Facilitated lexical processing accuracy and reaction times following repetitive Transcranial Magnetic Stimulation in dementia of the Alzheimer type: a case study

ABSTRACT

We investigated the potential effects of high-frequency (10 Hz) repetitive Transcranial Magnetic Stimulation (rTMS) of the bilateral Dorsolateral Prefrontal Cortex (DLPFC) (30-sessions; 2-sessions/day) on improving lexical processing in one participant with mild – Alzheimer’s disease (hereafter dementia of the Alzheimer type-DAT). Increased accuracy and faster reaction times (RTs) were reported in a lexical-decision task (LDT) up to 2-months post-intervention. The current findings indicate that high-frequency stimulation of the DLPFC might be a potential therapeutic tool to improve lexical processing in mild-DAT.

KEYWORDS:

• Dementia of the Alzheimer Type (DAT)
• Transcranial Magnetic Stimulation (TMS)
• reaction times
• lexical processing
• lexical decision task

1. Introduction

1.1. Clinical and language profile of DAT

DAT is a progressive neurodegenerative disease accounting for an estimated 60% to 80% of dementia cases (Association, 2019). Brain abnormalities (senile plaques, neurofibrillary tangles), are the hallmark of the disease, resulting in synaptic and neuronal loss, which cause the symptoms associated with DAT (DeTure & Dickson, 2019). Episodic memory deficits (e.g., loss of recent memories) appear several years before the disease onset due to impairment of hippocampus and other related structures (Bejanin et al., 2017; Rao et al., 2022 review). The progression of DAT leads to compromised semantic memory, due to lesions in the temporal lobe and other structures (Bejanin et al., 2017). Moreover, prefrontal cortex and medial temporal lobe deterioration leads to executive function (EF) (processing speed, conflict-resolution) and working memory (WM) decline (Kumar et al., 2017; Guarino et al., 2019 review; Zokaei & Husain, 2019). Patients’ progressive cognitive decline and the presence of other limitations (e.g., apraxia), result in inability to perform daily activities and reliance on a caregiver.

Language difficulties are also present in DAT due to the underlying cognitive limitations. One aspect of impaired language in DAT involves lexical processing, meaning speakers’ ability to distinguish between words of their language and pseudowords, which are formations that look like words but are not. Snyder et al. (1996) reported mild-DAT participants to accept pseudowords (1-letter difference from words e.g., helk ← held) and non-words (rearrangement of letters, e.g., atth ← that) as words to a significantly greater extent than controls in an acceptability-judgment task (AJT), while recognition of real words was intact. Additionally, mild-to-moderate-DAT participants were significantly slower and less accurate in detecting non-words than controls in a LDT due to cognitive dysfunction (Madden et al., 1999). More recently, Azevedo et al. (2015) reported comparable accuracy in words, pseudowords (e.g., glonet/legal, pronounceable, uninterpretable) and non-words (e.g., tlkpm/illegal, unpronounceable) between controls and DAT participants. Yet, DAT participants yielded significantly slower RTs in pseudowords compared to non-words and words. Impaired pseudoword processing has also been reported in Mild Cognitive impairment (MCI), a condition with similar EF limitations. In a relevant study, Manouilidou et al. (2016) found Slovene MCI participants to be less accurate and slower than controls in detecting pseudowords (interpretable, illegal e.g., *umiralec “die-er”) under time pressure in a chronometric task. This was attributed to reduced processing speed and impaired conflict resolution.

1.2. rTMS in healthy adults and DAT

rTMS has emerged as a prominent non-pharmacological and noninvasive procedure, due to its ability to modulate neuronal activity and cortical excitability (Rossi et al., 2009). Previous studies provide evidence on how rTMS influences cognition in healthy adults and clinical populations when applied over the DLPFC, due to its involvement in WM (Brunoni & Vanderhasselt, 2014 review), EFs (Panikratova et al., 2020), and language processing (Hertrich et al., 2021 review). In healthy adults, improved WM (Beynel et al., 2019), faster RTs in Stroop color – word task (Li et al., 2017), and reduced verb naming-latency (Cotelli et al., 2010) have been reported. Improved cognitive performance after rTMS over the DLPFC was also observed in MCI (Jiang et al., 2021 review).

