This Santa Barbara meeting was so rich in substantive material, with novel instrumentation and diversified clinical applications and interpretations, that it inevitably points to new useful approaches and unrecognized or unanswered problems. Examining several of these topical clusters individually, one can discern an orderly pattern of "Need to know" and of alternative competitive channels of research and development that may evolve into complementary, computer implemented instrumentation styles in the hospital, the clinic, and the home. By deliberately fitting together and loosely coordinating these several independent efforts, we may find an "exordium", an introduction to the next epoch of fruitful development.

Let us examine some of the individual features that have emerged.

1. "Cart" Modularization

The meeting has sought a design for the ideal cart. This is bound to be a self-defeating effort because several different ideals that are mutually exclusive are being sought. Just as the search for the "universal antidote" for poison has turned up some utility mixtures that are widely applicable to many cases but are still inappropriate in special cases and less than ideal in others, our answer lies in modular adaptation and retrofitability of our carts so that in minutes the cart can be a good fit for the patient's problems.

2. Transponsive versus Transductive Designs

Our carts have usually embodied a transductive design philosophy. The patient is a complex signal information source to be cleverly and accurately interrogated by one or several channels of similar or diverse information gathering instrumentation. The results are then to be compared, sifted and interpreted, sometimes with an element of temporal coherency and episodal segregation, more often not.

By allowing the cart to become transponsive, to ask questions to which answers are required instead of passively listening, and by allowing the computer to initiate questions on-line and in real time, in terms of the data coming in, we achieve a much more powerful "conversational" style of diagnosis and analysis. Whether this conversation is participated in by the physician or technician is optional. Smart computers can learn to imitate, or to serve the physician without exerting much initiative.

3. Homeodynamic versus Homeostatic Approach

It is inescapably evident that much of our present electro diagnostic hardware and software treats the patient as "1 stationary".

We end up with a composite representative ECG or VCG complex to analyze and call the variat ~ns "noise". We assign a single systolic and diastolic blood pressure and hematocrit. Actually features abstracted from time-series data vary greatly from second to second, minute to minute, and hour to hour, and are often different for different parts of the body. We have made the excuse that we simply are unable to function in a sea of information and must take our chances with an arbitrary small sample of representative values. Our microcomputers can now crunch data very rapidly, however, so that we can even now convert much of our temporal variance noise into important signal, and can do an enormous lot more if we will open the door to the homeodynamic image of the patient as a creature that changes meaningfully on a circadian basis but even more important, changes in response to the body's moment by moment real status.

4. Computer Vocalization and Phonemic Recognition

The technology of human speech simulation by computer has progressed so rapidly within the last three years that we are likely to underestimate the utility of this comfortable and versatile communication link between computer and human. Speech generation chips with vocabulary of several hundred words, or custom designed sound patterns now cost a few dollars instead of the several hundred dollars that they cost three or four years ago, so that they can be added as casually as we add an additional the simple CPU. The target of the speech may be/patient, who can be given instructions, asked questions , or offered words of courtesy and encourage- ment."Thank you", "that was fine", "please try to lie very still when I say 'hold' until I tell you to relax".The message may be addressed to the physician or to the technician, and in some cases is not appropriate for the patient's ears. The patient may be all too well aware of the meaning of an atrial fibrillation warning, a notification that the implanted defibrillator has been called into action!

The converse technology of computer recognition and acceptance of instructions in human speech pattern form is still in an emerging state form and can only be used with special precautions. Training to an individual operator's speech patterns is still generally desirable and a limited, carefully selected vocabulary is required. Here we can hope for rapid advance but only after several years of development. Key emergency words can be introduced that are universally recognizable, e.g. Help!, Stop, More, Less, Call the Doctor! Remember that computer speech can easily be made multilingual.

The related technology of Robotic Vision is developing rapidly and has many applications in our "Carts", but this transfer of technology is slow to occur because of the cultural gap between medicine and Automated Manufacture.

5. The Home Medical Computer Cart

We are apt to be so immersed in the development of elegant, sophisticated and hence expensive hospital or clinic oriented medical computer carts that we may lose sight of the enormous market and opportunity for nationwide health improvement at reduced cost that inheres in the mass-produced modularly adaptable computer medical cart for home use, operated by family members or the patient himself. This class of cart, modularly adapted for diagnostic, how-goes-it, chronic monitoring and/or care, or for rehabilitation exercising1 can be offered to the patient or his health insurer for perhaps ten dollars rental per day, a month's care for the price of one day in the hospital. Medical, legal, social, financial and management problems with this new category of health care are staggering in their variety and complexity, but the benefits are well worth the efforts of meeting the challenge.

6. Custom Tailoring of the Medical Computer Cart

We have seen the development of utility computer software for a particular package of diagnostic hardware that can be user-modified to adjust parametric emphasis or deemphasis of particular disease interpretations without requiring deep understanding of the prograimnatic design, and without invalidating the careful testing and clinical utility runs that such an expensive program requires for certification and acceptance. Extension of this concept allows us to expect the emergence of cart programs for hospital or for home ~tse that are individually and/or dynamically adjusted for the patient and for his or her progress into recovery or into progressively more severe disease and complications. One can also envision risk-benefit adjustment of these programs for individual patient desires or needs.Done individually, such programs would cost many thousands of dollars, but done once on an adaptable, personally repatternable design basis, even a quarter million dollar program would be easily affordable if applied to only a few thousand patients. This friendly computer that knows the patient and his personal history, his ability and willingness to pay for personalized medical attention and his health goals and cooperativeness can do much to recreate the image of a medical care system that cares about the patient and his family personally and acts as his advocate, not his adversary.

