1Background & Mechanism of Action
Repetitive transcranial magnetic stimulation (rTMS) modulates cortical excitability via long-term potentiation (LTP)- and long-term depression (LTD)-like mechanisms. In stroke rehabilitation, rTMS targets the interhemispheric imbalance: ischemic injury reduces ipsilesional excitability while disinhibiting the contralesional hemisphere, generating transcallosal inhibition that further suppresses recovery.
Two strategies: (1) Excitatory (ipsilesional) — high-frequency rTMS (≥5 Hz) or intermittent theta burst stimulation (iTBS) promotes LTP-like neuroplasticity; (2) Inhibitory (contralesional) — low-frequency rTMS (≤1 Hz) or continuous TBS (cTBS) suppresses pathological transcallosal inhibition. TBS delivers bursts of 3 pulses at 50 Hz every 200 ms, achieving comparable effects in 3–10 minutes at 80% AMT versus 20–30 minutes at 120% rMT for conventional rTMS. [ref-1-1]
rTMS 透過快速交變磁場調節皮質興奮性，作用類似 LTP 與 LTD。中風後病灶側皮質興奮性下降， 對側半球過度活躍並透過胼胝體抑制患側，形成「半球間不平衡」。治療策略分兩類： （1）興奮性：對患側 M1 施予高頻（≥5 Hz）或 iTBS，強化皮質可塑性； （2）抑制性：對健側施予低頻（≤1 Hz）或 cTBS，降低過度活躍。 TBS 僅需 3–10 分鐘即可達到傳統 rTMS 20–30 分鐘的效果，強度更低（80% AMT），耐受性更佳。
Table 1-1. Comparison of Major rTMS Protocols
| Protocol | Frequency | Target Site | Mechanism | Duration/Session |
|---|---|---|---|---|
| HF-rTMS | ≥5 Hz (5–20 Hz) | Ipsilesional M1 | LTP-like excitation | 20–30 min |
| LF-rTMS | 1 Hz | Contralesional M1 | LTD-like inhibition | 10–20 min |
| iTBS | 50 Hz bursts @ 200 ms | Ipsilesional M1 / L-IFG | LTP-like (rapid) | 3 min (600 p) |
| cTBS | 50 Hz bursts, continuous | Contralesional M1 / L-PPC | LTD-like (rapid) | 40 s (600 p) |
| Bilateral (dual) | HF-L + LF-R | Bilateral DLPFC or M1 | Restore balance | Per modality |
References
- Jiang T, et al. Theta burst stimulation: what role does it play in stroke rehabilitation? BMC Neurol. 2024. PMID 38297193
2Motor Recovery
Upper Limb. Tang et al. (2022) analyzed 15 RCTs (N=449): excitatory rTMS over the ipsilesional hemisphere improved FMA-UE significantly (MD=5.88, 95% CI 3.32–8.43). Critical timing effect: benefit significant only within 3 months of stroke onset. [ref-2-1] Chen et al. (2022) reviewed 25 RCTs (N=535) in chronic stroke: upper limb SMD=0.49; 600-pulse protocol yielded the strongest effect (SMD=0.75). [ref-2-2]
Lower Limb. Network meta-analysis of 26 RCTs (N=943): LF-rTMS improved FMA-LE (SMD=0.34, CI 0.11–0.58); balance (SMD=0.95) and walking speed (SMD=0.36). [ref-2-3]