The effect of rTMS on cognitive performance in DAT has been extensively studied, demonstrating promising results regarding its effectiveness (Šimko et al., 2022 metanalysis). Improved episodic memory and processing speed were reported up to 5-months post-treatment (Haffen et al., 2012), following rTMS (10 Hz) over the left-DLPFC (see Eliasova et al., 2014/stimulation of the right-inferior frontal gyrus). Bagattini et al. (2020) found that high-frequency (20 Hz) rTMS over the left-DLPFC, combined with cognitive training, significantly improved associative memory and visuospatial reasoning in DAT up to 12-weeks after the baseline-assessment. Additionally, Mano (2022) observed improved EFs in mild-to-moderate-DAT, up to 2-weeks post-treatment, following 10 sessions of rTMS (10 Hz) over the bilateral DLPFC, yet participants’ performance returned to baseline 6-weeks post-treatment. A recent study reported that cognitive improvement in DAT is associated with increased plasticity and cortical reactivity post-stimulation (e.g., Li et al., 2021).

rTMS has also been found to improve language in DAT. In a randomized, sham-controlled study, Cotelli et al. (2008) applied online-rTMS (20 Hz) over the bilateral DLPFC during a picture-naming task. Mild-DAT and moderate-DAT groups showed significantly increased accuracy in action-naming, while improved object-naming was observed only in moderate-DAT. In another double-blind study, Cotelli et al. (2011) found auditory sentence-comprehension to be significantly improved after high-frequency (20 Hz) rTMS over the DLPFC, for 10 to 20 sessions, in moderate-DAT, with the effect persisting up to 2-months post-intervention. This finding might indicate that rTMS can induce a long-term modulation of the cortical synapses related to the language network, resulting in effective language processing (Cotelli et al., 2011). More recently, effects of rTMS on improving complex sentence-comprehension were also observed in DAT (Manouilidou et al., in press).

While the effect of rTMS on other cognitive domains in DAT has been extensively studied, there is limited evidence of improved language abilities. We would like to emphasize that language impairment is a major issue that hampers communication in DAT. Thus, investigating the therapeutic potential of rTMS to overcome and delay the progression of language deficits in DAT is crucial. The current study aims to address this gap by examining whether high-frequency rTMS over the bilateral DLPFC can improve the manifested lexical processing deficits in DAT.

1.3. Processing of morphologically complex words and DLPFC stimulation

The current study is based on Manouilidou et al. (2016) and aims to assess lexical processing in DAT through a non-chronometric AJT and a chronometric LDT, using various types of words and pseudowords.

According to Schreuder and Baayen (1995), and Fruchter and Marantz (2015), speakers, when presented with a lexical item (e.g., reader), initially decompose it into its smaller constituents (reader + er). This is regardless of morphological complexity, as evidence suggests that a word such as corner also gets decomposed into corn + er. After initial, blind decomposition, the lexical processor recombines the elements and validates them for syntactic (licensing stage) and semantic compatibility (recombination stage). According to Manouilidou et al. (2016) this process requires that the speakers know and can access linguistic rules in the mental lexicon, but they also need to activate cognitive abilities such as conflict resolution and inhibition, in order to handle mismatches and conflicting representations that might arise. This becomes especially relevant when processing pseudowords which are interpretable (e.g., *umiralec “*die-er”), but they are non-existent because they violate word formation rules. Processing pseudowords increases demands on EFs. Specifically, the lexical processor when encountering an interpretable pseudoword (e.g., umiralec) creates a representation “someone who dies,” which conflicts with the linguistic rules (affix −ec attaches only to action verbs bral− “to read” and not to state ones umiral− “to die”). Thus, it needs to overcome this conflicting representation and to switch its attention (a process requiring inhibition) to the linguistic rules in order to detect the incompatible combination of the two elements (e.g., −ec + non-action verbs such as “die”) (a process requiring conflict resolution) and successfully decide whether or not the stimulus exists. The role of EFs in lexical processing suggests potential involvement of the DLPFC as well. rTMS over the DLPFC has been found to improve EFs in DAT. Thus, enhancing DLPFC appears to be a promising approach to investigate how the improvement of EFs could extend in other domains, such as lexical processing. In the current study the AJT provides evidence regarding the mental knowledge of the linguistic rules which involves controlled processing with low demands on EFs. On the other hand, the LDT provides information about processing speed, meaning how quickly speakers recognize a stimulus, access their mental lexicon to find matches, inhibit conflicting representations and resolve conflicts to decide on word status. As the process unfolds in time pressure, it involves automatic processing with high demands on EFs.