7. Matrix Inversion and Problem Oriented Cart Design

Medical computer carts have traditionally been designed to measure a predetermined package of data with only minor variations: for example, it might measure the twelve lead electrocardiogram for a predetermined number of beats or alternatively the vectorcardiogram using a specified orthogonal lead set, and will then do its best to analyse this data without being concerned about individual patient data and history or of medical special concerns. We should fully expect that these computers will soon call up additional resources of testing in terms of their findings, so as to provide immediately needed additional and simultaneous tests in keeping with the precepts of problem-oriented diagnostic analysis.

8. Supervision of Surgically Implanted or Patient Portable Computerized Equipment

For a decade or more, the principal surgically implanted electronic I' computerized" equipment has been the relatively "dumb" continuous or I "demand" cardiac pacemaker. This equipment served an important purpose but we are now seeing the emergence of a new generation of "smart" reprogrammable implants designed to do a variety of tasks in addition to cardiac pacing; alarms, programmed injections, pain alleviators, closed loop physiological control and a variety of sensory and motor transductions are within the scope of implantable computers. A key to the success of these devices is the development of very low power demanding computers, that with a requirement of perhaps a dozen microamperes from a single, usually lithium, high voltage battery cell, can run for years without recharging or replacement. Even more elaborate computers can be used if they are allowed to "sleep" until turned on by a very low power demanding surveillance system.

The present trend is toward reprogramming and "fine-tuning" these units periodically to keep them optimized for the patient's changing health status and the latest medical opinions of optimal computer participation. Presently this is generally being done rather arbitrarily by magnetically coupled input and responsive verification of recorded status, done in a hands-on physician-supervised session. Our computer carts should embody facilities for doing this carefully and adaptively, noting the patient response over an appreciable period of sampling.

With this level of cart supervision of indwelling electronic programming, we can begin to allow ourselves to think of medical diagnosis and treatment introducing the control concepts of margins of stability, internal control loop gain, and breaking out of limit cycle episodal disease manifestations. Even the very promising chronobiological scheduling of medication and therapy lends itself to implementation by this style of supervisory cart or its portable counterpart.

9. Phase-locked Loops and their Medical Implementation Phase-lock-loop technology has a largely unrecognized role in medical diagnosis and treatment. The relatively simple VCRS (Voluntary Cardiorespiratory Synchronization) procedure implemented by a simple computer or even by the patient after a few lessons on "Going on Autopilot" can give us a much more precise and sophisticated Epitome Electrocardiogram or Vectorcardiogram as well as a valuable non-loop opening approach to autonomic control loop gain. The locking together through temporal vulnerability windows of several physiological and neurohumoral processes can give insight into many psychosomatic disease compiaxes as well as possible internal crises involving phasic superposition of individually survivable events that in superposition can be catastrophic. A handy illustration of this class of phase-locked events can be found in the clustering of PVC beats within the VCRS patient breathing cycle.

10. The Biomedical Computerized Storm Warning Concept

In designing our computerIzed Hospital, Home and Implantable Medical Control or Diagnostic devices, we are often guilty of neglecting failure modes in our enthusiasm for better and more comprehensive functional modes. The old admonition for designs to "fail safe", "fail soft" and "fail seldom" is often neglected.

We are now facing a relatively unfamiliar triple failure mode threat against which we must provide safety.

We have all heard of non-replicate redundancy strategies to protect against malfunction or nonfunction and occasionally we can afford to use triple or greater redundancy to support majority logic failure or malfunction identification. Only occasionally can we implant self-repairing components and circuitry, but we should incorporate warnings of failures even though they have been dealt with by one of the satisfactory strategems.

We now face three electronic enemies against whIch we should armor our electronic computer carts and implants. The worst enemy is our own equipment. We could use very inexpensive body surface electrodes if electrographic amplifiers did not have to protect themselves against huge overloads incidental to defibrillation or electrosurgery and similar high energy insults. We could inexpensively incorporate in several interface generators a premonitory warning signal a few milliseconds in advance of the fact to disconnect, by-pass and otherwise "batten down the hatches" and warn the interpretive computer to protect itself and be wary of incoming data.

Second in the hierarchy of disruptive electronic signals is the rapidly increasing Electromagnetic Environmental Interference that is the concern of our Electromagnetic Compatibility Engineers. We live in a sea of unsensed but often frightfully intense, often pulsatile radiated energy in the bands from broadcast up to the microwave bands of radar. To some of these, our computers are inevitably, and often erratically, responsive. Special protective measures are indicated.

Least familiar, but real, are the deliberate, superpowerful bursts of electromagnetic energy that may accompany anticlpatable military actions. These cqn destroy, as well as temporarily inconvenience, medical computers, especially those operating at low power levels. Rudimentary protection against these catastrophic pulses is probably justified.

Conclusions and Working Plans

It would appear as a workable clustering of effort that we should undertake "Cart" designs in three different, but coordinated, "tracks".

Track I represents refinement and wide utilization of the present successes in simulation, enhancing and automating the functions of the good physician. This will provide very good medical care for the next five years while Track II is implemented.

Track II represents the computer participating implementation of new medical science reformulated on the presumption of algorithmic insight into disease processes, their diagnosis and therapy. This is the Biomedical, Biomimetic High Tech track as distinguished from the Electronic Computer High Tech path.

Track III represents the public participation in computer-assisted health care in the autonomous, medically prescribed, modularly adaptable, computer-implemented systems with or without electronic linkage to "higher medical authority".