Multimodal deep learning for dementia classification using text and audio.

Dementia is a progressive neurological disorder that affects the daily lives of older adults, impacting their verbal communication and cognitive function. Early diagnosis is important to enhance the lifespan and quality of life for affected individuals. Despite its importance, diagnosing dementia is a complex process. Automated machine learning solutions involving multiple types of data have the potential to improve the process of automated dementia screening. In this study, we build deep learning models to classify dementia cases from controls using the Pitt Cookie Theft dataset from DementiaBank, a database of short participant responses to the structured task of describing a picture of a cookie theft. We fine-tune Wav2vec and Word2vec baseline models to make binary predictions of dementia from audio recordings and text transcripts, respectively. We conduct experiments with four versions of the dataset: (1) the original data, (2) the data with short sentences removed, (3) text-based augmentation of the original data, and (4) text-based augmentation of the data with short sentences removed. Our results indicate that synonym-based text data augmentation generally enhances the performance of models that incorporate the text modality. Without data augmentation, models using the text modality achieve around 60% accuracy and 70% AUROC scores, and with data augmentation, the models achieve around 80% accuracy and 90% AUROC scores. We do not observe significant improvements in performance with the addition of audio or timestamp information into the model. We include a qualitative error analysis of the sentences that are misclassified under each study condition. This study provides preliminary insights into the effects of both text-based data augmentation and multimodal deep learning for automated dementia classification.

Effects of polyallylamine-coated nanoparticles on the optical and photochemical properties of rose bengal.

BACKGROUND: Inasmuch as optical and photochemical properties of a photosensitizer can be modified upon association with the nanoparticle (NP), we wondered whether the effectiveness of phototherapeutic rose bengal (RB) was affected upon tethering to the sodium lanthanide fluoride NP with an outer polyallylamine (PAH) coat. METHODS: RB molecules were electrostatically bound to the NaYF 4 :Gd 3+ :Nd 3+ NPs with inner silica and outer PAH coats. The products were analyzed for their size, shape and zeta potential using transmission electron microscopy and dynamic light scattering instrument. Ultraviolet-visible absorption spectrometry and fluorescence spectrometry were used to examine the spectral properties. Photodynamic effect in terms of singlet oxygen generation was quantitatively determined using the indicator 1,3-diphenylisobenzofuran (DPBF). Photocytotoxicity mediated by NP-bound RB was tested using A549 cells (Student's t test was used for statistical evaluation). RESULTS: NP-bound RB had the major absorbance peak at 561 nm, in comparison with 549 nm for free RB, accompanied with a significant decrease in absorptivity. The molar extinction coefficient becomes 36 000 M -1 cm -1 , only ~35% of that for free RB. Fluorescence spectral analyses showed a paradoxical decrease in the emission with higher NP concentrations even at very low dilutions. Most importantly, the association of RB with these NPs drastically increased its singlet oxygen production upon irradiation. The interaction of RB with PAH coat could partly account for this enhancement, given our finding that PAH in solution also caused a drastic rise in DPBF reactivity by free RB. These NPs exhibited strong photocytotoxic effects, and their promise in photodynamic therapy was addressed. CONCLUSION: Our findings provide evidence that the PAH coat plays a key role in enhanced biological activities of RB delivered via NPs, including the increase in singlet oxygen production and photocytotoxic effects.

Identification of the receptor of oncolytic virus M1 as a therapeutic predictor for multiple solid tumors.

Over the last decade, oncolytic virus (OV) therapy has shown its promising potential in tumor treatment. The fact that not every patient can benefit from it highlights the importance for defining biomarkers that help predict patients' responses. As particular self-amplifying biotherapeutics, the anti-tumor effects of OVs are highly dependent on the host factors for viral infection and replication. By using weighted gene co-expression network analysis (WGCNA), we found matrix remodeling associated 8 (MXRA8) is positively correlated with the oncolysis induced by oncolytic virus M1 (OVM). Consistently, MXRA8 promotes the oncolytic efficacy of OVM in vitro and in vivo. Moreover, the interaction of MXRA8 and OVM studied by single-particle cryo-electron microscopy (cryo-EM) showed that MXRA8 directly binds to this virus. Therefore, MXRA8 acts as the entry receptor of OVM. Pan-cancer analysis showed that MXRA8 is abundant in most solid tumors and is highly expressed in tumor tissues compared with adjacent normal ones. Further study in cancer cell lines and patient-derived tumor tissues revealed that the tumor selectivity of OVM is predominantly determined by a combinational effect of the cell membrane receptor MXRA8 and the intracellular factor, zinc-finger antiviral protein (ZAP). Taken together, our study may provide a novel dual-biomarker for precision medicine in OVM therapy.

Tumors driven by RAS signaling harbor a natural vulnerability to oncolytic virus M1.

Oncolytic viruses are potent anticancer agents that replicate within and kill cancer cells rather than normal cells, and their selectivity is largely determined by oncogenic mutations. M1, a novel oncolytic virus strain, has been shown to target cancer cells, but the relationship between its cancer selectivity and oncogenic signaling pathways is poorly understood. Here, we report that RAS mutation promotes the replication and oncolytic effect of M1 in cancer, and we further provide evidence that the inhibition of the RAS/RAF/MEK signaling axis suppresses M1 infection and the subsequent cytopathic effects. Transcriptome analysis revealed that the inhibition of RAS signaling upregulates the type I interferon antiviral response, and further RNA interference screen identified CDKN1A as a key downstream factor that inhibits viral infection. Gain- and loss-of-function experiments confirmed that CDKN1A inhibited the replication and oncolytic effect of M1 virus. Subsequent TCGA data mining and tissue microarray (TMA) analysis revealed that CDKN1A is commonly deficient in human cancers, suggesting extensive clinical application prospects for M1. Our report indicates that virotherapy is feasible for treating undruggable RAS-driven cancers and provides reliable biomarkers for personalized cancer therapy.

Endless Pursuit: Evolution of Insulin Administration.

Diabetes mellitus is a chronic metabolic health condition affecting the steady state of blood sugar level. The usual method of administration is subcutaneous injection of insulin. There are several ways to subcutaneously inject insulin, such as syringes, insulin pens, and insulin pumps. However, subcutaneous injections of insulin can lead to discomfort, pain and local infection. This review focuses on traditional methods of insulin administration, non-invasive approaches, and new insulin therapy technologies, and the advantages and disadvantages of these approaches, as well as future development prospects are also discussed.