|
 ABSTRACT
The Brain Research Institute is an interdisciplinary research unit of the UCLA Medical School, supporting basic research in fields which contribute to an understanding of brain mechanisms and behavior. In 1960 the School of Medicine was relatively young, having graduated its first class in 1955. Among the early professors to affiliate with the new medical school was Dr. H. W. Magoun, whose own research interests were in the nervous system. Under his leadership, a formal proposal was prepared requesting the University of California to establish a Brain Research Institute (BRI). The proposal was approved in December 1959 and by mid-1961, sixty-seven members from fourteen University departments (12 in the Medical School) moved into a new BRI building contiguous with the Medical School, and the Neuropsychiatric Institute.
When the Brain Research Institute at UCLA was planned in the mid 50's, very few scientists of any discipline recognized the potential of automated data processing as a research tool, and the use of computers by neurophysiologists seemed like science fiction. Dr. Ross Adey, a physician, professor of anatomy, and a ham radio operator, understood the potential of the computer. Adey convinced Magoun and Dr. Jack French, the BRI's new director, that the computer revolution would bring enormous changes to neurophysiology. In 1960, the Institute made a commitment to obtain funding from NIH to pursue computing research, and I was the electrical engineer who wrote the engineering aspects of our proposal: to establish a Data Processing Laboratory whose major focus was the design and installation of a small general purpose computer facility centered around an automated analog-digital conversion center. By the time the Institute officially opened in 1961, the BRI was awaiting funding from the Institute of Neurological Diseases and Blindness (NINDB) for its successful grant application, “The Application of Computing Techniques to Brain Function,” to develop new computer methods for treating data derived from brain study. This was the first support of computer research that NINDB had funded, and there was much controversy at NIH over the size of the grant, which was about $290,000.
The goal of the grant was to introduce computing techniques to analyze electrical activity recorded from nervous structures in a variety of living forms ranging from invertebrate to man. The BRI had ninety projects and in over half of the laboratories, electrical signals of nervous activity were observed and recorded. Electrical activity recorded from nervous tissue was classified as “spikes” or “waves”. The term “spike” referred to the fast electrical action potentials of discharging nerve cells, called neurons. “Waves” were slow rhythmic potential changes as seen in the electroencephalogram (EEG) which represented on-going oscillations from the brain. The goals of investigators studying these signals were varied, and included: relating electrical events in single nerve fibers to the physio-chemical processes that occur in the transmission of the nerve signal; monitoring electrical events in order to map pathways in the nervous system; correlating EEG changes in electrical activity with different behavioral or physiological states; classifying characteristic EEG patterns for clinical diagnosis; and monitoring electrical activity during different stages of sleep, conditioning or learning - to cite a few. All of these experiments were characterized by great masses of data which were largely assessed by the naked eye. The digital computer could provide a powerful tool for neurophysiologists to study these phenomena, but the incompatibility between the continuously time-variant nature of these physiological quantities and the discrete character code of the digital computer were major obstacles. Our immediate aim in designing the laboratory was to provide a facility to computerize the collection, editing, and processing of these data.
The newly formed Data Processing Laboratory of the Brain Research Institute was specifically designed to make the latest high speed computing techniques available in the BRI. But an analog to digital conversion system was needed to bridge the gap. In 1961, to make use of digital computers, the recorded analog data had to be digitized and stored on digital magnetic tape in computer readable format. We decided to make several analog magnetic tape recorders available for lending to individual laboratories, and to have a central facility, in DPL, with a versatile conversion system able to produce data tapes in IBM format. We wanted to enable our investigators to utilize the very advanced computing facilities available to the UCLA community at the time. We also planned to equip a small digital techniques laboratory to develop special purpose devices to investigate digitizing data at the source, and to implement the remote transmission of data stored on digital magnetic tape, between laboratories and computers. Additionally, we decided to rent a small general purpose computer and investigate the application of “on line” computing techniques to neuro-physiological data.