2. Study design

2.1. Participant

One female (P1), aged 75 years old (2021 recruitment date), native Slovene-speaker, right-handed, with 12 years of education, diagnosed with mild-DAT according to the criteria of Dubois et al. (2007), was recruited in the Neurology Clinic, University Medical Center of Ljubljana, Slovenia. P1 presented with a progressive episodic memory loss, starting 3 years prior to recruitment. Magnetic resonance imagining (MRI) revealed widespread brain atrophy, pronounced in the medial temporal lobe, including the hippocampus and entorhinal cortex, but no hydrocephalus. P1 had normal TSH, T3, T4, B12, folate, and other relevant biochemistry. By the date of recruitment, P1’s cognitive impairment extended to other abilities, including language (as demonstrated in the relevant sections of MMSE). A transient mild but short-lasting cognitive improvement was noted after introducing treatment with acetylcolinestarase inhibitor which had to be stopped due to the gastroenterological side effects.

P1 was eligible to participation by, also, fulfilling the following criteria: (1) a diagnosis of DAT according to the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s and Related Disorders Association (NINCDS-ADRDA), and (2) a mild level of general cognitive dysfunction (Montreal Cognitive Assessment/MoCA score between 18 and 25).

The exclusion criteria included (1) potentially confounding psychiatric disorders [e.g., psychotic symptoms or disorders, alcohol or illegal drug abuse, severe depression as assessed with the Geriatric Depression Scale (GDS): 20–30]. A diagnosis of a major depressive episode was also excluded after psychiatric evaluation. (2) potentially confounding neurological disorder (e.g., stroke, epilepsy, traumatic brain injury), and (3) writing/reading disability sufficient to impair performance in the assessment.

Additional neuropsychological evaluation was administered to gather information regarding P1’s cognitive abilities before treatment. Mini-Mental State Examination test MMSE (Folstein et al., 1975) and MoCA were conducted to detect cognitive dysfunction (e.g., orientation, language). EFs (e.g., processing speed) were measured using the Stroop task (word, color, word-color; Stroop, 1935). Forward and Backward Digit Span (Sattler & Ryan, 2009) were used to evaluate participant’s attention span and WM capacity. Language, as well as attention, inhibition etc., were assessed with category, letter, and category-switching Verbal Fluency Tasks. Participant’s performance in the assessments revealed cognitive dysfunction. Table 1 presents participants’ demographic and neuropsychological information at baseline.

Table 1. Participant’s demographics and neuropsychological scores as measured at baseline evaluation. Data are shown as raw scores, number of correct responses and T-scores.

