Automatic speech recognition in vitreo-retinal surgery. A project for a prototypal computer-based voice-controlled vitrectomy machine.

In the past half decade automatic speech recognition techniques, software and hardware technology have matured enough to support sophisticated medical applications. The project described aimed at introducing a computer-based, voice-controlled prototype system in a simulated vitreo-retinal surgery scenario. The aim was to provide the surgeon with a tool that could significantly improve the quality and ease of work and shorten the duration of intervention. The speech recognition system allows voice entry of simple commands to simulate surgical instrument control, including the infusion pump, vitreous cutter and diathermy. The project relies on a Markov-based, speaker-dependent, commercial isolated-word recognizer, and consists of a specific recognition vocabulary and application software, created and developed by the authors. Results have been encouraging. The system performed well under the test conditions, proving robust, simple to use and accurate (over 97% average word recognition rate). On the basis of their experience, the authors believe that automatic speech recognition technology, though suffering from some limitations such as the need for training, speaker dependence and a relatively small vocabulary, and requiring extensive testing under operating conditions, merits further development and opens new perspectives for a possible new generation of surgical instruments.