Automatic Speech Classifier for Mild Cognitive Impairment and Early Dementia

The World Health Organization estimates that 50 million people are currently living with dementia worldwide and this figure will almost triple by 2050. Current pharmacological treatments are only symptomatic, and drugs or other therapies are ineffective in slowing down or curing the neurodegenerative process at the basis of dementia. Therefore, early detection of cognitive decline is of the utmost importance to respond significantly and deliver preventive interventions. Recently, the researchers showed that speech alterations might be one of the earliest signs of cognitive defect, observable well in advance before other cognitive deficits become manifest. In this article, we propose a full automated method able to classify the audio file of the subjects according to the progress level of the pathology. In particular, we trained a specific type of artificial neural network, called autoencoder, using the visual representation of the audio signal of the subjects, that is, the spectrogram. Moreover, we used a data augmentation approach to overcome the problem of the large amount of annotated data usually required during the training phase, which represents one of the most major obstacles in deep learning. We evaluated the proposed method using a dataset of 288 audio files from 96 subjects: 48 healthy controls and 48 cognitively impaired participants. The proposed method obtained good classification results compared to the state-of-the-art neuropsychological screening tests and, with an accuracy of 90.57%, outperformed the methods based on manual transcription and annotation of speech.

Development and Validation of an Analytical HPLC Method to Assess Chemical and Radiochemical Purity of [68Ga]Ga-NODAGA-Exendin-4 Produced by a Fully Automated Method

Background: Glucagon-like peptide 1 receptor (GLP-1R) is preferentially expressed in pancreatic islets, especially in β-cells, and highly expressed in human insulinomas and gastrinomas. In recent years several GLP-1R–avid radioligands have been developed to image insulin-secreting tumors or to provide a tentative quantitative in vivo biomarker of pancreatic β-cell mass. Exendin-4, a 39-amino acid peptide with high binding affinity to GLP-1R, has been labeled with Ga-68 for imaging with positron emission tomography (PET). Preparation conditions may influence the quality and in vivo behavior of tracers. Starting from a published synthesis and quality controls (QCs) procedure, we have developed and validated a new rapid and simple UV-Radio-HPLC method to test the chemical and radiochemical purity of [68Ga]Ga-NODAGA-exendin-4, to be used in the clinical routine. Methods: Ga-68 was obtained from a 68Ge/68Ga Generator (GalliaPharma®) and purified using a cationic-exchange cartridge on an automated synthesis module (Scintomics GRP®). NODAGA-exendin-4 contained in the reactor (10 µg) was reconstituted with HEPES and ascorbic acid. The reaction mixture was incubated at 100 °C. The product was purified through HLB cartridge, diluted, and sterilized. To validate the proposed UV-Radio-HPLC method, a stepwise approach was used, as defined in the guidance document released by the International Conference on Harmonization of Technical Requirements of Pharmaceuticals for Human Use (ICH), adopted by the European Medicines Agency (CMP/ICH/381/95 2014). The assessed parameters are specificity, linearity, precision (repeatability), accuracy, and limit of quantification. Therefore, a range of concentrations of Ga-NODAGA-exendin-4, NODAGA-exendin-4 (5, 4, 3.125, 1.25, 1, and 0.75 μg/mL) and [68Ga]Ga-NODAGA-exendin-4 were analyzed. To validate the entire production process, three consecutive batches of [68Ga]Ga-NODAGA-exendin-4 were tested. Results: Excellent linearity was found between 5–0.75 μg/mL for both the analytes (NODAGA-exendin-4 and 68Ga-NODAGA-exendin-4), with a correlation coefficient (R2) for calibration curves equal to 0.999, average coefficients of variation (CV%) <2% (0.45% and 0.39%) and average per cent deviation value of bias from 100%, of 0.06% and 0.04%, respectively. The calibration curve for the determination of [68Ga]Ga-NODAGA-exendin-4 was linear with a R2 of 0.993 and CV% <2% (1.97%), in accordance to acceptance criteria. The intra-day and inter-day precision of our method was statistically confirmed using 10 μg of peptide. The mean radiochemical yield was 45 ± 2.4% in all the three validation batches of [68Ga]Ga-NODAGA-exendin-4. The radiochemical purity of [68Ga]Ga-NODAGA-exendin-4 was >95% (97.05%, 95.75% and 96.15%) in all the three batches. Conclusions: The developed UV-Radio-HPLC method to assess the radiochemical and chemical purity of [68Ga]Ga-NODAGA-exendin-4 is rapid, accurate and reproducible like its fully automated production. It allows the routine use of this PET tracer as a diagnostic tool for PET imaging of GLP-1R expression in vivo, ensuring patient safety.