	Baseline
Age	75
Education	12
MoCA	21
MMSE	25
Digit Span – forward Number of spans Span-length
	6
	5
Digit Span – backward Number of spans Span-length
	6
	5
Verbal Fluency – phonemic	17
Verbal Fluency – category	14
Verbal Fluency – category-switching Number of words produced Correct-switches
	7
	6
Stroop – word	33
Stroop – color	22
Stroop – color-word T-score Interference
	33
	40
GDS	15

2.2. Methodology

2.2.1. Materials

The stimuli consisted of 90 words (e.g., bralec ← bral + ec “reader”), 204 decomposable pseudowords (*umiralec ← umiral + ec “die-er”), 30 non-words (e.g., *fekarna), and 144 filler nouns (e.g., čokolada “chocolate”) (1:1 existent and non-existent words ratio) (following Manouilidou et al., 2016). Pseudowords consisted of illegal, interpretable, and pronounceable strings that violated word formation rules of Slovene and were created by using existent word-bases and affixes. Non-words followed the orthographic and phonotactic rules of Slovene, but they were uninterpretable. We matched the frequency of words with the bases of all pseudowords by using the Gigafida 2.0 corpus, while there were slight differences in length, with pseudowords being slightly longer than words and non-words. The stimuli were randomly split in two lists [List-1 and List-2/162 stimuli (45-words +15-non-words +102-pseudowords) + 72-fillers/List)] for counterbalancing purposes, as well as to simplify and shorten the tasks. The two lists were presented alternately to P1 during the evaluations. The same materials were used for the LDT and AJT.

2.2.2. Experimental procedure

The OpenSesame software (Mathôt et al., 2012) was used to design and conduct the tasks.

The linguistic evaluation was completed in one session. To avoid external factors, such as fatigue which could slow down P1’s RTs, the LTD was performed first, followed by the AJT. Experiments were presented via a portable computer (12.1,” 1280 × 800 pixels). The session lasted approximately 30–40 min, including breaks.

The LDT was divided into 2 blocks with a short break in between. Instructions were displayed on the screen and were read to P1 by a linguist examiner. P1 was instructed to press as quickly and as accurately as possible either “YES” for existent words, or “NO” for non-existent words. Prior to stimulus presentation, a fixation point was presented in the middle of the screen (500 ms). The stimuli were displayed (size 36, lower case, black font/white background) on the middle of the screen. Words, pseudowords, non-words were presented in randomized order and remained on the screen until P1’s response or until the 4000 ms time limit had elapsed. A training session, including 12 items (6/words, 6/non-words), preceded the main task. Feedback was given during training (green-line/correct-response, red-line/wrong-response), but not during the experimental session. Participant’s responses and RTs were automatically recorded by OpenSesame. The AJT procedure was identical, but without a response-time limit.

2.3. rTMS stimulation protocol

rTMS was delivered with an air-cooled butterfly (figure of 8) coil (172 × 92 mm/Cool-B65 MagVenture). The stimulation intensity was set at 51%, corresponding to 100% of P1’s resting motor threshold (RMT), which was measured prior to treatment using electromyography of the first dorsal interosseous muscle (FDI) of the dominant hand, by delivering single biphasic TMS pulses to the FDI hotspot in the primary motor cortex. RMT was defined as the minimum stimulation intensity required to induce a response of at least 50 μV peak to peak in at least 50% of the trials. Each of the rTMS sessions (one per target region), consisted of 25 trains of high-frequency repetitive biphasic TMS pulses (40 pulses/train at 10 Hz; inter train interval of 11s; inter pulse internal of 10 ms; 1000 pulses per session). Target regions of the DLPFC (L-DLPFC & R-DLPFC) were stimulated sequentially, first session always at L-DLPFC and second at R-DLPFC; stimulation targets corresponded to the F3 (L-DLPFC) and F4 (R-DLPFC) electrode positions according to the international 10–20 electrode placement system. P1 underwent daily treatment for 3 weeks (2 sessions/day; 5 days/per week × 3 weeks = 30 sessions/30.000 cumulative pulses delivered). The two sessions together lasted approximately 20–30 min, including preparation. The number of daily pulses, i.e., 2000 pulses was well within the limits considered safe for rTMS (Rossi et al., 2009).