UCLA has had a long and established history in the field of computing, dating back to 1949 when the National Bureau of Standards supported design and construction of the Standard's Western Automatic Computer (SWAC) electronic digital equipment on campus (which continued in service until 1966). In 1956 the Western Data Processing Center was created under the joint sponsorship of the University and the International Business Machines Corporation (IBM). A 709 computer was installed in the Graduate School of Management to support research and education in the application of computing to business management. In the same period, the School of Engineering and the Department of Mathematics recruited Dr. Gerald Estrin to initiate a computer engineering program and to help expand scientific computing at UCLA. With the aid of a grant from the National Science Foundation the University established the Campus Computing Facility in 1961. In fact, in 1961, three IBM 7090 computers arrived at UCLA: one for the Western Data Processing Center, one for the Campus Computing Center, and one for a new Health Science Computing Center. Dr. Wilfred J. Dixon, Professor of Biostatistics in the Medical Center, had also obtained support from NIH for computation on medical data, and obtained a large mainframe to be housed in the Medical School. These three computers attracted the attention of the Defense Advanced Research Project Agency (DARPA) and generated the early research that led to the now famous ARPA network. I recall from conversations with Dr. Dixon that the campus mainframe-providers were very disturbed by the vast quantities of input and output data the biologists and medical people used. They believed that only small amounts of input or output data were required by scientists, assuming that the computer's speed and sophistication made appropriate use of clever algorithms. This was just prior to business applications of high speed computers, and most physical scientists believed at that time that the medical people were asking the wrong questions or did not have a discipline that was suitable for computation!
In 1960/61, the largest laboratory in the BRI was the Space Biology Laboratory (SBL), which pioneered the development of data acquisition and computing techniques. SBL programs planned to study the effects of environmental stresses, likely to be encountered in space flight, on the brain mechanisms of animals and man. In addition to having written the neurophysiological sections of the DPL grant to NIH, Dr. Adey also obtained support from the Air Force and the National Aeronautics and Space Administration to establish SBL for studying information storage mechanisms in brain systems tested under conditions of simulated ballistic flight. A major segment of the SBL program studied the electrophysiological correlates of behavior by means of electrodes implanted chronically in normal behaving animals. Dr. Adey's aim was to find the EEG relationship to behavioral states of wakefulness, alertness, and decision making. The Data Processing Laboratory was designated as the computation center to support activities of the SBL staff.
In 1961, another new program for the BRI was the Clinical Neurophysiology Unit. On a mandate from Congress, NIH had set up a program designed to support facilities in which the techniques of basic research could be applied to clinical practice. In response, the National Institute of Neurological Diseases and Blindness funded the BRI to investigate basic neurophysiological methods suitable for human patients who could receive therapeutic intervention in the nervous system for temporal lobe epilepsy. In this program the techniques of basic neurophysiology, as developed in animal experimentation, were used to study patients with neurological and psychiatric disorders. There was particular interest in correlating recordings obtained from electrodes placed on the skull of patients, with activity recorded from the cortex and deep brain structures. An operating theater, ten bed suite, electrophysiological recording unit, and the basic facilities of the Brain Research Institute were made available. In particular, the computerization of patient records would make possible studies of implanted electrodes from human patients by Dr. Mary A. B. Brazier, who had joined the BRI at that time. Dr. Brazier had been an electroencephalographer at the Massachusetts General Hospital, and had frequent contact with her colleagues at MIT. In the early days of DPL we had frequent visits with members of the Communications Biophysics Group of the MIT Research Laboratory of Electronics (1, 2).