Leveraging Publicly Available Information to Analyze Information Operations

Traditionally, a significant part of assessing information operations (IO) relies on subject matter experts’ time- intensive study of publicly available information (PAI). Now, with massive amounts PAI made available via the Internet, analysts are faced with the challenge of effectively leveraging massive quantities of PAI to draw meaningful conclusions. This paper presents an automated method for collecting and analyzing large amounts of PAI from China that could better inform assessments of IO campaigns. We implement a multi-model system that involves data acquisition via web scraping and analysis using natural language processing (NLP) techniques with a focus on topic modeling and sentiment analysis. After conducting a case study on China’s current relationship with Taiwan and comparing the results to validated research by a subject matter expert, it is clear that our methodology is valuable for drawing general conclusions and pinpointing important dialogue over a massive amount of PAI.

Automated light-induced synthesis of 89Zr-radiolabeled antibodies for immuno-positron emission tomography

Clinical production of 89Zr-radiolabeled antibodies (89Zr-mAbs) for positron emission tomography imaging relies on the pre-conjugation of desferrioxamine B (DFO) to the purified protein, followed by isolation and characterization of the functionalized intermediate, and then manual radiosynthesis. Although highly successful, this route exposes radiochemists to a potentially large radiation dose and entails several technological and economic hurdles that limit access of 89Zr-mAbs to just a specialist few Nuclear Medicine facilities worldwide. Here, we introduce a fully automated synthesis box that can produce individual doses of 89Zr-mAbs formulated in sterile solution in < 25 min starting from [89Zr(C2O4)4]4– (89Zr-oxalate), our good laboratory practice-compliant photoactivatable desferrioxamine-based chelate (DFO-PEG3-ArN3), and clinical-grade antibodies without the need for pre-purification of protein. The automated steps include neutralization of the 89Zr-oxalate stock, chelate radiolabeling, and light-induced protein conjugation, followed by 89Zr-mAb purification, formulation, and sterile filtration. As proof-of-principle, 89ZrDFO-PEG3-azepin-trastuzumab was synthesized directly from Herceptin in < 25 min with an overall decay-corrected radiochemical yield of 20.1 ± 2.4% (n = 3), a radiochemical purity > 99%, and chemical purity > 99%. The synthesis unit can also produce 89Zr-mAbs via the conventional radiolabeling routes from pre-functionalized DFO-mAbs that are currently used in the clinic. This automated method will improve access to state-of-the-art 89Zr-mAbs at the many Nuclear Medicine and research institutions that require automated devices for radiotracer production.

Automated histopathological evaluation of pterygium using artificial intelligence

PurposeThis study aimed to evaluate the efficacy of a new automated method for the evaluation of histopathological images of pterygium using artificial intelligence.MethodsAn in-house software for automated grading of histopathological images was developed. Histopathological images of pterygium (400 images from 40 patients) were analysed using our newly developed software. Manual grading (I–IV), labelled based on an established scoring system, served as the ground truth for training the four-grade classification models. Region of interest segmentation was performed before the classification of grades, which was achieved by the combination of expectation-maximisation and k-nearest neighbours. Fifty-five radiomic features extracted from each image were analysed with feature selection methods to examine the significant features. Five classifiers were evaluated for their ability to predict quantitative grading.ResultsAmong the classifier models applied for automated grading in this study, the bagging tree showed the best performance, with a 75.9% true positive rate (TPR) and 75.8% positive predictive value (PPV) in internal validation. In external validation, the method also demonstrated reproducibility, with an 81.3% TPR and 82.0% PPV for the average of four classification grades.ConclusionsOur newly developed automated method for quantitative grading of histopathological images of pterygium may be a reliable method for quantitative analysis of histopathological evaluation of pterygium.

