Clinical Scorecard: Surface EMG may restore speech after laryngectomy

At a Glance

Key Highlights

• In a pilot study, synthesized speech had a mean word error rate of 20% for the laryngectomy patient and 15% for the healthy reference speaker.
• Five listeners transcribed silently articulated date-and-time phrases; correct word recognition was 80% for the patient and 85% for the reference speaker.
• The patient's presurgical voice was reproduced using about 8 minutes of previously recorded audio, but personalized output was assessed qualitatively rather than with a separate intelligibility measure.
• Analyses found that 9 of 20 electrodes in the patient and 11 of 18 usable electrodes in the reference speaker captured 95% of multichannel signal information; electrode selection was participant-specific.
• The study included two participants, a restricted phrase set, and one recording session; larger studies and repeated testing are needed to establish broader applicability.

Guideline-Based Recommendations

Diagnosis

• Study methods: speech intelligibility was assessed using word error rates from transcriptions by five native English-speaking listeners.

Management

• The pilot study demonstrated surface EMG speech synthesis; it did not establish routine clinical use.
• Voice personalization used presurgical recordings when available and was separate from the speaker-independent decoding process.

Monitoring & Follow-up

• The authors identified decoder stability across days and repeated electrode placement as unresolved questions.
• Personalized voice output was demonstrated qualitatively; the authors noted that quantitative intelligibility assessment remains necessary.

Risks

• The study included two participants, one session, and date-and-time phrases rather than spontaneous conversation.
• The personalization process included an intermediate text-extraction step that the researchers identified as an area for future development.

Patient & Prescribing Data

Study participant: one 60-year-old man tested about 2 months after total laryngectomy. Prior radiation-related tissue damage prevented electrolarynx use; his tracheoesophageal voice prosthesis had been removed after leakage, aspiration, and infections. The source reports no prescribing data.

The patient silently articulated 499 date-and-time phrases while signals were recorded from 20 electrodes on the face, jaw, and neck. Synthesized speech had a 20% mean word error rate. The study did not establish effectiveness for routine or spontaneous communication.

Clinical Best Practices

• Research interpretation: the authors presented the findings as preliminary feasibility evidence, not a validated clinical recommendation.
• The authors cautioned that participant-specific electrode selections should not be treated as a standard placement configuration.
• The reported intelligibility results were for synthesized speech before personalization; personalized voice output was demonstrated qualitatively.
• The authors identified larger studies, open conversational speech, and repeated testing as needed to establish broader applicability.

Related Resources & Content

• Laryngoscope Investigative Otolaryngology (study source; article URL not provided)