Improving generalization performance of electrocardiogram classification models.

Objective.Recently, many electrocardiogram (ECG) classification algorithms using deep learning have been proposed. Because the ECG characteristics vary across datasets owing to variations in factors such as recorded hospitals and the race of participants, the model needs to have a consistently high generalization performance across datasets. In this study, as part of the PhysioNet/Computing in Cardiology Challenge (PhysioNet Challenge) 2021, we present a model to classify cardiac abnormalities from the 12- and the reduced-lead ECGs.Approach.To improve the generalization performance of our earlier proposed model, we adopted a practical suite of techniques, i.e. constant-weighted cross-entropy loss, additional features, mixup augmentation, squeeze/excitation block, and OneCycle learning rate scheduler. We evaluated its generalization performance using the leave-one-dataset-out cross-validation setting. Furthermore, we demonstrate that the knowledge distillation from the 12-lead and large-teacher models improved the performance of the reduced-lead and small-student models.Main results.With the proposed model, our DSAIL SNU team has received Challenge scores of 0.55, 0.58, 0.58, 0.57, and 0.57 (ranked 2nd, 1st, 1st, 2nd, and 2nd of 39 teams) for the 12-, 6-, 4-, 3-, and 2-lead versions of the hidden test set, respectively.Significance.The proposed model achieved a higher generalization performance over six different hidden test datasets than the one we submitted to the PhysioNet Challenge 2020.

Deep contrastive learning of molecular conformation for efficient property prediction.

Data-driven deep learning algorithms provide accurate prediction of high-level quantum-chemical molecular properties. However, their inputs must be constrained to the same quantum-chemical level of geometric relaxation as the training dataset, limiting their flexibility. Adopting alternative cost-effective conformation generative methods introduces domain-shift problems, deteriorating prediction accuracy. Here we propose a deep contrastive learning-based domain-adaptation method called Local Atomic environment Contrastive Learning (LACL). LACL learns to alleviate the disparities in distribution between the two geometric conformations by comparing different conformation-generation methods. We found that LACL forms a domain-agnostic latent space that encapsulates the semantics of an atom's local atomic environment. LACL achieves quantum-chemical accuracy while circumventing the geometric relaxation bottleneck and could enable future application scenarios such as inverse molecular engineering and large-scale screening. Our approach is also generalizable from small organic molecules to long chains of biological and pharmacological molecules.

Evaluating the Image Quality of Neck Structures Scanned on Chest CT with Low-Concentration-Iodine Contrast Media.

BACKGROUND: The purpose of this study was to investigate and compare the image quality of low-concentration-iodine (240 mgI/mL) contrast media (CM) and high-concentration-iodine (320 mgI/mL) CM according to the radiation dose. METHODS: A total of 366 CT examinations were examined. Based on an assessment of quantitative and qualitative parameters by two radiologists, the quality was compared between Group A (low-concentration-iodine CM) and Group B (high-concentration-iodine CM) images of thyroid gland, sternocleidomastoid muscle (SCM), internal jugular vein (IJV), and common carotid artery (CCA). Another subgroup analysis compared Group a, (using ≤90 kVp in Group A), and Group b, (using ≥100 kVp in Group B) for finding the difference in image quality when the tube voltage is lowered. RESULTS: Image quality did not differ between Groups A and B or between Groups a and b. The signal-to-noise ratio and contrast-to-noise ratio were significantly higher for Group B than Group A for the thyroid gland, IJV, and CCA. No statistical differences were found in the comparison of all structures between Groups a and b. CONCLUSION: There was no significant difference in image quality based on CM concentration with variable radiation doses. Therefore, if an appropriate CT protocol is applied, clinically feasible neck CT images can be obtained even using low-concentration-iodine CM.

Protein transfer learning improves identification of heat shock protein families.