DPL was the first integrated electronic and computer laboratory established for the express purpose of developing automated technology for nervous system research. DPL was designed to be a resource for the conception and implementation of new computer based methods by neurophysiologists and engineers in the BRI. I was driven by the need to make computer systems easier to use by medical scientists, rather than by “hacker” motivation; i.e. the need to program computers in the most clever or economical way possible. I wanted to bring computer technology to bear on the problems of neurophysiology and wanted a DPL that would enable experimenters to gain research strength and perspective from the large computing installations on our campus. This approach to computing systems was rare in 1960 and stemmed from my engineering education in which I was sensitized to a systems engineering approach to problem solving. Of course, the problem of asking meaningful neurophysiological questions still remained, and I was very dependent on my physiological colleagues.
My personal interest in brain waves stemmed from my position as an electrical engineer in the electroencephalographic department of the Neurological Institute in New York City in 1951. I had just received my Ph.D. in electrical engineering from the University of Wisconsin and my husband had obtained his Ph.D. in 1950. He was employed at the Institute for Advanced Study (IAS) in Princeton on the Electronic Computer Project. I also worked on the IAS computer project for several months, but sought a position in my own specialty. I had trouble obtaining traditional employment as an electrical engineer because my commitment was not taken seriously; a common attitude towards women's careers in the 50's. During my graduate days at Wisconsin, I had heard a talk about the electroencephalogram and believed that the study of this electrical phenomenon should be of interest to an electrical engineer. One of the compensations for the daily four hour commute between my home in Princeton, N.J., and my work in New York City, was being able to participate in the social and intellectual life of the Institute for Advanced Study (IAS) where the computing machine, called the Johniac, was being built under the leadership of John von Neumann. Frequently we went out to dinner with the von Neumanns, and Johnny, as everyone called him, always queried me about the status of my EEG research. Johnny's interest in the brain stemmed from his interest in automata. In 1954 we had the opportunity to go to Israel to build a version of the Johniac at the Weizmann Institute of Science, where my computer experience continued (3).
While employed at the Neurological Institute in 1951, I redesigned an electronic frequency analyzer for bioelectric signals. The frustrations of adjusting analog filter circuits in the early 50's, coupled with my new experiences in digital logic design, put me in a good position to introduce electronic digital techniques for recording the impulse firing pattern of neurons in 1960 (4). Consequently, I became involved with the use of computers by the neurophysiological community, and joined the BRI in 1960. When the NINDB awarded the Brain Research Institute our first computer grant, it included funding for my proposal: to design and implement an analog to digital conversion system for DPL. This system had the versatility to digitize both ongoing electrical activity amplitudes and the time interval between neuronal discharges. Both EEG and spike train investigators could make analog data tapes to format into IBM digital tapes for analysis at the Health Science Computing Facility. The facilities and capabilities of the DPL in the early 60's are described in reference (5). Renewals of this grant by NINDB provided 18 years of continuous support and enabled DPL to consistently remain a leader in the development of computer methods for neurophysiological research.
During the years 1960-1975, the period covered by this conference, DPL made pioneering contributions in the areas of: analog to digital conversion; automated data acquisition and preprocessing; signal analysis, interactive graphics, modeling and simulation, laboratory computer systems, and distributed processing systems. Throughout those years, over 100 investigators consulted with DPL staff on the introduction of computing techniques, and DPL activities contributed to the computer awareness of many more.
During this period, nervous system research evolved from a fragmented quantitative descriptive discipline to an interdisciplinary neuroscience. In the next paragraphs, I will elaborate further on the type of computing system research done in the BRI-DPL during the three five-year periods between 1960 and 1975. In my review, I will not dwell on the computer technology of the system, details of which can be found in my references. The neurophysiological questioned addressed by BRI investigators, and the role of computers as a tool in seeking answers, can be reviewed in the BRI Annual Reports (6). A brief overview of the major advancements of the laboratory follows.
