The advent of PC-based voice-recognition software for medical dictation began over 10 years ago. In those early days, due to the limitations of computer speed and memory, the learning curve was steep and adaptation was limited. The "training period" was measured in weeks and there were professional one-on-one trainers to teach the use of voice recognition for dictation purposes. Despite such great efforts, the error rates were high and a great deal of time was needed to correct errors by voice commands or keyboard entry. The voice-recognition medical records dictation system fell short of expectation, until recently.

Nowadays, commercial voice-recognition software has made tremendous strides, taking advantage of the increased processing speed of parallel computing and vast memory size – Moore's Law at work. It is now feasible to setup a low cost medical dictation system, and get it to work in days. The learning curve is much abbreviated and with improved accuracy. Nevertheless, dictating into a voice recognition system requires the user to speak into the microphone clearly and distinctly, at a pre-determined pace. Although such speech cadence appears artificial, it is not difficult to get accustomed to and is not an impediment to its use.

The voice-recognition software is also capable of acting as a voice-activated system to control computer functions, such as selecting desktop icons, opening and closing menus, and switching between open windows. Moreover, the system has a self-learning capability: When a user corrects an error with either keyboard strokes or voice commands, the software recognizes the error and will improve its accuracy next time.

The advent of PC-based voice-recognition software for medical dictation began over 10 years ago. In those early days, due to the limitations of computer speed and memory, the learning curve was steep and adaptation was limited. The "training period" was measured in weeks and there were professional one-on-one trainers to teach the use of voice recognition for dictation purposes. Despite such great efforts, the error rates were high and a great deal of time was needed to correct errors by voice commands or keyboard entry. The voice-recognition medical records dictation system fell short of expectation, until recently.

Nowadays, commercial voice-recognition software has made tremendous strides, taking advantage of the increased processing speed of parallel computing and vast memory size – Moore's Law at work. It is now feasible to setup a low cost medical dictation system, and get it to work in days. The learning curve is much abbreviated and with improved accuracy. Nevertheless, dictating into a voice recognition system requires the user to speak into the microphone clearly and distinctly, at a pre-determined pace. Although such speech cadence appears artificial, it is not difficult to get accustomed to and is not an impediment to its use.

The voice-recognition software is also capable of acting as a voice-activated system to control computer functions, such as selecting desktop icons, opening and closing menus, and switching between open windows. Moreover, the system has a self-learning capability: When a user corrects an error with either keyboard strokes or voice commands, the software recognizes the error and will improve its accuracy next time.

I will describe this personal journey in setting up a functional medical dictation system from inexpensive, off-theshelf voice-recognition software and using it for the last two years, together with a hands-on demonstration. I purchased a regular, non-medical, version of the voice recognition software for US $100, while the medical version of the voice recognition software costs well over US $1,000.

The key advantage of this software is the use of "Macro Commands", in which key phrases specified by a user can trigger the software to type a series of pre-determined sentences or phrases. For example, I equated the Macro Command "Normal Back Exam" to the typing of the following normal physical examination findings for low back pain:

  • Straight Raising Test Negative Bilaterally

  • Motor Strength 5/5

  • Deep Tendon Reflexes Symmetrical

  • No Neurological Deficits

Similarly, my Macro Command for "Normal Exam" will generate the following text:

  • Heart – Regular Rate and Rhythm

  • Lungs – Clear

  • Abdomen –– + Bowel Sound, Non-tender, Non-distended, No Masses

  • Extremities – No Edema, Dorsalis Pedis & Posterior Tibial Pulses Intact Bilaterally.

And my Macro Command for "Low Back Pain Work Restrictions" will generate:

  • No lifting over 5 kg.

  • No repetitive bending or twisting of the spine

  • No climbing ladders

Thus, by using user-specified Macro Commands, this voice recognition software will save me time by not requiring me to dictate the entire text. Such Macro Commands are difficult to implement in any manual dictation system unless a dedicated transcriptionist is used.

Additionally, having an electronic medical file makes the job of searching for specific words or phrases easy. For example, it takes less than a fraction of a second to find all my patient files in which a medication "propanolol" is mentioned. Such a voice-recognition dictation system complements well with an enterprise Electronic Health Records system to achieve "total automation."

From personal experience, it took an initial 20 hours of training the software – reading, dictating, correcting, and setting up macro commands – to get started. Thereafter, there was a period of two months of daily usage fraught with frustrating frequent typographical errors requiring repetitive corrections and re-training the software – a period during which the software "learned" and "adapted" to my personal language and dictation style, including my accent. Following this phase came the reward – this software became an integral part of daily medical practice as its accuracy continued to improve and it reached a "smooth and comfortable" level about six months after initial usage. At about the one-year mark, the incremental improvements diminished, nearing a plateau. Now, more than two and half years into my journey, the voice recognition software is an integral part of my repertoire of indispensible medical tools, as it takes me more time to hand write my office notes (even with medical abbreviations) as compared to have my office notes typed by the voice recognition system. Therefore, I regard it as invaluable as my 25 year-old stethoscope. It is definitely an investment well made.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.