Heat shock proteins (HSPs) play a pivotal role as molecular chaperones against unfavorable conditions. Although HSPs are of great importance, their computational identification remains a significant challenge. Previous studies have two major limitations. First, they relied heavily on amino acid composition features, which inevitably limited their prediction performance. Second, their prediction performance was overestimated because of the independent two-stage evaluations and train-test data redundancy. To overcome these limitations, we introduce two novel deep learning algorithms: (1) time-efficient DeepHSP and (2) high-performance DeeperHSP. We propose a convolutional neural network (CNN)-based DeepHSP that classifies both non-HSPs and six HSP families simultaneously. It outperforms state-of-the-art algorithms, despite taking 14-15 times less time for both training and inference. We further improve the performance of DeepHSP by taking advantage of protein transfer learning. While DeepHSP is trained on raw protein sequences, DeeperHSP is trained on top of pre-trained protein representations. Therefore, DeeperHSP remarkably outperforms state-of-the-art algorithms increasing F1 scores in both cross-validation and independent test experiments by 20% and 10%, respectively. We envision that the proposed algorithms can provide a proteome-wide prediction of HSPs and help in various downstream analyses for pathology and clinical research.

Kaleidoscopic fluorescent arrays for machine-learning-based point-of-care chemical sensing.

Multiplexed analysis allows simultaneous measurements of multiple targets, improving the detection sensitivity and accuracy. However, highly multiplexed analysis has been challenging for point-of-care (POC) sensing, which requires a simple, portable, robust, and affordable detection system. In this work, we developed paper-based POC sensing arrays consisting of kaleidoscopic fluorescent compounds. Using an indolizine structure as a fluorescent core skeleton, named Kaleidolizine (KIz), a library of 75 different fluorescent KIz derivatives were designed and synthesized. These KIz derivatives are simultaneously excited by a single ultraviolet (UV) light source and emit diverse fluorescence colors and intensities. For multiplexed POC sensing system, fluorescent compounds array on cellulose paper was prepared and the pattern of fluorescence changes of KIz on array were specific to target chemicals adsorbed on that paper. Furthermore, we developed a machine-learning algorithm for automated, rapid analysis of color and intensity changes of individual sensing arrays. We showed that the paper sensor arrays could differentiate 35 different volatile organic compounds using a smartphone-based handheld detection system. Powered by the custom-developed machine-learning algorithm, we achieved the detection accuracy of 97% in the VOC detection. The highly multiplexed paper sensor could have favorable applications for monitoring a broad-range of environmental toxins, heavy metals, explosives, pathogens.

Full Color Tunable Aggregation-Induced Emission Luminogen for Bioimaging Based on an Indolizine Molecular Framework.

By taking advantage of a unique mechanism of aggregation-induced emission (AIE) phenomena, AIE luminogens (AIEgens) have been provided as a solution to overcome the limitations of conventional fluorophores bearing the feature of aggregation-caused quenching (ACQ) phenomena. Especially, AIEgens paved the way to develop fluorogenic probes ideal for fluorescent imaging in live cell conditions. Despite the high demand for discovery of new AIEgens, it is still challenging to find a versatile molecular platform to generate diverse AIEgens. Herein, we report a new colorful molecular framework, Kaleidolizine (KIz), as a molecular platform for AIEgen generation. The KIz system allows systematic tuning of the emission wavelength from 455 to 564 nm via perturbation of the electron density of substituents on the indolizine core. Increasing the water fraction of the KIz solution in the THF/water mixture induces the fluorescence intensity increase up to 120-fold. Crystal structure analysis, computational calculations, and solvatochromism studies suggest that a synergistic effect between the intramolecular charge transfer and restriction of intramolecular rotation acts as the AIE mechanism in the KIz system. Conjugation of the triphenylphosphonium moiety to KIz allows successful development of triphenylphosphonium (TPP)-KIz for real-time bioimaging of innate mitochondria in live cells, thereby revealing the potential of KIz as a versatile molecular platform to generate fluorogenic probes based on AIE phenomena. We do believe the KIz system could serve as a new, reliable, and generally applicable molecular platform to develop various AIEgens having desired photophysical properties along with an excellent signal-to-noise ratio and with experimental convenience especially for fluorogenic live cell imaging.

The comparative study of resonance disorders for Vietnamese and Korean cleft palate speakers using nasometer.