1960-1965: Analog-Digital Conversion
The first high speed analog-digital conversion system on the UCLA campus was housed in DPL. In 1961, we purchased an analog-digital (A-D) conversion unit, built by the Airborne Instruments Laboratory, to meet DPL design specifications. Our original choice was to rent a computer from the Digital Equipment Corporation, because of DEC's ability to input analog data. But DEC would not rent equipment, and our budget did not allow us to purchase. Our conversion system included an analog tape playback unit, the 16 channel A-D multiplexer and converter, a storage unit, and a digital output tape, all integrated into a system by DPL staff. A storage device was necessary to hold data read from an analog tape, as the digital tape was stopped and started to prepare the spaces and format required by logical records and files. We decided to use a small computer as the storage device, because a computer both controlled data flow and ran computer programs. We rented a Control Data Corporation 160a which provided the mass storage to allow the timing and tape format required for IBM-compatible digital tapes which we processed on the Health Sciences 7090. In a typical off-line operation, the computer controlled a tape-to-tape conversion system by sampling and quantifying information recorded on analog magnetic tape and assembling it into a series of IBM characters which were recorded 200 bits per inch on digital magnetic tape. To digitize the analog data properly, rules of sampling theory were observed involving appropriate regard for the frequency spectrum of the original data, and the information to be extracted. These factors dictated such converter requirements as sampling rate, sampling aperture and quantization level. A control unit supplied the timing and control signals and the logic that provided the system with flexibility to handle either EEG or spike electrical activity. (7)
The system was designed to serve the needs of many investigators employing different data acquisition methods and assumed that suitable record and play-back equipment existed to reproduce their recorded signals. There were three primary off-line modes of operation: for spectral analyses, averaging responses, and inter-spike interval computing. The spectral analysis mode was heavily used by programs for correlation and spectral analysis which were all done on the IBM 7090. The use of the power spectrum was introduced to the study of the electroencephalogram in 1962 by Dr. Don Walter, a significant contributor to DPL throughout its history. EEG data typically covered frequencies from 3 to 40 cps, in the 20-100 microvolt range and was amplified by recording amplifiers. With the advent of the fast Fourier transform (FFT), devised by Cooley and Tukey (8), spectral calculations became quicker and easier, and became used extensively in studying the structure of EEG data (9, 10). The measures which we normally calculated were the autospectrum, cross-spectrum (both amplitude and phase), and coherence. These measures allowed estimates of the distribution of power or variance as a function of frequency and of the interrelationships of activity in given frequency bands among different signals to be assessed.
The small memory size and relatively slow speed of the CDC 160a was a limiting factor for spectral problems which required very high-speed computing, but the computer was programmed to operate in an on-line fashion for average response calculations or for microelectrode studies. Microelectrode techniques recorded electrical activity from single nerve cells surrounding the electrode. The difference in recorded amplitudes of pulses in a spike train were the basis for discriminating between several discharging neurons recorded by the same electrode. These “spike” signals were in the millivolt range and required a frequency response to 30 kc/s to reproduce accurately their waveshape; however, bandwidths of 5 or 10 kc/s were sufficient to extract information as to firing pattern. It had been hypothesized that the temporal distribution of spikes represented a coding of information. Precise information on how nerves conducted signals had come from microelectrode studies in the peripheral nervous system, where neurons increased their firing rate as a function of stimulus intensity. Microelectrode studies were also made from deep brain structures, the electrode were positioned by geometrical coordinates as with gross electrodes. It was of great interest to know whether the subtle variations in the response of brain cells to different modes of stimuli were the equivalent of some “coding” process, and much activity was
REFERENCES
Note: OCR errors may be found in this Reference List extracted from the full text article. ACM has opted to expose the complete List rather than only correct and linked references.
| |
1
|
Communications Biophysics Group of Research Laboratory of Electronics and W.M. Siebert, "Processing Neuroelectric Data". Massachusetts Instit. Technol., Tech. Rep. 351. Technology Press, Cambridge, Mass., 1959, 121 p.
|
| |
2
|
Brazier, M.A.B. (Editor), "Computer Techniques in EEG Analysis". Electroenceph. clin. Neurophysiol., 1962, Suppl. 20:I08 p.
