Fumio Shichijo

Artificial neural networks for source localization in the human brain

SummarySource localization in the brain remains an ill-posed problem unless further constraints about the type of sources and the head model are imposed. Human head is modeled in various ways depending critically on the computing power available and/or the required level of accuracy. Sophisticated and truly representative models may yield more accurate results in general, but at the cost of prohibitively long computer times and huge memory requirements. In conventional source localization techniques, solution source parameters are taken as those which minimize an index of performance, defined relative to the model-generated and clinically measured voltages. We propose the use of a neural network in the place of commonly employed minimization algorithms such as the Simplex Method and the Marquardt algorithm, which are iterative and time consuming. With the aid of the error-backpropagation technique, a neural network is trained to compute source parameters, starting from a voltage set measured on the scalp. Here we describe the methods of training the neural network and investigate its localization accuracy. Based on the results of extensive studies, we conclude that neural networks are highly feasible as source localizers. A trained neural networks independence of localization speed from the head model, and the rapid localization ability, makes it possible to employ the most complex head model with the ease of the simplest model. No initial parameters need to be guessed in order to start the calculation, implying a possible automation of the entire localization process. One may train the network on experimental data, if available, thereby possibly doing away with head models.

Computed Tomography Controlled Aspiration Surgery for Hypertensive Intracerebral Hemorrhage

To overcome the disadvantages of conventional surgery, we developed our own method of computed tomography controlled stereotactic aspiration surgery for hypertensive intracerebral hemorrhage. A new ultrasonic surgical aspirator was also developed to facilitate aspiration of a dense clot in the acute stage. Between 1980 and 1988, 437 patients with hypertensive intracerebral hemorrhage underwent aspiration surgery. Computed tomography controlled stereotactic aspiration surgery can be performed in hematomas of any site or any stage with minimal tissue damage, even in aged patients. Aspiration surgery led to a favorable clinical experience and outcome, especially in the acute stage.

Dipole source localization of MEG by BP neural networks

SummaryThe purpose of this study was to examine the usefulness of BP neural networks for source localization of MEG. Since the performance of this method does not depend on the complexity of brain parameters and source models, a homogeneous brain model and a single current dipole source are assumed for convenience. Localization accuracy was examined in relation to the configuration and scale of the network. As a result, average error for position and moment estimations was within 2%, while the maximum error was about 5%. It was therefore concluded that the neural network method was useful for MEG source localization, though some improvements are still necessary.

EEG Analysis Using Neural Networks to Detect Changes of Brain Conditions during Operations

In some cases, temporary or permanent occlusion of the arteries that participate in providing cerebral blood flow is needed during surgery. It is important to monitor the brain condition to prevent from a sequela. EEG (electroencephalogram) analysis is one of the most familiar methods. The goal of this study is to develop integrated EEG analysis software. The software consists of data acquisition, waveform memory management, and data analysis components. At this point in time, it can display waveform, spectrum and topography changes of each frequency component for any electrode group, and DRT (deviation ratio topography) in real time. Moreover, the system warns the operator via alarm when the EEG changes significantly.

Three dimensional solid model reconstruction for neurosurgical approach

Two different types of solid models of neurosurgical pathologies by 3-D reconstruction were presented and examined, and their clinical usefulness was appraised by neurosurgical simulation. Cranial vessel models were proposed by solidifying liquid photopolymer based on 3-D MR angiography data. These models provided total visual and tactile perceptions of the anatomy. Aneurysm model well demonstrated the root of its origin and was useful in preoperative planning. Three dimensional demonstration of arterial angles gave more informations for performing transarterial embolization of aneurysm. This method had advantages in visualising more complicated or minute structures such as cerebral vessels. Intracranial pathologies and the skull base anatomy were demonstrated by milling urethane block based on CT data. These models were useful in deciding the appropriate direction of surgical approach for lesions located in the middle fossa. This method had another advantage to demonstrate the details of anatomical relationships between macroscopic pathologies and surrounding structures. Milled models were useful for the practice of preoperative simulation because urethane was easily processed in comparison with solidified polymer.

Application of neural networks to brain dynamics identification by EEG

We have constructed a multilayered neural network system that identifies brain dynamics from electroencephalogram (EEG) data by error backpropagation (BP) learning. EEG data in the rest state with closed eyes and open eyes are measured with electrodes that are placed by the international 10-20 system. The brain dynamics are embedded in the neural networks. The developed system discriminates the dynamics of the brain dynamics of the brain activities associated with open eyes from those with closed eyes.

CT-Controlled Stereotactic Aspiration in Hypertensive Brain Hemorrhage

CT-guided stereotactic aspiration surgery for the evacuation of hypertensive brain hemorrhage (HBH) has been introduced recently. From December, 1980, to April, 1986, we used aspiration surgery in 375 cases of HBH. The 6-month postoperative outcome of aspiration surgery for acute basal-ganglionic hemorrhage was compared with that after conventional surgery of all Japan and those after conservative treatment. In basal-ganglionic hemorrhage, aspiration surgery led to a favorable clinical experience and outcome, providing the possibility of a new avenue of surgical treatment of HBH.

Analysis for Brain Activities during Operations Using Measured EEG

EEG (electroencephalogram) analysis is one of the most familiar methods. The goal of this study is to develop the integrated EEG analysis software. In some cases, temporary or permanent occlusion of the arteries, which participates the cerebral blood flow, is needed during surgery. It is important to monitor the brain condition to prevent from a sequel. In this study, we propose moving average type neural networks (MANN) are applied to analyze the EEG, too. In this method, measured EEG is divided into overlapped period of 30 seconds. A MANN is trained to predict EEG by using of previous 5 samples in each period. After the training, connection weights of each MANN are compared by inner product to evaluate changes of brain condition. According to the result of analysis of a clinical data, the usefulness of the MANN analysis and DR (deviation ratio) analysis was confirmed

Analysis of premovement components of movement-related cortical potentials in patients with Parkinson's disease or brain tumors.

We have recorded movement-related cortical potentials (MRCPs) preceding voluntary finger extension in patients with Parkinsons disease and in patients with brain tumors that may involve the motor system. Three components of potential shifts preceding the voluntary movement were identified in all subjects. There were no differences between the healthy subjects and those with Parkinsons disease, or between pre- and postoperative recordings from the same parkinsonian patients in terms of onset latencies and gradients of these potential shifts. There was no evidence of MRCPs from the depth electroencephalogram recorded from the thalamic ventrolateral (VL) nucleus. After removal of the cerebellar dentate nucleus along with a tumor, MRCPs were still identified. After removal of tumors from the supplementary motor cortex and frontal association cortex, a part of the MRCPs preceding the voluntary movement was absent. The generation of MRCPs was considered not to be greatly affected by the cerebellar system because: (1) no MRCPs were detected from the thalamic VL nucleus, which is the relay nucleus of the dentato-rubro-thalamic pathway, believed to be related to voluntary movement; (2) destruction of the VL nucleus did not cause any changes in the MRCPs recorded at other sites, and (3) the MRCPs did not disappear when the dentate nucleus, an output system, was excised.

Computer Tomography-Controlled Stereotactic Surgery

In computed tomography (CT)-controlled stereotactic surgery, the coordinate system of the CT scanner is applied to determine the target depth and direction as well as for readjustment of final probe direction. This method can be used for all types of stereotactic surgery for the brain.