Automatic initial and final segmentation in cleft palate speech of Mandarin speakers

ChatGPT for shaping the future of dentistry: the potential of multi-modal large language model

The ChatGPT, a lite and conversational variant of Generative Pretrained Transformer 4 (GPT-4) developed by OpenAI, is one of the milestone Large Language Models (LLMs) with billions of parameters. LLMs have stirred up much interest among researchers and practitioners in their impressive skills in natural language processing tasks, which profoundly impact various fields. This paper mainly discusses the future applications of LLMs in dentistry. We introduce two primary LLM deployment methods in dentistry, including automated dental diagnosis and cross-modal dental diagnosis, and examine their potential applications. Especially, equipped with a cross-modal encoder, a single LLM can manage multi-source data and conduct advanced natural language reasoning to perform complex clinical operations. We also present cases to demonstrate the potential of a fully automatic Multi-Modal LLM AI system for dentistry clinical application. While LLMs offer significant potential benefits, the challenges, such as data privacy, data quality, and model bias, need further study. Overall, LLMs have the potential to revolutionize dental diagnosis and treatment, which indicates a promising avenue for clinical application and research in dentistry.

[Establishment and application of mandarin cleft palate speech database]

OBJECTIVE

This research aims to collect speech samples from patients with cleft palate, establish a mandarin-based database of cleft palate speech after sample analysis and classification, and provide a reference for the diagnosis of hypernasal or cleft palate, clinical education, and standard training for professional speech therapists and related research.

METHODS

A total of 768 speech samples were collected from patients and volunteers from the Speech Therapy Center, West China Hospital of Stomatology, between May 2016 and March 2018. These samples were edited and categoried before being saved into the cleft lip and palate biologic information database.

RESULTS

A mandarin-based database of cleft palate speech was established from 768 subjects, including 456 children (male 227, female 229), 312 adults (male 178, female 134), 369 normal speech voices, 155 low-level hypernasal samples, 102 moderate-level hypernasal samples, 142 high-level hypernasal samples, and 64 512 words, 24 576 phonemes, and 7 680 numbers.

CONCLUSIONS

This study first established a mandarin-based database of cleft palate speech, which has enormous value for the education of speech pathology of cleft palate in mandarin and further research.

The Correlation Between Consonant Articulation and Velopharyngeal Function in Patients With Unoperated Submucous Cleft Palate

Submucous cleft palate (SMCP) is a congenital condition related to abnormal muscle attachments within the soft palate. Even though hypernasality and nasal emission, the primary symptom of SMCP, have been well-documented, research on articulation errors is currently lacking. The object of this study was to examine the consonant articulation in patients with unoperated SMCP and its possible correlation with the velopharyngeal function. This study analyzed the perceptual speech assessment and nasopharyngoscopy data of 338 patients with unoperated SMCP between years 2008 and 2016 retrospectively. The mean age of the patients was 13.27 years. About 125 patients showed velopharyngeal competence (VPC), 174 velopharyngeal incompetence (VPI), and 39 marginal VPI (MVPI). Among the 338 patients, 137 presented normal articulation, 124 consonant omission, 25 compensatory articulation, 36 consonant weakening, 51 consonant substitution, and 36 posterior placement. About 70.40% VPC patients, 17.24% VPI patients, and 48.72% MVPI patients demonstrated normal articulation. About 59.20% VPI patients showed consonant omission. The results demonstrated that consonant articulation varied among patients with unoperated SMCP and was correlated with their velopharyngeal functions, while consonant omission was the most common articulation error. Velopharyngeal closure pattern was not associated with either velopharyngeal function or consonant articulation.