BACKGROUND: Nasalance is used to evaluate the velopharyngeal incompetence in clinical diagnoses using a nasometer. The aim of this study is to find the nasalance differences between Vietnamese cleft palate children and Korean cleft palate children by measuring the nasalance of five oral vowels. METHODS: Ten Vietnamese cleft palate children after surgery, three Vietnamese children for the control group, and ten Korean cleft palate children after surgery with the same age participated in this experimentation. Instead of Korean control, the standard value of Korean version of the simplified nasometric assessment procedures (kSNAP) was used. RESULT: The results are as follows: (1) the highest nasalance score among the Vietnamese normal vowels is the low vowel /a/; however, that of Korean normal vowels is the high vowel /i/. (2) The average nasalance score of Korean cleft palate vowels is 18% higher than that of Vietnamese cleft palate vowels. There was a nasalance score of over 45% among the vowels /e/ and /i/ in Vietnamese cleft palate patients and /i/, /o/, and /u/ in Korean cleft palate patients. CONCLUSION: These different nasalance scores of the same vowels seem to cause an ethnic difference between Vietnamese and Korean cleft palate children.

The Effects of Bolus Consistency in Pharyngeal Transit Duration during Normal Swallowing.

OBJECTIVE: To examine the effects of age, gender and bolus consistency in normal populations using the temporal measurement of Pharyngeal Transit Duration (PTD), which reflects the duration of bolus flow from the ramus of the mandible to the upper esophageal sphincter. METHOD: 40 normal and healthy subjects had Videofluoroscopic Swallowing Examinations (VFSEs) of 5 ml thin and nectar thick liquids, and puree consistencies. A slow motion and frame by frame analysis was performed. Three-way analysis of variance (ANOVA) was performed to examine the main effect and interactions, and paired t-tests for the three consistency comparisons. RESULTS: Older subjects had a significantly longer PTD than younger subjects (p<0.01). In addition, men had significantly shorter PTDs than women (p<0.01). Puree showed a significantly longer PTD than the other two consistencies, regardless of age and gender (p<0.05). CONCLUSION: PTD is an indicative of motor weakness in pharyngeal swallowing secondary to aging. In addition, the results supported the assumption that there is a functional difference in pharyngeal swallowing between men and women. It is expected that the results of this study will be used for further investigation of patients with dysphagia.

Highly sensitive electrochemical detection of proteins using aptamer-coated gold nanoparticles and surface enzyme reactions.

A novel electrochemical detection methodology is described for the femtomolar detection of proteins which utilizes both DNA aptamer-functionalized nanoparticles and a surface enzymatic reaction. Immunoglobulin E (IgE) was used as a model protein biomarker, which possesses two distinct epitopes for antibody (anti-IgE) and DNA aptamer binding. A surface sandwich assay format was utilized involving the specific adsorption of IgE onto a gold electrode surface that was pre-modified with a monolayer of aptamer-nanoparticle conjugates followed by the specific interaction of alkaline phosphatase (ALP) conjugated anti-IgE. To clearly demonstrate the signal enhancement associated with nanoparticle use, anodic current measurements of the ALP catalyzed oxidation of the enzyme substrate 4-aminophenylphosphate (APP) were also compared with electrode surfaces upon which the aptamer was directly attached. The detection of an unlabelled protein at concentrations as low as 5 fM is a significant improvement compared to conventional electrochemical-based immunoassay approaches and provides a foundation for the practical use and incorporation of nanoparticle-enhanced detection into electrochemical biosensing technologies.

Solitary extrahepatic intraabdominal metastasis from hepatocellular carcinoma after liver transplantation.

A liver transplantation is a treatment option in selected patients with hepatocellular carcinoma (HCC). Despite the adequate selection of candidates, recurrences of HCC may still develop. Solitary extrahepatic metastasis from HCC after a liver transplantation is rare. Here we report two cases of HCC demonstrated extrahepatic recurrence to the adrenal gland and spleen, respectively, within one year after a liver transplantation. Since the treatment of solitary extrahepatic metastasis from HCC after a liver transplantation is not standardized, surgical resection was performed. In the case of HCC adrenal metastasis, innumerable intrahepatic metastases were found two months after the adrenalectomy. And 16 months after adrenalectomy, the patient expired due to tumor progression and hepatic failure. In the case of HCC splenic metastasis, postoperative radiation therapy was performed. However, two recurrent HCC nodules were found 15 months after the splenectomy and received transarterial chemoembolization (TACE). And 29 month after the splenectomy, the patient also expired as same causes of former patient.

Voice onset time variations of korean stop consonants in aphasic speakers.