P1 was informed at recruitment that she would receive either active-rTMS or sham-stimulation, thus she was blind to her treatment allocation. The linguist who administrated the AJT and LDT, and the rTMS therapist knew which treatment P1 underwent. The psychologist, who conducted the neuropsychological assessment, was blind to P1’s allocation. Finally, according to P1, rTMS was well-tolerated and no minor or major adverse effects were reported.

Ethics approval was obtained from the Commission for Medical Ethics of the Republic of Slovenia (protocol: 0120–166/2020/11). P1 and her caregiver provided written informed consent prior to participation.

2.4. Assessments and objectives

The neuropsychological and linguistic evaluations conducted a week prior to treatment (baseline), immediately post-intervention (W4), 2-weeks post-intervention (W6), and 2-months post-intervention (W12) seeking for potential long-term effects.

Τhe primary outcome measure of the study was to evaluate P1’s performance in the linguistic tasks at W4, W6, and W12 compared to baseline. Specifically, we investigated the potential effect of treatment on increasing accuracy and reducing LDT-latency in AJT and LDT.

Finally, the secondary outcome measure was to observe participant’s performance in the neuropsychological assessments (e.g., MMSE) before and post-intervention.

3. Results

3.1. Statistics – analysis

The Cochran’s Q test (cohrans.q function/nonpar R package) was used to compare accuracy between baseline and post-intervention evaluations. For post-hoc pairwise comparisons the McNemar χ2 test was conducted (mcnemar.test function/stats R package). Friedman Anova test (friedman.test R function) was used to test whether RTs among the evaluation time points were significantly different, followed by the Wilcoxon sign ranks (pairwise.wilcox.test/stats package & wilcoxsign test/coin package in R) for pairwise comparisons. In all comparisons RTs from the correct responses were used. P-values were Bonferroni corrected for multiples comparisons.

To detect differences in participant’s accuracy between the AJT and the LDT, we conducted Mann – Whitney U test. Kruskal – Wallis H test with post-hoc comparisons (Mann-Whitney U test – corrected for multiple comparisons) was conducted to compare P1’s accuracy and RTs on words, pseudowords, and non-words in both tasks. The tests were conducted using the statistical software SPSS 23.

3.2. Baseline comparisons

P1’s overall accuracy in the LDT was 73.5% and significantly differed (U = 10530, z = −4.69, p < .001) from her accuracy in the AJT (92.6% correct).

With respect to AJT, analysis revealed no statistically significant difference in accuracy among the different stimulus types (H (2) = 2.614, p = .271), with participant’s performance in words (97.7% correct) being comparable to both non-words (93.3% correct) and pseudowords (90% correct), and no differences between the last two being observed.

Regarding LDT, analysis revealed that P1 performed significantly worse in pseudowords (59.8% correct) compared to both words (97.7% correct) (U = 1423, z = −4.68, p < .001) and non-words (93.3% correct) (U = 508, z = −2.51, p = .036), while no dissociation was found between words and non-words (U = 322, z = −.82, p > .05).

In terms of RTs, significant differences were observed among the different stimulus types (H (2) = 65.531, p < .001) in the LDT. P1 was significantly slower in detecting pseudowords (mean-RT 2496 ms) when compared to words (mean-RT 1382 ms) (U = 150, z = −7.74, p < .001), and non-words (mean-RT 2001 ms) (U = 248, z = −2.43, p = .045). Finally, P1 was significantly faster in detecting words compared to non-words (U = 64, z = −1.43, p < .001).

3.3. Primary outcome measures

3.3.1. AJT – accuracy

Figure 1 summarizes participant’s overall mean accuracy (A) and per stimulus category (B). No significant difference was detected in participant’s overall accuracy (χ2(3, N = 162) = 3.41, p = .332) across all four evaluations (A). P1’s accuracy in the category of non-words remained comparable across the baseline, W4, W6 and W12 evaluations (χ2(3, N = 15) = 1.28, p = .732). Same findings were reported for both pseudowords (χ2(3, N = 102) = 5.87, p = .117) and words (χ2(3, N = 45) = 2, p = .572) (B).