Spasticity. Fan et al. (2022), 42 RCTs (N=2,108): rTMS+CR reduced Modified Ashworth Scale (SMD=0.65–1.29). More than 10 sessions outperformed ≤10 sessions. [ref-2-4]
上肢：Tang 等（2022）分析 15 項 RCT（N=449），興奮性 rTMS 顯著改善 FMA-UE（MD=5.88），效益僅在中風後 3 個月內統計顯著，強調早期介入的重要性。600 脈衝方案效果最佳（SMD=0.75）。
下肢：26 項 RCT（N=943）網路統合分析顯示，低頻 rTMS 可改善下肢 FMA（SMD=0.34），平衡（SMD=0.95）與步速（SMD=0.36）改善較明顯。
痙攣：42 項 RCT（N=2,108）顯示 rTMS 能顯著降低改良 Ashworth 量表分數（SMD=0.65–1.29），超過 10 次療程效果更佳，副作用輕微。
Table 2-1. rTMS for Motor Recovery — Evidence Summary
| Domain | Best Protocol | Evidence | Key Effect | Timing | GRADE |
|---|---|---|---|---|---|
| Upper limb | HF / iTBS, ipsilesional M1, 600 p | 15–25 RCTs, N=449–535 | FMA-UE MD=5.88; SMD=0.49 | <3 months | Moderate |
| Lower limb | LF 1 Hz, contralesional M1 | 26 RCTs, N=943 | FMA-LE SMD=0.34 | Any phase | Low |
| Spasticity | LF or HF, >10 sessions | 42 RCTs, N=2,108 | MAS SMD=0.65–1.29 | Acute–Chronic | Low |
| Balance / Gait | rTMS (mixed) | 25 RCTs, N=535 | Balance SMD=0.95; Gait SMD=0.36 | Chronic | Low–Moderate |
References
- Tang Z, et al. Excitatory rTMS Over the Ipsilesional Hemisphere for Upper Limb Motor Function After Stroke. Front Neurol. 2022. PMID 35795793
- Chen G, et al. Effects of rTMS on Sequelae in Patients with Chronic Stroke. Front Neurosci. 2022. PMID 36340781
- Xie YJ, et al. rTMS for Lower Extremity Motor Function in Stroke: Network Meta-Analysis. Neural Regen Res. 2020. PMID 33269766
- Fan J, et al. Effectiveness and Safety of rTMS on Spasticity After UMN Injury. Front Neural Circuits. 2022. PMID 36426136
3Aphasia
Cheng et al. (2024) synthesized 47 RCTs (N=2,190); low-frequency rTMS (1 Hz) targeting the right inferior frontal gyrus (R-IFG) was employed in 40 of 47 studies. All language domains improved, effects persisting to 12 months. Bilateral stimulation sustained improvements ≥2 months. [ref-3-1]
Xie et al. (2025), 30 RCTs (N=1,597): auditory comprehension (MD=1.94), naming (MD=1.53), repetition (MD=1.79), spontaneous speech (MD=1.97), all p<0.001. Safety: 4 adverse events / 819 patients (0.49%). [ref-3-2]
Optimal protocol: 1 Hz LF-rTMS over R-IFG combined with speech-language therapy (SLT), acute or recovery phase.
Cheng 等（2024）納入 47 項 RCT（N=2,190），主要方案為低頻 rTMS（1 Hz）施打於右側額下回（R-IFG），各語言功能均顯著改善，效益可持續至 12 個月。Xie 等（2025）納入 30 項 RCT（N=1,597），rTMS 合併言語治療後四大語言功能均有顯著改善，安全性極佳（副作用率僅 0.49%）。建議首選方案：低頻 1 Hz rTMS（R-IFG）合併言語治療，急性或恢復期介入。
Table 3-1. rTMS for Post-Stroke Aphasia — Language Domain Improvements
