20260 research outputs found
Sort by
A ResNet-LSTM hybrid model for predicting epileptic seizures using a pretrained model with supervised contrastive learning
Get PDFIn this paper, we propose a method for predicting epileptic seizures using a pre-trained model utilizing supervised contrastive learning and a hybrid model combining residual networks (ResNet) and long short-term memory (LSTM). The proposed training approach encompasses three key phases: pre-processing, pre-training as a pretext task, and training as a downstream task. In the pre-processing phase, the data is transformed into a spectrogram image using short time Fourier transform (STFT), which extracts both time and frequency information. This step compensates for the inherent complexity and irregularity of electroencephalography (EEG) data, which often hampers effective data analysis. During the pre-training phase, augmented data is generated from the original dataset using techniques such as band-stop filtering and temporal cutout. Subsequently, a ResNet model is pre-trained alongside a supervised contrastive loss model, learning the representation of the spectrogram image. In the training phase, a hybrid model is constructed by combining ResNet, initialized with weight values from the pre-trained model, and LSTM. This hybrid model extracts image features and time information to enhance prediction accuracy. The proposed method’s effectiveness is validated using datasets from CHB-MIT and Seoul National University Hospital (SNUH). The method’s generalization ability is confirmed through Leave-one-out cross-validation. From the experimental results measuring accuracy, sensitivity, and false positive rate (FPR), CHB-MIT was 91.90%, 89.64%, 0.058 and SNUH was 83.37%, 79.89%, and 0.131. The experimental results demonstrate that the proposed method outperforms the conventional methods
The LKB1–TSSK1B axis controls YAP phosphorylation to regulate the Hippo–YAP pathway
Get PDFThe Hippo pathway’s main effector, Yes-associated protein (YAP), plays a crucial role in tumorigenesis as a transcriptional coactivator. YAP’s phosphorylation by core upstream components of the Hippo pathway, such as mammalian Ste20 kinase 1/2 (MST1/2), mitogen-activated protein kinase kinase kinase kinases (MAP4Ks), and their substrate, large tumor suppressor 1/2 (LATS1/2), influences YAP’s subcellular localization, stability, and transcriptional activity. However, recent research suggests the existence of alternative pathways that phosphorylate YAP, independent of these core upstream Hippo pathway components, raising questions about additional means to inactivate YAP. In this study, we present evidence demonstrating that TSSK1B, a calcium/calmodulin-dependent protein kinase (CAMK) superfamily member, is a negative regulator of YAP, suppressing cellular proliferation and oncogenic transformation. Mechanistically, TSSK1B inhibits YAP through two distinct pathways. Firstly, the LKB1–TSSK1B axis directly phosphorylates YAP at Ser94, inhibiting the YAP–TEAD complex’s formation and suppressing its target genes’ expression. Secondly, the TSSK1B–LATS1/2 axis inhibits YAP via phosphorylation at Ser127. Our findings reveal the involvement of TSSK1B-mediated molecular mechanisms in the Hippo–YAP pathway, emphasizing the importance of multilevel regulation in critical cellular decision-making processes
Mechanism underlying and prevention of electrode migration in cochlear implants
No full textPurpose: We investigate the clinical manifestations, mechanisms, and methods of preventing electrode migration in Cochlear Implantation (CI) patients, based on our practical experience with this problem. Study design: This is a retrospective study in a single center. Methods: We retrospectively reviewed electrode migration in 4 (0.75%) of 532 patients who underwent CI at our tertiary institution from January 2002 to December 2022. Pre- and post-operative pure-tone audiometry, word recognition score, aided functional gain test, and sound field speech intelligibility test were evaluated. Results: All four patients underwent CIs with the straight electrode type. The following events or symptoms were observed in the patients before confirming electrode migration: an increase in high-frequency thresholds during the post-operative aided functional gain test and a decline in scores on the sound field speech intelligibility test. Electrode migration was confirmed through transocular view X-ray or temporal bone computer tomography. Two patients showed coiled electrodes within the mastoid cavity; while in the others, the electrodes were observed to be floating inside the cavity. To prevent migration of electrodes due to these issues, we mixed bone paste collected during the drilling of the mastoid cavity with glue and used it to secure the electrodes in place. Conclusion: Electrode migration can result in a decrease in hearing ability and may necessitate a revision surgery to adjust the electrode placement. The main factors affecting electrode placement include the position of electrode within the mastoid cavity and the elasticity of straight electrodes. It is important for surgeons to recognize the factors that increase the risk of electrode migration and to take preventative measures to reduce this risk
A De-Identification Model for Korean Clinical Notes: Using Deep Learning Models
Get PDFTo extract information from free-text in clinical records due to the patient's protected health information PHI in the records pre-processing of de-identification is required. Therefore we aimed to identify PHI list and fine-tune the deep learning BERT model for developing de-identification model. The result of fine-tuning the model is strict F1 score of 0.924. Due to the convinced score the model can be used for the development of a de-identification model
Variation in the Allergenicity of Scrambled, Boiled, Short-Baked and Long-Baked Egg White Proteins