Figure 1. Mean overall accuracy (A) & mean accuracy per stimulus category (B) across the evalution time points in AJT.

Each y-axis demonstrates P1’s mean accuracy [Panel-A (overall-accuracy); Panel-B (accuracy per stimulus-category)] from baseline to post-treatment evaluations in AJT. Each x-axis represents either the four evaluation time points (Panel-A) or the different types of stimuli (Panel-B). Error bars show standard deviations.

3.3.2. LDT – accuracy

P1’s mean overall accuracy in LDT at the four evaluations is presented in Figure 2(a). A significant difference was found in participant’s overall accuracy (χ2(3, N = 162) = 59.04, p < .001) across all evaluation time points. Post-hoc pairwise comparisons showed a significantly increased LDT-accuracy at W4 compared to baseline (χ2(1) = 23.59, p < .001), which remained at both W6 (χ2(1) = 20.48, p < .001) and W12 (χ2(1) = 24.73, p < .001).

Figure 2. Mean overall accuracy (A) & mean accuracy per stimulus category (B) across the evalution time points in LDT.

Each y-axis demonstrates P1’s mean accuracy [Panel-A (overall-accuracy); Panel-B (accuracy per stimulus-category)] from baseline to post-treatment evaluations in LDT. Each x-axis represents either the four evaluation time points (Panel-A) or the different types of stimuli (Panel-B). Asterisk (*) = statistically significant changes from baseline to post-intervention. Error bars show standard deviations.

Figure 2(b) illustrates participant’s mean accuracy in the categories of no-words, pseudowords, and words at baseline, W4, W6, and W12. With respect to non-words, analysis revealed no significant differences in P1’s accuracy across baseline, W4, W6, and W12 evaluations (χ2(3, N = 15) = .6, p = .896). On the other hand, analysis revealed that P1’s performance significantly changed from baseline to all post-treatment evaluations (χ2(3, N = 102) = 65.85, p < .001) in terms of accuracy in the category of pseudowords. P1’s accuracy significantly increased from baseline to W4 (χ2(1) = 36.21, p < .001). At W6 (χ2(1) = 16.82, p < .001) and W12 (χ2(1) = 19.11, p < .001) participant’s accuracy decreased, yet it remained significantly better compared to baseline. Regarding words, analysis revealed no differences from baseline to W4 (χ2(1) = 3.12, p = .462), W6 (χ2(1) = .25, p > .05) and W12 (χ2(1) = 0, p > .05).

3.3.3. LDT – RTs

P1’s overall average RTs (ms) at the 4 assessments are presented in Figure 3(a). Significant changes were recorded from baseline to the post-intervention evaluations (χ2(3) = 47.96, p < .001). Post-hoc pairwise comparisons revealed that LDT-latency was significantly reduced at W4 compared to baseline (T = 958, z = 6.40, p < .001), with this observation persisting at both W6 (T = 1730, z = 4.58, p < .001) and W12 (T = 1231, z = 6.03, p < .001).

Figure 3. Mean overall RTs (A) & mean RTs per stimulus category (B) across the evaluation time points in LDT.

Mean RTs [Panel-A (overall-RTs); Panel-B (RTs per stimulus-category) in LDT from baseline to post-treatment are presented in y-axis. The x-axis represents either the different evaluations (Panel-A) or the different types of stimuli (Panel-B). Asterisk (*) = statistically significant changes from baseline to postintervention. Error bars represent standard-deviations.

