoked by displacement of contralateral vibrissa from vibrissa/barrel cortex in five rats. Current source-density (CSD) analysis combined with principal component analysis (PCA) was used to determine the time course of laminar-specific transmembrane currents during the evoked response. 2. The potential complex consisted of biphasic fast components followed by long-lasting slow waves. It began with activity in supragranular cells consisting of a source in layers I-II and a sink in layers IV-V; this was followed by activation of the infragranular cells with a paired sink and source in layers I-IV and V-VI, respectively. The slow-wave sequences also began in the supragranular cells followed by infragranular neurons. 3. We propose that the fast components reflect sequential intralaminar depolarization processes, and the slow waves, hyper- or repolarization processes. These results suggest that a basic neuronal circuit, consisting of sequential activation of the supragranular and then the infragranular pyramidal cells, gives rise to the field potentials evoked by physiological stimulation. This is consistent with our previous studies of direct cortical responses (DCR) and pathological discharges of the penicillin focus.

Laminar interactions in rat motor cortex during cyclical excitability changes of the penicillin focus

Laminar interactions between neurons in rat motor cortex during cyclical seizure episodes in the penicillin focus were studied using a combination of current source-density (CSD) and principal component analysis (PCA), combined with computer-based physical modeling. These data suggest that all phases of cyclical seizure phenomena are produced by interactions between two distinct populations of neurons, the same neuronal circuits previously reported to give rise to the direct cortical response (DCR) and electrically evoked interictal penicillin spikes (EIIS). The first population consists of small pyramidal cells in the supragranular layer, and the second population consists of larger pyramidal cells in the infragranular layers with apical dendrites extending to the cortical surface. The supragranular cells serve as a trigger zone for initiating both spontaneous interictal spikes (IIS) and polyspike bursts (PSB) during seizures. Fast activity in the supragranular cells is typically followed by a hyperpolarizing slow wave that may be the result of Ca2+-activated K+ currents. This slow wave increases during seizures, possibly reflecting changes in extracellular Ca2+ associated with seizure onset and termination. The monophasic response of infragranular cells is similar for both IIS and PSB and consists of a large depolarizing shift followed by a rapid but partial repolarization period and a subsequent gradual repolarization period lasting several hundred milliseconds. The infragranular response is similar in polarity and morphology to the intracellularly recorded paroxysmal depolarization shift (PDS) and may indicate that these deeper neurons are mainly responsible for this phenomena in neocortex. Finally, there is a marked postictal slow oscillation between the supra- and infragranular layers. This oscillation appears first and largest in the supragranular cells and may reflect a disturbance in excitatory feedback in these cells produced by the disinhibitory effect of penicillin, a disturbance capable of pathologically synchronizing the epileptic neuronal aggregate sufficiently for activation of the spike-generating mechanism and subsequent seizures.

Investigation of multiple simultaneously active brain sources in the electroencephalogram

We present a method of investigating multiple simultaneously active brain sources that overlap both in space and time in the scalp electroencephalogram (EEG). In order to identify the contributions of the individual brain sources to measured potentials, we applied principal component analysis and various methods of rotating the principal components including a newly developed rotation procedure using frequency criteria. We related the results of these multivariate statistical techniques to a new physical model using multiple current dipoles with fixed anatomical locations and time-varying activities. We thus are able to study 3-dimensional location, time activity and interaction of multiple simultaneously active brain sources in the scalp EEG.

Laminar cortical interactions during epileptic spikes studied with principal component analysis and physiological modeling

The direct cortical responses (DCR) to electrical stimulation and electrically evoked interictal penicillin spikes (EIIS) were studied in the same rats using current source-density (CSD) analysis to directly compare regions of neuronal depolarization and hyperpolarization in neocortex. Principal component analysis (PCA) was further used to evaluate patterns of covariance in the CSD that were characteristic of interactions between pyramidal cell populations with spatially and temporally distinct transmembrane currents. A physical model was applied to the physiological interpretation of PCA results and the optimal model parameters used to estimate neuronal generators of recorded laminar field potentials. The data suggested that the DCR and EIIS were produced by the same neuronal circuit which could be represented by two anatomically distinct populations of pyramidal cells. The first of these populations was situated in the upper and middle layers (supragranular pyramidal neurons) and formed a dipolar CSD pattern that reversed polarity in layers II and III. The second deeper population (infragranular pyramidal neurons) extended throughout most of the cortical thickness and formed a dipolar CSD pattern that reversed polarity in layer V. We propose that excitatory intracortical connections of supragranular pyramidal cells may pathologically synchronize depolarization within the epileptic focus. In this way, supragranular pyramidal cells may provide a trigger mechanism for interictal spikes in neocortex.

Long-term evaluation of human teeth after Le Fort I osteotomy: a histologic and developmental study

Transient pulpal vascular ischemia and direct injury to the apices of the teeth have been implicated as the causes of degenerative and atrophic pulpal changes in experimental animals after Le Fort I osteotomy despite the presence of collateral circulation. The long-term clinical effect of these pathologic changes in human teeth has not been studied. Seventeen maxillary third molar teeth from 10 patients whose postsurgical follow-up ranged from 6 months to 78 months (mean, 40 months) were extracted. The long-term biologic effects of Le Fort I osteotomy on the pulp and on the development of teeth were retrospectively evaluated with clinical and standard histologic techniques. Normal teeth from patients who were not operated on were used as controls. Histologic examination revealed an intact pulpal circulation and minimal pathologic changes in the pulpal tissue. Clinical and radiographic studies showed that the growing teeth developed normally after surgery. The Le Fort I downfracture procedure had little discernible long-term effect on the pulp and on the development of human third molar teeth.

Early color deprivation and subsequent color vision in a dichromatic monkey

A squirrel monkey (Saimiri sciureus) was reared for the first 4 months of life in a dim, colorless environment. Following an additional 10 months of normal visual experience, tests of color vision and spectral sensitivity were run on this animal and a control subject. The results suggest that the adult expression of dichromatic color vision does not depend on color experience during the first 4 months of life.