| Language Domain | Studies | Effect Size (MD) | Stroke Phase | GRADE |
|---|---|---|---|---|
| Auditory Comprehension | 30 RCTs, N=1,597 | MD=1.94 (p<0.001) | All phases | Moderate |
| Naming | 22–30 RCTs, N≥1,229 | MD=1.53 (p<0.001) | All phases | Moderate |
| Repetition | 22–30 RCTs, N≥1,228 | MD=1.79 (p<0.001) | Acute–Sequelae | Moderate |
| Spontaneous Speech | 17–30 RCTs, N≥1,046 | MD=1.97 (p<0.001) | Acute–Recovery | Moderate |
References
- Cheng et al. rTMS for Post-Stroke Non-Fluent Aphasia: SR & Meta-Analysis. Front Neurol. 2024. PMID 38751884
- Xie L, et al. Efficacy and Safety of rTMS for Post-Stroke Aphasia. Front Neurol. 2025. PMID 41189653
4Post-Stroke Depression
Shao et al. (2021), 7 RCTs (N=351): rTMS significantly reduced depression scores (SMD=−1.15, 95% CI −1.62 to −0.69); remission OR=3.46 (p<0.001). [ref-4-1]
Wang et al. (2025) NMA of 12 RCTs: Bilateral DLPFC rTMS ranked first (SUCRA 95.9%, SMD=−2.33, CI −3.16 to −1.50), HF-rTMS L-DLPFC (SUCRA 65%, SMD=−1.62), iTBS (SUCRA 61.9%, SMD=−1.57), LF-rTMS (SUCRA 27.1%, SMD=−0.80). Recommend: 4 weeks bilateral DLPFC, 20 min/session. [ref-4-2]
Shao 等（2021）整合 7 項 RCT（N=351），rTMS 顯著改善憂鬱症狀（SMD=−1.15），緩解率為對照組的 3.46 倍，安全性佳、無癲癇發作。Wang 等（2025）網路統合分析（12 項 RCT，4 種方案）顯示：雙側 DLPFC rTMS 效果最佳（SUCRA 95.9%，SMD=−2.33），建議連續 4 週雙側 DLPFC rTMS（左高頻＋右低頻），每次 20 分鐘。
Table 4-1. Network Meta-Analysis: rTMS Protocols for PSD (Wang et al. 2025)
| Protocol | Target | SMD vs Sham (95% CI) | SUCRA | GRADE |
|---|---|---|---|---|
| Bilateral rTMS | L-DLPFC (HF) + R-DLPFC (LF) | −2.33 (−3.16 to −1.50) | SUCRA 95.9% | Moderate |
| HF-rTMS | Left DLPFC | −1.62 (−2.11 to −1.13) | SUCRA 65.0% | Moderate |
| iTBS | Left DLPFC | −1.57 (−2.56 to −0.57) | SUCRA 61.9% | Moderate |
| LF-rTMS | Right DLPFC | −0.80 (−1.36 to −0.23) | SUCRA 27.1% | Low |
References
- Shao D, et al. Efficacy of rTMS for Post-Stroke Depression. Braz J Med Biol Res. 2021. [Q3 — quality noted] PMID 33470386
- Wang Y, et al. rTMS for Treating Post-Stroke Depression: Network Meta-Analysis. BMC Psychiatry. 2025. PMID 40665248
5Post-Stroke Dysphagia
Wen et al. (2022), 11 RCTs (N=463): rTMS + swallowing therapy produced the largest effect size across all domains reviewed (SMD=2.15, 95% CI 1.61–2.70). Frequency effect (p=0.008): 3 Hz (SMD=2.28) > 5 Hz (SMD=1.87) > 1 Hz (SMD=1.26). LF-rTMS alone failed to improve swallowing vs controls. [ref-5-1]
Optimal protocol: HF-rTMS (3–5 Hz) combined with swallowing exercises.
