Predicting language treatment response in bilingual aphasia using neural network-based patient models.

Predicting language therapy outcomes in bilinguals with aphasia (BWA) remains challenging due to the multiple pre- and poststroke factors that determine the deficits and recovery of their two languages. Computational models that simulate language impairment and treatment outcomes in BWA can help predict therapy response and identify the optimal language for treatment. Here we used the BiLex computational model to simulate the behavioral profile of language deficits and treatment response of a retrospective sample of 13 Spanish-English BWA who received therapy in one of their languages. Specifically, we simulated their prestroke naming ability and poststroke naming impairment in each language, and their treatment response in the treated and the untreated language. BiLex predicted treatment effects accurately and robustly in the treated language and captured different degrees of cross-language generalization in the untreated language in BWA. Our cross-validation approach further demonstrated that BiLex generalizes to predict treatment response for patients whose data were not used in model training. These findings support the potential of BiLex to predict therapy outcomes for BWA and suggest that computational modeling may be helpful to guide individually tailored rehabilitation plans for this population.

BiLex: A computational approach to the effects of age of acquisition and language exposure on bilingual lexical access.

Lexical access in bilinguals can be modulated by multiple factors in their individual language learning history. We developed the BiLex computational model to examine the effects of L2 age of acquisition, language use and exposure on lexical retrieval in bilingual speakers. Twenty-eight Spanish-English bilinguals and five monolinguals recruited to test and validate the model were evaluated in their picture naming skills in each language and filled out a language use questionnaire. We examined whether BiLex can (i) simulate their naming performance in each language while taking into account their L2 age of acquisition, use and exposure to each language, and (ii) predict naming performance in other participants not used in model training. Our findings showed that BiLex could accurately simulate naming performance in bilinguals, suggesting that differences in L2 age of acquisition, language use and exposure can account for individual differences in bilingual lexical access.

A Computational Account of Bilingual Aphasia Rehabilitation.

Current research on bilingual aphasia highlights the paucity in recommendations for optimal rehabilitation for bilingual aphasic patients (Roberts & Kiran, 2007; Edmonds & Kiran, 2006). In this paper, we have developed a computational model to simulate an English-Spanish bilingual language system in which language representations can vary by age of acquisition (AoA) and relative proficiency in the two languages to model individual participants. This model is subsequently lesioned by varying connection strengths between the semantic and phonological networks and retrained based on individual patient demographic information to evaluate whether or not the model's prediction of rehabilitation matched the actual treatment outcome. In most cases the model comes close to the target performance subsequent to language therapy in the language trained, indicating the validity of this model in simulating rehabilitation of naming impairment in bilingual aphasia. Additionally, the amount of cross-language transfer is limited both in the patient performance and in the model's predictions and is dependent on that specific patient's AoA, language exposure and language impairment. It also suggests how well alternative treatment scenarios would have fared, including some cases where the alternative would have done better. Overall, the study suggests how computational modeling could be used in the future to design customized treatment recipes that result in better recovery than is currently possible.

Using computational patients to evaluate illness mechanisms in schizophrenia.

BACKGROUND: Various malfunctions involving working memory, semantics, prediction error, and dopamine neuromodulation have been hypothesized to cause disorganized speech and delusions in schizophrenia. Computational models may provide insights into why some mechanisms are unlikely, suggest alternative mechanisms, and tie together explanations of seemingly disparate symptoms and experimental findings. METHODS: Eight corresponding illness mechanisms were simulated in DISCERN, an artificial neural network model of narrative understanding and recall. For this study, DISCERN learned sets of autobiographical and impersonal crime stories with associated emotion coding. In addition, 20 healthy control subjects and 37 patients with schizophrenia or schizoaffective disorder matched for age, gender, and parental education were studied using a delayed story recall task. A goodness-of-fit analysis was performed to determine the mechanism best reproducing narrative breakdown profiles generated by healthy control subjects and patients with schizophrenia. Evidence of delusion-like narratives was sought in simulations best matching the narrative breakdown profile of patients. RESULTS: All mechanisms were equivalent in matching the narrative breakdown profile of healthy control subjects. However, exaggerated prediction-error signaling during consolidation of episodic memories, termed hyperlearning, was statistically superior to other mechanisms in matching the narrative breakdown profile of patients. These simulations also systematically confused autobiographical agents with impersonal crime story agents to model fixed, self-referential delusions. CONCLUSIONS: Findings suggest that exaggerated prediction-error signaling in schizophrenia intermingles and corrupts narrative memories when incorporated into long-term storage, thereby disrupting narrative language and producing fixed delusional narratives. If further validated by clinical studies, these computational patients could provide a platform for developing and testing novel treatments.