The central role of the left inferior longitudinal fasciculus in the face-name retrieval network.

Unsuccessful retrieval of proper names (PNs) is commonly observed in patients suffering from neurological conditions such as stroke or epilepsy. While a large body of works has suggested that PN retrieval relies on a cortical network centered on the left anterior temporal lobe (ATL), much less is known about the white matter connections underpinning this process. Sparse studies provided evidence for a possible role of the uncinate fasciculus, but the inferior longitudinal fasciculus (ILF) might also contribute, since it mainly projects into the ATL, interconnects it with the posterior lexical interface and is engaged in common name (CN) retrieval. To ascertain this hypothesis, we assessed 58 patients having undergone a neurosurgery for a left low-grade glioma by means of a famous face naming (FFN) task. The behavioural data were processed following a multilevel lesion approach, including location-based analyses, voxel-based lesion-symptom mapping (VLSM) and disconnection-symptom mapping. Different statistical models were generated to control for sociodemographic data, familiarity, biographical knowledge and control cognitive performances (i.e., semantic and episodic memory and CN retrieval). Overall, VLSM analyses indicated that damage to the mid-to-anterior part of the ventro-basal temporal cortex was especially associated with PN retrieval deficits. As expected, tract-oriented analyses showed that the left ILF was the most strongly associated pathway. Our results provide evidence for the pivotal role of the ILF in the PN retrieval network. This novel finding paves the way for a better understanding of the pathophysiological bases underlying PN retrieval difficulties in the various neurological conditions marked by white matter abnormalities.

An articulatory study of differences and similarities between stuttered disfluencies and non-pathological disfluencies.

While stuttering-like disfluencies have long interested researchers, little is known about their articulatory realisation. Yet, such a description seems crucial in order to better understand the differences between stuttering-like and other disfluencies, and their underlying motor mechanisms. Hence, we aimed to compare the articulatory supraglottic activity in stuttered disfluencies with non-pathological disfluencies by using Electromagnetic Articulography (EMA) data. To achieve this comparison, two criteria were used for the description of articulatory activity during the disfluency. The first focused on the characterisation of the movements observable during the disfluency. This criterion gives an indication of the presence or absence of articulatory movement as well as interarticulator coupling. The second criterion provided information on the retention and anticipation of the sound preceding and following the disfluency. In order to strengthen our analysis, both criteria were examined using two complementary methodologies. The first was an expert-based analysis, while the second used a set of metrics based on velocity and acceleration. These analyses were conducted independently, and the final results of our study corresponded to the conclusions of the comparison of both methodology results. The results we obtained show that stuttered disfluencies and non-pathological disfluencies do have common characteristics. However, stuttered disfluencies and non-pathological disfluencies produced by Persons Who Stutter (PWS) present some particularities, mainly in terms of retention and anticipation, and the presence of spasmodic movements. Results are discussed in the light of different models of stuttering.

A two-case study of coarticulation in stuttered speech. An articulatory approach.

This study aims to describe the coarticulatory behaviour in stuttered speech from an articulatory point of view. Its purpose is to assess the nature of transitions between a stuttered phone and preceding and subsequent phones. Two persons who stutter were recorded by means of an electromagnetic articulograph while reading a text. The vertical movements of upper and lower lips, tongue body, tongue tip and mandible were extracted. They were then analysed during a stuttering moment and linked to the acoustic type of disfluency. Our findings showed several configurations of coarticulatory behaviour in terms of supraglottic articulatory movements. While disfluencies can be the result of a disrupted coarticulatory configuration, no systematicity has been found. Moreover, all acoustic types of disfluencies are represented in several coarticulatory configurations. Therefore, a stuttering-like disfluency is not always due to a coarticulatory disturbance, since correct coarticulatory patterns can be observed both between the disfluent sound and its previous and subsequent sounds. Furthermore, they suggest that the acoustic classification of disfluencies does not seem important for the coarticulatory behaviour.

Disruption of the left anterior ventrolateral temporal cortex along with the inferior longitudinal fasciculus impairs the ability to retrieve famous face names: Evidence from three longitudinal case studies.

The cortical distribution of the proper name (PN) retrieval network has been widely studied contrary to its connectional anatomy. Here, we report the case of three patients with a low-grade glioma damaging the mid-to-anterior part of the left temporal lobe. A longitudinal behavioural assessment showed that the surgery caused a long-lasting decline in PN retrieval performances in all patients. Furthermore, a detailed analysis of surgery-related structural disconnections revealed that interruption of the inferior longitudinal fasciculus was the unique common denominator.

Advancing Stuttering Detection via Data Augmentation, Class-Balanced Loss and Multi-Contextual Deep Learning.

Stuttering is a neuro-developmental speech impairment characterized by uncontrolled utterances (interjections) and core behaviors (blocks, repetitions, and prolongations), and is caused by the failure of speech sensorimotors. Due to its complex nature, stuttering detection (SD) is a difficult task. If detected at an early stage, it could facilitate speech therapists to observe and rectify the speech patterns of persons who stutter (PWS). The stuttered speech of PWS is usually available in limited amounts and is highly imbalanced. To this end, we address the class imbalance problem in the SD domain via a multi-branching (MB) scheme and by weighting the contribution of classes in the overall loss function, resulting in a huge improvement in stuttering classes on the SEP-28 k dataset over the baseline (StutterNet). To tackle data scarcity, we investigate the effectiveness of data augmentation on top of a multi-branched training scheme. The augmented training outperforms the MB StutterNet (clean) by a relative margin of 4.18% in macro F1-score ( F1). In addition, we propose a multi-contextual (MC) StutterNet, which exploits different contexts of the stuttered speech, resulting in an overall improvement of 4.48% in F1 over the single context based MB StutterNet. Finally, we have shown that applying data augmentation in the cross-corpora scenario can improve the overall SD performance by a relative margin of 13.23% in F1 over the clean training.