Wen 等（2022）整合 11 項 RCT（N=463），rTMS 合併傳統吞嚥訓練效果顯著（SMD=2.15），為各領域中效果量最大者。頻率對療效有顯著影響（p=0.008）：3 Hz 效果最佳（SMD=2.28），1 Hz 效果最差（SMD=1.26）；低頻 rTMS 單獨使用無顯著改善。建議首選方案：高頻 rTMS（3–5 Hz）合併吞嚥訓練。
Table 5-1. rTMS for Post-Stroke Dysphagia — Frequency Comparison
| Frequency | SMD | vs Control | GRADE |
|---|---|---|---|
| 3 Hz (optimal) | 2.28 | Significant (p<0.008) | Moderate |
| 5 Hz | 1.87 | Significant | Moderate |
| 1 Hz | 1.26 | Significant (weakest) | Low–Moderate |
| LF alone (<1 Hz) | ≈0 | NOT significant | Low |
References
- Wen et al. The Effectiveness of rTMS for Post-Stroke Dysphagia: SR & Meta-Analysis. Front Hum Neurosci. 2022. PMID 35370584
6Visuospatial Neglect & Cognitive Impairment
Visuospatial Neglect. Yang et al. (2023), 11 RCTs: LF-rTMS (0.5–1 Hz) or cTBS to ipsilesional left hemisphere improved short-term line bisection (SMD=−1.10, CI −1.84 to −0.37) and cancellation tests (SMD=1.08). Long-term: line bisection SMD=−1.25, cancellation SMD=1.45. Modified cTBS (801 pulses, L-PPC) outperformed standard 600-pulse protocol. [ref-6-1]
Cognitive Impairment. rTMS produced significant cognitive benefit in chronic stroke (SMD=0.68, CI 0.32–1.05). iTBS to L-DLPFC improved executive function, memory, and attention. Evidence rated Low due to heterogeneity. [ref-6-2]
視空間忽略：Yang 等（2023）分析 11 項 RCT，對病灶側施予低頻 rTMS（0.5–1 Hz）或 cTBS，顯著改善線段二等分測試（SMD=−1.10）與刪除測試（SMD=1.08），長期效益更優（分別為 −1.25 與 1.45）。801 脈衝改良型 cTBS（左後頂葉）優於標準 600 脈衝方案。
認知功能：慢性中風期 rTMS 對認知有顯著改善（SMD=0.68）；左 DLPFC 的 iTBS 可改善執行功能、記憶與注意力，GRADE 評級為低。
Table 6-1. rTMS for Visuospatial Neglect & Cognitive Impairment
| Outcome | Protocol | RCTs | ST SMD | LT SMD | GRADE |
|---|---|---|---|---|---|
| Line Bisection | LF 1 Hz / cTBS, L-hemisphere | 11 | −1.10 | −1.25 | Low–Moderate |
| Cancellation Test | LF 1 Hz / cTBS, L-PPC | 11 | 1.08 | 1.45 | Low–Moderate |
| Cognitive Function | iTBS, L-DLPFC | ~25 | 0.68 | Limited | Low |
References
- Yang FA, et al. Short- and Long-Term Effects of rTMS on Poststroke Visuospatial Neglect. Am J Phys Med Rehabil. 2023. PMID 36730575
- Jiang T, et al. Theta Burst Stimulation in Stroke Rehabilitation. BMC Neurol. 2024. PMID 38297193
7Theta Burst Stimulation: Practical Alternative
Jiang et al. (2024), 33 RCTs. Standard TBS: 600 pulses at 80% AMT. Motor: iTBS to ipsilesional M1, FMA-UE gains persist 3 months; "priming iTBS" (cTBS pre-conditioning) enhances outcomes. Aphasia: iTBS over left IFG, effects lasting ≥3 months. Neglect: modified 801-pulse cTBS, L-PPC superior to standard 600-pulse. Cognition: iTBS to L-DLPFC. Spasticity: cerebellar targeting effective. [ref-7-1]
TBS vs conventional rTMS: 3–10 min vs 20–30 min; 80% AMT vs 120% rMT; no evidence of inferiority across domains.
