Radiological Reporting That Combine Continuous Speech Recognition with Error Correction by Transcriptionists

We evaluate the usefulness of a method in which re-ports created using CSR corrected by transcriptionists, a method that we suspect will allow transcriptionists of greater and lesser experience to transcribe and create radiological reports speedily and accurately. MATERIAL AND METHODS Two transcriptionists each were assigned to 2 groups according to their years of transcription experience: the two in Group A had more than 10 years ’ experience, and the two in Group B had less than 3 months ’ expe-rience. All four listened to the same 100 dictation files (voice linked text files) and created radiological reports using both conventional human transcription and a combined method in which they corrected errors in re-ports created using CSR. The transcriptionists were al-lowed unlimited opportunities to listen to the dictation and playing and recording times of dictated speech were the same. A hundred dictation files (7349 words) were divided into 10 series, each of which comprised 5 reports of plain radiographs, 3 reports of computed tomographic (CT) examination and 2 reports of mag-netic resonance (MR) imaging examination: each file ranged from 615 to 863 words. During one month, the transcriptionists completed 200 reports; each day, eac