therapy4speech.aiRecovery is a dose problem
What the evidence says works – and what patients actually receive.
- Trials with about 8.8 hours of therapy a week showed significant effects; trials with about 2 hours a week did not (Bhogal et al. 2003)
- Best outcomes at 20–50 total hours, 3–5+ days a week, with prescribed home practice – more intensity than usual clinical services provide (RELEASE Collaborators 2022, 959 patients, 25 trials)
- For children with speech sound disorders, three sessions a week beat one with the same total number of sessions (Allen 2013)
- 69 % of therapists can only deliver therapy once a week (Hegarty 2018)
Home digital practice works
The intensity that trials call for is hard to deliver in a clinic – and possible at home.
Can a phone score speech and mouth movements?
Consumer face tracking is not a clinical instrument by itself – so we treat landmarks as the raw sensor and calibrate per patient.
Video kinematics
Published facial-movement analysis reached 87–96 % accuracy in orofacial impairment assessment in stroke and ALS (Guarin 2022; Bandini 2018).
Audio scoring
Closed-vocabulary, prompted exercises make the target known; fine-tuned models for children's articulation reached ICC 0.98 against therapist transcription (Kim 2025).
Our validation plan
Pilot scores are compared with blinded therapist ratings (target agreement ≥ 0.8), with per-patient baseline calibration and therapist review of flagged attempts.
References
- Bhogal SK, Teasell R, Speechley M. Intensity of aphasia therapy, impact on recovery. Stroke. 2003.
8.8 h/week in effective trials vs ~2 h/week in trials without effect - Brady MC et al. Speech and language therapy for aphasia following stroke. Cochrane Database Syst Rev. 2016.
high-intensity, high-dose therapy associated with better outcomes - Breitenstein C et al. Intensive speech and language therapy in patients with chronic aphasia (FCET2EC). Lancet. 2017.
≥10 h/week for 3 weeks improved verbal communication - RELEASE Collaborators. Dosage, intensity and frequency of language therapy for aphasia. Stroke. 2022.
959 patients, 25 trials: best gains at 20–50 h total, 3–5+ days/week with home practice - Cavanaugh R et al. Aphasia treatment dosage in US outpatient rehabilitation. Am J Speech Lang Pathol. 2021.
median 7.5 h per episode of care vs ~20 h in research protocols - Palmer R et al. Aphasia therapy dose in UK community services. PLOS ONE. 2018.
guideline ≈ 45 min/day; delivered ≈ 32 min/week - Allen MM. Intervention efficacy and intensity for children with speech sound disorder. J Speech Lang Hear Res. 2013.
3×/week outperformed 1×/week with the same 24 sessions - Hegarty N et al. Frequency of therapy delivered by UK speech and language therapists. Int J Lang Commun Disord. 2018.
69 % deliver therapy once a week - Des Roches CA et al. Effectiveness of an impairment-based individualized rehabilitation program using an iPad-based software platform. Front Hum Neurosci. 2015.
~4 h/week of home practice produced significant language gains (n = 51) - Palmer R et al. Self-managed computerised speech and language therapy for chronic aphasia (Big CACTUS). Lancet Neurol. 2019.
278 patients; computer therapy improved word finding - Lang S et al. Digital add-ons and home-exercise adherence – review of 10 RCTs. Arch Physiother. 2022.
adherence improved in 7 of 10 trials - Mazeas A et al. Gamification and physical activity – meta-analysis of 16 RCTs. JMIR. 2022.
gamified programmes outperformed non-gamified - Norbury CF et al. The impact of nonverbal ability on prevalence and clinical presentation of language disorder. J Child Psychol Psychiatry. 2016; Bishop DVM et al. CATALISE. PLOS ONE. 2016.
developmental language disorder ≈ 7 % of children (≈ 1 in 14) - World Stroke Organization Global Fact Sheet 2022; Frederick A et al. 2022.
post-stroke aphasia burden – millions of adults worldwide - Deutsches Ärzteblatt, 19 June 2026: Logopäden-Mangel verschärft Wartezeiten.
shortage occupation; 132 days to fill a vacancy; 1.9 M prescriptions/year - Guarin DL et al. 2022; Bandini A et al. 2018 – video kinematics in ALS/stroke.
facial-movement analysis reached 87–96 % accuracy in orofacial assessment - Kim H et al. 2025 – automated scoring of children's articulation.
fine-tuned models reached ICC 0.98 vs SLP transcription
Evidence compiled by the therapy4speech.ai team for the MIT Professional Education CTO Program, 2025–2026. Summaries are ours; please consult the original publications.
If someone you love needed speech therapy – which therapist would you choose?
Pilot programme 2027 – we are looking for partner clinics: two pilot sites, a clinical advisory board and therapists who want to shape the product. Pricing on request.