|
| |
3
|
Estrin, Thelma, "Computers, Neuroscience and Women: (1949-1999)". Proceedings of the Annual Conference IEEE Engineering in Medicine and Biology Soc., 1984, pp. 831-836.
|
| |
4
|
Estfin, Thelma, "Recording the Impulse Firing Pattern of Neurons Utilizing Digital Techniques". Digest of the 1961 International Conference on Medical Electronics, 1961, p. 9.
|
| |
5
|
Estrin, Thelma, W. Ross Adey, Mary A.B. Brazier and R.T. Kado, "Facilities in a Brain Research Institute for Acquisition, Processing and Digital Computation of Neuro-Physiological Data". Proceedings of the Conference on Data Acquisition and Processing in Biology and Medicine, Pergamon Press, 1963, pp. 191-207.
|
| |
6
|
Brain Research institute, University of California, Los Angeles, Annual Reports 1-15, July 1961-June 1976.
|
| |
7
|
Estrin, Thelma, "A Conversion System for Neuroelectric Data". Electroencephalography and Clinical Neurophysiology, Vol. 14, 1962, pp. 414-416.
|
| |
8
|
Cooley, J.W. and J.W. Tukey, "An Algorithm for the Machine Calculation of Complex Fourier Series", Math. Comput., Vol. 19, 1965, p. 297.
|
| |
9
|
Brazier, M.A.B. and D.O. Walter, Ed., "Evaluation of Bioelectfical Data from Brain Nerve and Muscle, II Part A. Frequency and Correlation Analysis". Handbook of Electroencephalography and Clinical Neurophysiology, Vol. 5, 1973.
|
| |
10
|
Dummermuth, G., W. Walz, G. ScoUo-Lauizzari, and B. Kleiner, "Spectral Analysis of EEG Activity in Different Sleep Stages in Normal Adults". Eur. Neurol., Vol. 7, 1972, pp. 265-296.
|
| |
11
|
Culler, G.J. and B.D. Fried, The TRW Two-Station, On-line Scientific Computor". TRW Space Technology Laboratories, Vol. iii, 1964.
|
| |
12
|
Estrin, Thelma, "Neurophysiological Research Using a Remote Time-Shared Computer". Data Acquisition and Processing in Biology and Medicine, Vol. 5, Pergamon Press, 1966, pp., 117-135.
|
 |
13
|
|
| |
14
|
Estrin, Thelma, "On-Line Electroencephalographic Digital Computing System". Electroenceph. clin. Neurophysiol., Vol 19, 1965, pp. 524-526.
|
| |
15
|
Estrin, Thelma and Robert Uzgalis, "Computerized Display of Spatio-Temporal EEG Patterns". IEEE Transactions on Bio-Medical Engineering, Vol. BME-16, No. 3, 1969, pp. 192-196.
|
| |
16
|
Sclabassi, R.J., R. Buchness, and T. Estfin, "Interactive Graphics in the Analysis of Neuronal Spike Train Data". Computers in Biology and Medicine, Vol. 6, I976, pp. 163-178.
|
| |
17
|
Estrin, Thelma, R. Sclabassi and R. Buchness, "Computer Graphic Applications to Neurosurgery". Proceedings First World Conference on Medical Information (MEDINFO), North Holland Publishing Co., 1974, pp. 831-836.
|
| |
18
|
Estrin, Thelma, J.V. Wegner, R. Bettinger," Computer Generated Brain Maps". Proceedings San Diego Biomedical Symposium, 1975, pp. 369-374.
|
| |
19
|
Buchness, R., T. Estrin and J. Sue, "Use of Mumps for Interactive Graphics". 1975 Mumps Users' Group Meeting.
|
|
Adobe Acrobat
QuickTime
Windows Media Player
Real Player
Pdf
J.3