OBJECTIVE: To compare the voice onset time (VOT) differences of Korean stops in the initial and intervocalic positions between the aphasic patients with peculiarities of aspiration and a control group. METHOD: We examined 15 aphasic patients (nine males, six females) who had suffered a stroke (average age 49.7 years) and 15 healthy controls (average age 47.4 years). An aphasia examination was made by an aphasia battery of three standard tests and VOT was analyzed instrumentally. Stop consonants in the initial and intervocalic position were measured to categorize them according to aphasia types, place of articulation, and manner of articulation. RESULTS: VOT of the aphasic patients with peculiarities of aspiration had a greater difference than that of the controls, indicating that the temporal non-coordination between the laryngeal adjustment and oral articulators of aphasic patients happens due to the VOT of stops in the initial and intervocalic positions (p<0.05). CONCLUSION: VOT of stop consonants in the initial position produced by aphasic patients tends to be proportional to their breathing. It can cause glottal width and make aphasic patients' VOT duration longer. Lastly, the method to measure the VOT of aphasic patients is more significant for the types of phonation than for the places of articulation, and makes it possible to induce abnormal VOT.

Unipolar and continuous cortical stimulation to enhance motor and language deficit in patients with chronic stroke: report of 2 cases.

BACKGROUND: Stroke often leads to permanent neurological deficit despite extensive therapeutic and rehabilitative efforts; and in chronic stages, it is difficult to restore neurological function. Recently, bipolar and intermittent cortical stimulation combined with rehabilitation has been reported to be effective in chronic hemiparetic stroke. However, optimal parameters of cortical stimulation for chronic stroke have not been determined to restore neurological function in chronic stages of stroke. CASE DESCRIPTIONS: We report 2 cases of chronic stroke whose neurological functions were improved by continuous, unipolar cortical stimulation combined with rehabilitation. Patient 1 presented motor and language deficit associated with cortical infarct in the left middle cerebral artery territory, which occurred 18 months before admission. Patient 2 presented motor deficit associated with subcortical infarct in the internal capsule 8 months before admission. Unipolar electrodes were implanted epidurally to cover the contralateral premotor and motor cortex in both cases, and an additional electrode was implanted over the Broca area in patient 1. Follow-up evaluation after 4 months of stimulation and concurrent rehabilitation showed improvement of Fugl-Meyer and FIM scores in both patients, and of speech and articulation in patient 1, whose electrode covered the Broca area. No adverse effects were observed during the period of treatment. CONCLUSION: Unipolar and continuous cortical stimulation with rehabilitation is beneficial for improving neurological deficit in selected cases of chronic stroke.

Acoustic effects of carbamazepine in benign rolandic epilepsy.

Subtle specific interference with language function has been suspected in benign rolandic epilepsy (BRE) because of the location of the seizure focus. The aim of this study was to investigate the language and speech problems in BRE and the changes after medication with carbamazepine. Benign rolandic epilepsy patients (n=11) who met ILAE criteria were assessed for standardized full articulation. Speech problems were present in BRE patients and worsened after medication. The error pattern of laryngeal articulation in BRE was exclusively the substitution of stop consonants, and these errors increased after medication (6% vs 13%). In this study, the variation in acoustic parameters after carbamazepine was similar to the hypokinetic pattern of articulation abnormalities. Our findings suggest that BRE patients treated with carbamazepine should be monitored for language and speech problems.

Reorganization of cortical language areas in patients with aphasia: a functional MRI study.

The purpose of this study is to delineate the pattern of reorganization of cortical language areas using functional magnetic resonance imaging (fMRI) after rehabilitation therapy in patients with aphasia. Six right-handed aphasic patients were investigated. Causes of aphasia were intracerebral hemorrhages of the left basal ganglia in 3 patients, cerebral infarction of the left MCA in 2, and surgical resection of the frontotemporal lobes to control intractable epilepsy in 1. An auditory sentence completion task was used to activate brain language areas during the fMRI. Three patients with left frontal lesions showed activation in the right inferior frontal lobes while performing language tasks, whereas the other 3, whose lesions located at subcortical areas, showed activation in the bilateral frontal and temporal lobes. Our results demonstrated the differences in interhemispheric reorganization of the language network depending on the location of the lesion in aphasic patients. While the patients with subcortical lesion showed tendency of bilateral frontal activation, those with cortical lesion showed activation of the right frontal lobe.