Jiang 等（2024）整合 33 項 RCT，標準 TBS 方案為 600 脈衝（80% AMT），每次僅需 3–10 分鐘。各領域摘要：運動 — 患側 M1 的 iTBS 效益可持續 3 個月；失語症 — 左 IFG 的 iTBS 效益持續 3 個月以上；視空間忽略 — 801 脈衝改良型 cTBS（左後頂葉）優於標準方案；痙攣 — 小腦靶點有效。TBS 因療程短、強度低、耐受性佳，已成為傳統 rTMS 的可行第一線替代方案。
Table 7-1. TBS Protocol Recommendations by Domain
| Domain | TBS Type | Target Site | Pulses | GRADE |
|---|---|---|---|---|
| Upper limb motor | iTBS (excitatory) | Ipsilesional M1 | 600 | Moderate |
| Aphasia | cTBS (R-IFG) or iTBS (L-IFG) | R-IFG / L-IFG | 600 | Moderate |
| Visuospatial neglect | cTBS (inhibitory) | L-posterior parietal | 801 modified | Low–Moderate |
| Cognitive impairment | iTBS (excitatory) | L-DLPFC | 600 | Low |
| Spasticity | iTBS / cTBS | Cerebellar or M1 | 600 | Low |
| Dysphagia | iTBS | Bilateral / cerebellar | 600 | Low |
References
- Jiang T, et al. Theta Burst Stimulation: What Role Does It Play in Stroke Rehabilitation? BMC Neurol. 2024. PMID 38297193
8Comprehensive Summary & GRADE Evidence Table
GRADE certainty ranges from Low (lower limb, spasticity, cognition) to Moderate (upper limb, aphasia, PSD, dysphagia). Key shared limitations: high heterogeneity, small individual RCT sample sizes, variable stroke phases and lesion locations, non-standardized outcome measures. OpenEvidence cross-check: pending. Future priorities: large multicenter RCTs with biomarker-guided patient selection (neuroimaging, CST integrity, MEP presence).
rTMS 效益方向一致，GRADE 從低（下肢、痙攣、認知）到中等（上肢、失語症、PSD、吞嚥）不等。共同限制：高異質性、小樣本、中風期程與病灶差異。需要大型多中心 RCT 及生物標記引導的個人化治療策略。OpenEvidence 交叉驗證：待確認。
Table 8-1. Comprehensive Evidence Summary: rTMS for Stroke Rehabilitation
| Domain | Optimal Protocol | Evidence | Key Effect Size | GRADE | OE Check |
|---|---|---|---|---|---|
| Upper limb motor | HF/iTBS, ipsilesional M1, <3 mo | 15–25 RCTs, N=449–535 | FMA-UE MD=5.88; SMD=0.49 | Moderate | Pending |
| Lower limb/gait | LF 1 Hz, contralesional M1 | 26 RCTs, N=943 | FMA-LE SMD=0.34 | Low | Pending |
| Spasticity | LF or HF, >10 sessions | 42 RCTs, N=2,108 | MAS SMD=0.65–1.29 | Low | Pending |
| Aphasia | LF 1 Hz, R-IFG + SLT | 47 RCTs, N=2,190 | MD=1.5–2.0 (all domains) | Moderate | Pending |
| Post-stroke depression | Bilateral DLPFC, 4 weeks | 12 RCTs | SMD=−2.33; OR=3.46 | Moderate | Pending |
| Dysphagia | HF 3 Hz + swallowing Tx | 11 RCTs, N=463 | SMD=2.15 | Moderate | Pending |
| Visuospatial neglect | LF/cTBS 801 p, L-PPC | 11 RCTs | SMD=−1.10 (ST); −1.25 (LT) | Low–Moderate | Pending |
| Cognitive impairment | iTBS, L-DLPFC | Limited RCTs | SMD=0.68 | Low | Pending |
★Take-Home Messages
Key Clinical Takeaways
- Timing is critical for motor recovery — excitatory rTMS for upper limb must begin within 3 months of stroke onset for significant benefit (GRADE: Moderate).
- 失語症是 rTMS 效果最一致的領域 — 低頻 1 Hz（R-IFG）合併言語治療可改善四大語言功能，安全性極佳（副作用率 <0.5%），效益持續 12 個月（GRADE: Moderate）。
- Bilateral DLPFC rTMS is preferred for PSD — NMA ranks it #1 (SUCRA 95.9%, SMD=−2.33) over all unilateral protocols (GRADE: Moderate).
- 吞嚥困難的效果量最大（SMD=2.15） — 高頻 3 Hz 合併吞嚥訓練為首選；低頻單獨使用無效（GRADE: Moderate）。
- TBS is a time-efficient first-line alternative — 3–10 min at 80% AMT vs 20–30 min for conventional rTMS, comparable efficacy across all domains.
- 主要限制與未來方向 — 各領域均面臨高異質性與小樣本問題。需要大型多中心 RCT 及生物標記引導的個人化治療策略。