I met Dr. Gerald Kien through Dr. Van Ness. Dr. Kien was a professor at Northwestern University Medical School and was responsible for their laboratory.
I helped build a medical records system and helped a colleague of his prepare a set of data demonstrating tremor for an image in a textbook.
(Note: there is a completely different company called TeleMed. This is about Telemed which was formed in 1969.)
Dr. Kien was in the process of forming a venture to do computer analysis of Electrocardiograms (ECG; occasionally abbreviated EKG) by computer. The voltage signals from an electrocardiograph cart would be translated to a signal to be transmitted using FM modulation over the telephone line. For the standard twelve lead electrocardiogram, three signals would be sent in each group. Each group would take approximately 10 seconds.
There were some questions about the technology--would the phone lines be steady and noise-free enough? Preliminary research showed that teletype signals worked well over the phone, and the phase shift from tests showed that the ECG would not be distorted.
Another consideration would be the number of phone lines required to handle the incoming calls. To help with this, I wrote a program called Probability of Busy that would calculate the probability of a busy call for a given call volume on a given number of incoming phone lines. We presumed that 5% probability of busy would be acceptable. This helped us provision the set of incoming lines for the anticipated peak number of
calls.
The choice for the computer was Sigma 5 from Scientific Data Systemslater to be acquired by Xerox. The Sigma 5 was claimed to be a good computer for instrumentation for the aerospace industry. It was a 32-bit word machine, with ability to access bytes, halfwords, and doublewords. Registers were treated as the lower 16 words of memory, but didn't actually access the core. This had the advantage of not requiring other instruction types to distinguish between registers and memory.
We had two nearly identical computers. One was for production, the other was for development and testing, and the occasional business reports.
The interrupt structure was built with interfacing to the physical world nicely, with 237 interrupt levels.
The addition of their analog to digital converter was essential to getting the analog ECG into digital form. It was a 16 bit nominal converter, but the ECG was adequately digitized at 10 bit resolution every 2 milliseconds.
The incoming phones were handled by analog modems whose main function was to answer the phone call. The modem would trigger a hardware interrupt and engage the three-channel FM-encoded ECG signals. Touchtones were decoded and turned into digital bytes for the patient ID, and an asterisk character signaling change to the next lead set.
The multiplexing A/D converter cost on the order of $40,000 at the beginning. As an indication of how far the industry has come, such devices run less than $10.
From the computer side, we had the ability to send a reverse channel signal that would light a bulb on the cart at the patient bedside. At first, we would send a flashing light signaling a good transmission. A bad call could be caused by patient's muscle contractions or signal malfunction and would be signaled by a steady light. I suggested that we reverse this and use flashing light indicating a bad call. This would account for failures not detected by the computer on our end, such as a failure in along the line. It would be better to signal bad for a good signal that signal good for a failed ECG transmission.
The Sigma series provided head-per-track disks, which was essential for real-time or near-real-time. Perhaps you remember the papers about putting the directory in the middle of a disk to reduce seek time? With a head-per-track disk, you don't need to read that paper. But careful management of writing to and reading from the disk would be required.
Disk space was limited and expensive. Thus, data had to be written to tape for rereading and eventually archival storage. There is a whole way of thinking about tape that is less present these days.
One had to keep in mind limited speed of tape, but also keep in mind that to not feed the tape fast enough, it would have to stop, then restart, costing a lot of time.
Also tape is totally sequential in nature, so careful design of record structure would be required.
Each machine had two tape drives.
Almost all program input came in through the card reader. There is a set of practices that go along with maintaining programs and documentation on punched cards. The first time you drop an deck of cards with an important program. Your colleagues will be sure to tell you to draw a diagonal line across the top of the deck so that there is at least a faint chance of recovering the order.
We quickly hit on a way of storing programs on the disk, and a simple line editor that would add, insert, or delete lines in a deck.
There were no screens to view anything on. All readable output went to the line printer. Listings of key programs would cover a corner or more of your desk. The sound of the printer would give a familiar sound by which you could often identify what was being printed.
That covers the hardware, except for the air conditioning and raised flooring. See Software at Telemed.