Figure 3(b) presents participant’s average RTs in the different types of stimuli at baseline, W4, W6 and W12. With respect to non-words, analysis revealed a significant change across the four evaluations (χ2(3) = 10.74, p = .01). Post-hoc comparisons showed that the participant was significantly faster in detecting non-words at W4 (T = 1, z = 3.109, p = .011) and W6 (T = 3, z = 2.970, p = .018) when compared to baseline. A slight increase was observed at W12, yet the difference remained significant (T = 6, z = 2.760, p = .035) compared to baseline. Statistically significant difference (χ2(3) = 45.98, p < .001) was found in participant’s average RTs in the category of pseudowords across baseline, W4, W6 and W12 evaluations. Post-comparisons revealed that P1 was significantly faster in detecting pseudowords at W4 (T = 145, z = 5.749, p < .001) and W6 (T = 214, z = 5.250, p < .001) when compared to baseline. An increase in the average RT was observed at W12, yet it remained significantly faster (T = 173, z = 5.548, p < .001) compared to baseline. Regarding words, no statistical differences were detected from baseline to W4 (T = 274, z = 1.399, p = .97), W6 (T = 543, z = −1.144, p > .05) and W12 (T = 322, z = −1.817, p = .42).

3.3.4. Secondary outcome measures

Table 2 presents P1’s performance in the neuropsychological assessments at baseline and post-intervention. Improved performance, up to W12, was observed in MMSE, Verbal Fluency Task-phonemic, Verbal Fluency Task-category, Verbal Fluency task-category switching, Stroop-color, and Stroop – color-word. Regarding the remaining assessments, participant’s performance either dropped slightly post-intervention or returned to baseline at W12.

Table 2. P1’s performance in the neuropsychological assessments at baseline and post-intervention evaluations. Data are shown as raw scores, number of correct responses and T-scores.

	Evaluations
	Baseline	W4	W6	W12
Neuropsychological assessments
MMSE	25	27	26	26
Digit Span – forward Number of spans Span-length
	6	6	6	5
	5	6	5	5
Digit Span – backward Number of spans Span-length
	6	6	6	5
	5	5	5	4
Verbal Fluency − phonemic	17	20	27	26
Verbal Fluency – category	14	15	19	15
Verbal Fluency – category-switching Number of words produced Correct-switches
	7	7	8	7
	6	6	8	6
Stroop – word	33	21	25	27
Stroop – color	22	19	25	25
Stroop – color-word T-score Interference
	33	38	39	39
	40	50	47	46
GDS	15	-	19	23

4. Discussion

We examined lexical processing in an individual with mild-DAT by conducting AJT and LDT tasks. The primary aim of the study was to investigate the feasibility of high frequency (10 Hz) rTMS over the bilateral DLPFC to improve accuracy and RTs. Moreover, the effect of rTMS on participant’s general cognitive abilities was examined. Below, we discuss the findings and implications of the study.

4.1. Baseline measures

P1 achieved very high accuracy in deciding on word status in the AJT. This performance suggests that during the early stages of DAT, individuals’ knowledge of the linguistic rules and their ability to distinguish between words, non-words and pseudowords remain intact under no time pressure despite the underlying cognitive limitations (in line with Manouilidou et al., 2016 for MCI).

P1’s performance was significantly compromised in the LDT compared to the AJT, with low accuracy in pseudowords, and slow RTs in pseudowords and non-words (as in Manouilidou et al., 2016 for MCI). This suggests preserved knowledge of linguistic rules but also slow processing speed which results from EFs impairment (as manifested by impaired Stroop performance). While P1 was initially able to decompose pseudowords into their constituents (umiral + ec), impaired processing speed delayed the inhibition of conflicting representations (someone who dies) and prevented the application of the linguistic rules, leading to failure in detecting and rejecting inappropriate combinations (*umiralec). Similarly, impaired processing speed delayed the detection of non-words (*fekarna) since they resemble words (lekarna “pharmacy”).

4.2. rTMS intervention

Our data support the efficacy of high-frequency (10 Hz) rTMS, for 30 sessions, over the bilateral DLPFC to improve language performance in mild-DAT and bring to light new insights by revealing facilitation effects on lexical processing which, to our knowledge, has remained overlooked up to this point. We will focus on the outcome observed in the LDT which is promising for the efficacy of rTMS in DAT.