FunOrder 2.0 – a fully automated method for the identification of co-evolved genes

Coevolution is an important biological process that shapes interacting species or even proteins – may it be physically interacting proteins or consecutive enzymes in a metabolic pathway. The detection of co-evolved proteins will contribute to a better understanding of biological systems. Previously, we developed a semi-automated method, termed FunOrder, for the detection of co-evolved genes from an input gene or protein set. We demonstrated the usability and applicability of FunOrder by identifying essential genes in biosynthetic gene clusters from different ascomycetes. A major drawback of this original method was the need for a manual assessment, which may create a user bias and prevents a high-throughput application. Here we present a fully automated version of this method termed FunOrder 2.0. To fully automatize the method, we used several mathematical indices to determine the optimal number of clusters in the FunOrder output, and a subsequent k-means clustering based on the first three principal components of a principal component analysis of the FunOrder output. Further, we replaced the BLAST with the DIAMOND tool, which enhanced speed and allows the future integration of larger proteome databases. The introduced changes slightly decreased the sensitivity of this method, which is outweighed by enhanced overall speed and specificity. Additionally, the changes lay the foundation for future high-throughput applications of FunOrder 2.0 in different phyla to solve different biological problems.

Automated Segmentation and Connectivity Analysis for Normal Pressure Hydrocephalus

Objective and Impact Statement. We propose an automated method of predicting Normal Pressure Hydrocephalus (NPH) from CT scans. A deep convolutional network segments regions of interest from the scans. These regions are then combined with MRI information to predict NPH. To our knowledge, this is the first method which automatically predicts NPH from CT scans and incorporates diffusion tractography information for prediction. Introduction. Due to their low cost and high versatility, CT scans are often used in NPH diagnosis. No well-defined and effective protocol currently exists for analysis of CT scans for NPH. Evans’ index, an approximation of the ventricle to brain volume using one 2D image slice, has been proposed but is not robust. The proposed approach is an effective way to quantify regions of interest and offers a computational method for predicting NPH. Methods. We propose a novel method to predict NPH by combining regions of interest segmented from CT scans with connectome data to compute features which capture the impact of enlarged ventricles by excluding fiber tracts passing through these regions. The segmentation and network features are used to train a model for NPH prediction. Results. Our method outperforms the current state-of-the-art by 9 precision points and 29 recall points. Our segmentation model outperforms the current state-of-the-art in segmenting the ventricle, gray-white matter, and subarachnoid space in CT scans. Conclusion. Our experimental results demonstrate that fast and accurate volumetric segmentation of CT brain scans can help improve the NPH diagnosis process, and network properties can increase NPH prediction accuracy.

Diagnostic utility of FGF-23 in mineral bone disorder during chronic kidney disease

Our data confirm that intact fibroblast growth factor 23 (iFGF-23) concentration is increased in patients with chronic kidney disease (CKD) and that it increases with disease progression (stages I-V). Therefore, iFGF-23 could be considered an early biomarker in the course of chronic kidney disease-mineral bone disorder (CKD-MBD), which has several aspects that make it potentially useful in clinical practice. The availability of an automated method for iFGF-23 assay may represent an added value in the management of the patient with CKD-MBD already from the early stages of the disease, before the increase of the routinely used laboratory parameters, 1-84 parathyroid hormone (PTH) and 25-OH-vitamin D (25-OH-vitD), which occur in more advanced stages of the disease.

Methods for determining color characteristics of vegetable raw materials. A review

Food product quality defines a complex of food product properties such size, shape, texture, color and others, and determines acceptability of these products for consumers. It is possible to detect defects in plant raw materials by color and classify them by color characteristics, texture, shape, a degree of maturity and so on. Currently, the work on modernization of color control systems has been carried out for rapid and objective measuring information about color of plant raw materials during their harvesting, processing and storage. The aim of the work is to analyze existing methods for determining color characteristics of plant raw materials described in foreign and domestic studies. Also, this paper presents the results of the experimental studies that describe the practical use of methods for measuring food product color. At present, the following methods for determining color characteristics by the sensor analysis principle are used: sensory, spectrophotometric and photometric. These methods have several disadvantages. Therefore, computer vision has found wide application as an automated method for food control. It is distinguished by high confidence and reliability in the process of determining freshness, safety, a degree of maturity and other parameters of plant raw materials that are heterogeneous in terms of the abovementioned indicators. The computer vision method is realized in the following systems: conventional, hyperspectral and multispectral. Each subsequent system is a component of the preceding one. Materials presented in the paper allow making a conclusion about the effectiveness of the computer vision systems with the aim of automatic sorting and determining quality of plant raw materials in the food industry.