Medical Report Generation in a Structured and Interactive Way Using Speech Driven Approach

Kuru, Kaya orcid iconORCID: 0000-0002-4279-4166, Arda, Kemal and Baykal, Nazife (2007) Medical Report Generation in a Structured and Interactive Way Using Speech Driven Approach. In: 4th National Conference on Medical Informatics, November,15-18, 2007, Belek-ANTALYA.

[thumbnail of VOR]
PDF (VOR) - Published Version

Official URL:


The primary objective of the present study is to develop a method, which is about the capturing of medical data in the field of medical reporting as an alternative interface additional to keyboard, mouse and transcriptionists. In this view, we built a pilot speech-interfaced application for medical reporting. In this study, we blend the advantages of isolated speech recognition (ISR) as being speaker-independent and continuous speech recognition (CSR) as being speaker-dependent together by restricting sequence of words suitable to the fields presented by interactive style in a very large vocabulary lexicon by a language model. We tried to point out how semi-structured (the understanding of open-ended and close-ended conditions) and interactive medical reporting benefited from SR technology by using CSR and ISR respectively. A complete interactivity is provided with an interface to remove look-away problem in radiology during examination as the user hears questions (text-tospeech) and answers these questions with his/her voice (speech-to-text) by using a headphone in a great concentration without the need to look at monitor and return computer to enter data by using mouse and keyboard without loosing hygienic environment.
Preliminary results have shown the success of SR technology at entering structured medical data with interactivity. The recognition process is done very fast and more accurately with the proposed and applied methodology when compared to the usage of user-dependent CSR in free text medical report generation. The same colon reports belonging to the same patients were generated by four different methods with the use of same user and same-trained speech files and the recognition success of SR is increased with the method that we proposed. The medical reports can be prepared more accurate with the proposed method by preventing possible lacks by guiding radiologist through examination with the system; hygienic working environments can be preserved. Over and above, the recognition success of SR can be increased to hundred percent in case of transforming the kinds of medical reports to exact structured formats, in other words, the answers of the asked questions are close-ended (interpretation is not permitted). Furthermore, medical data are collected in a structured format to be evaluated for further studies.
This study in radiology is significant as pointing out that a voice-enabled, structured, interactive reporting system can be used by giving the experts a means to produce medical reports quickly, efficiently, and much more cheaply than those, which are currently in widespread use as transcriptionists

Repository Staff Only: item control page