First, in terms of accuracy, the results demonstrated that high-frequency rTMS over the DLPFC improved participant’s performance in the LDT. Significantly increased accuracy was observed in the impaired category of pseudowords, with the effect remaining up to 2-months post-intervention (Figure 2(b)).

The second unique effect obtained in the current study involves processing speed as measured in an LDT (RTs). While it is known that processing speed improved in cognitive assessments, such as the Stroop task (Haffen et al., 2012), to our knowledge there is no data about processing speed in relation to lexical processing. Our findings suggest that rTMS over the bilateral DLPFC is an effective tool for faster LDT-latency in mild-DAT. Faster RTs were observed in the impaired categories of non-words and pseudowords post-intervention, with the effect persisting up to 2-months post-treatment (Figure 3(b)).

The observed improvement might result from the effect of rTMS on P1’s overall cognitive abilities, such as EFs, rather than solely on language abilities. We assume that rTMS over the DLPFC led to the improvement of processing speed, an ability involved in lexical processing. Specifically, we suggest that improved processing speed, combined with P1’s linguistic knowledge, led to improved lexical processing post-treatment, with P1 becoming faster and thus more accurate in applying her knowledge of linguistic rules to resolve the conflict arising from the combination of incompatible elements (*umiralec “*dier”), as well as suppressing conflicting representations during licensing and recombination stages. Indication of improved EFs post-treatment which could correlate with improved lexical processing, was observed in P1’s performance in Stroop color-word, Stroop-word, and Verbal Fluency (phonemic & category). P1 demonstrated increased scores on these tests, with the improvement persisting up to 2-months post-intervention. Improved lexical processing resulting from enhanced EFs following DLPFC stimulation adds further support to the involvement of DLPFC in language processing (Hertrich et al., 2021).

Finally, the duration of the effects on lexical processing in mild-DAT appears to be promising. Contrary to previous studies (e.g., Mano, 2022) that investigated cognitive abilities, such as attention, we sought long-term effects of rTMS on language abilities in mild-DAT. The current findings revealed evidence of sustained beneficial effects on LDT-accuracy and LDT-latency in mild-DAT, up to two-months post-treatment. These findings add further support to Cotelli et al. (2011) that reported improved language performance in DAT up to two-months post-treatment. We assume that, similarly to other cognitive abilities, the long-term language improvement may be attributed to the ability of rTMS to induce long-term enhancement of brain plasticity and cortical reactivity (Li et al., 2021).

4.3. Limitations

The present study revealed promising findings, yet there are limitations to consider. The lack of a double-blind design could lead to a biased interpretation of the results. We tried to compensate for this limitation, by minimizing the linguist’s presence in the experimental procedures (automatic recording of P1’s responses and RTs with guidance being provided only during training). Moreover, the psychologist was blind to P1’s treatment allocation. Finally, we collected data from one participant that received active stimulation, thus a sham-controlled study with a larger sample is recommended for future research.

Conclusion

This is the first study to report promising results regarding improved lexical processing in terms of processing speed and accuracy in mild-DAT post-intervention. Our findings suggest that high-frequency (10 Hz) rTMS over the bilateral DLPFC, applied for 30 sessions (2 sessions/day; 15 days) appears to be an effective tool to improve lexical processing in mild-DAT. Increased pseudoword accuracy and faster RTs in non-word and pseudoword processing were observed in LDT, with the beneficial effects persisting up to two-months post-intervention.

Acknowledgements

We wish to thank P1 for her participation and dedication in completing the study as well as her caregivers for their commitment. Moreover, we would like to thank Tjaša Mlinarič, for her help in conducting the neuropsychological assessment before and post-treatment.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data set of this study is available from the corresponding author, CM upon request.

Additional information

Funding

The current research was supported by the following ARIS (Slovenian Research Agency) grants: projects J6-1806 (C. Manouilidou), J5-4590 (J. Bon), and programmes P6-0218 (C. Manouilidou), P5-0110 (J. Bon), P3-0171 (Z. Pirtošek).