Get PDFBACKGROUND: Hen's egg white (HEW) is the most common cause of food allergy in children which induces mild to fatal reactions. The consultation for a proper restriction is important in HEW allergy. We aimed to identify the changes in HEW allergenicity using diverse cooking methods commonly used in Korean dishes. METHODS: Crude extract of raw and 4 types of cooked HEW extracts were produced and used for sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), enzyme-linked immunosorbent assay (ELISA), and ELISA inhibition assays using 45 serum samples from HEW allergic and tolerant children. Extracts were prepared; scrambled without oil for 20-30 seconds in frying pan without oil, boiled at 100°C for 15 minutes, short-baked at 180°C for 20 minutes, and long-baked at 45°C for 12 hours with a gradual increase in temperature up to 110°C for additional 12 hours, respectively. RESULTS: In SDS-PAGE, the intensity of bands of 50-54 kDa decreased by boiling and baking. All bands almost disappeared in long-baked eggs. The intensity of the ovalbumin (OVA) immunoglobulin E (IgE) bands did not change after scrambling; however, an evident decrease was observed in boiled egg white (EW). In contrast, ovomucoid (OM) IgE bands were darker and wider after scrambling and boiling. The IgE binding reactivity to all EW allergens were weakened in short-baked EW and considerably diminished in long-baked EW. In individual ELISA analysis using OVA+OM+ serum samples, the median of specific IgE optical density values was 0.435 in raw EW, 0.476 in scrambled EW, and 0.487 in boiled EW. Conversely, it was significantly decreased in short-baked (0.406) and long-baked EW (0.012). Significant inhibition was observed by four inhibitors such as raw, scrambled, boiled and short-baked HEW, but there was no significant inhibition by long-baked HEW (IC50 > 100 mg/mL). CONCLUSION: We identified minimally reduced allergenicity in scrambled EW and extensively decreased allergenicity in long-baked EW comparing to boiled and short-baked EW as well as raw EW. By applying the results of this study, we would be able to provide safer dietary guidence with higher quality to egg allergic children
Novel Textbook Outcomes following emergency laparotomy: Delphi exercise
Get PDFBackground: Textbook outcomes are composite outcome measures that reflect the ideal overall experience for patients. There are many of these in the elective surgery literature but no textbook outcomes have been proposed for patients following emergency laparotomy. The aim was to achieve international consensus amongst experts and patients for the best Textbook Outcomes for non-trauma and trauma emergency laparotomy. Methods: A modified Delphi exercise was undertaken with three planned rounds to achieve consensus regarding the best Textbook Outcomes based on the category, number and importance (Likert scale of 1–5) of individual outcome measures. There were separate questions for non-trauma and trauma. A patient engagement exercise was undertaken after round 2 to inform the final round. Results: A total of 337 participants from 53 countries participated in all three rounds of the exercise. The final Textbook Outcomes were divided into ‘early’ and ‘longer-term’. For non-trauma patients the proposed early Textbook Outcome was ‘Discharged from hospital without serious postoperative complications (Clavien–Dindo ≥ grade III; including intra-abdominal sepsis, organ failure, unplanned re-operation or death). For trauma patients it was ‘Discharged from hospital without unexpected transfusion after haemostasis, and no serious postoperative complications (adapted Clavien–Dindo for trauma ≥ grade III; including intra-abdominal sepsis, organ failure, unplanned re-operation on or death)’. The longer-term Textbook Outcome for both non-trauma and trauma was ‘Achieved the early Textbook Outcome, and restoration of baseline quality of life at 1 year’. Conclusion: Early and longer-term Textbook Outcomes have been agreed by an international consensus of experts for non-trauma and trauma emergency laparotomy. These now require clinical validation with patient data
Longitudinal immune kinetics of COVID-19 booster versus primary series vaccination: Insight into the annual vaccination strategy
Get PDFBackground: Data on the durability of booster dose immunity of COVID-19 vaccines are relatively limited. Methods: Immunogenicity was evaluated for up to 9–12 months after the third dose of vaccination in 94 healthy adults. Results: Following the third dose, the anti-spike immunoglobulin G (IgG) antibody response against the wild-type was boosted markedly, which decreased gradually over time. However, even 9–12 months after the booster dose, both the median and geometric mean of anti-spike IgG antibody levels were higher than those measured 4 weeks after the second dose. Breakthrough infection during the Omicron-dominant period boosted neutralizing antibody titers against Omicron sublineages (BA.1 and BA.5) and the ancestral strain. T-cell immune response was efficiently induced and maintained during the study period. Conclusions: mRNA vaccine booster dose elicited durable humoral immunity for up to 1 year after the third dose and T-cell immunity was sustained during the study period, supporting an annual COVID-19 vaccination strategy
Area-level deprivation and handwashing behavior during the COVID-19 pandemic: A multilevel analysis on a nationwide survey in Korea
No full textThis study investigated the association between area deprivation level and performance of handwashing behavior during the COVID-19 pandemic in Korean adults. This study used data from the 2015 Population and Housing Census data to measure area deprivation level. The 2020 Korea Community Health Survey was used for all other variables, including hand hygiene behavior (August through November 2020). The association between area deprivation level and practice of handwashing behavior was examined using multilevel logistic regression analysis. The study population comprised 215,676 adults aged 19 years or above. Compared to the least area deprived group, the most deprived group was more likely to not wash hands after using the restroom (OR 1.43, 95% CI 1.13–1.82), after coming home (OR 1.85, 95% CI 1.43–2.39), and using soap (OR 1.55, 95% CI 1.29–1.84). The findings suggest the importance of considering area deprivation in implementing policies that promote handwashing, particularly during a pandemic
Automatic Sleep Stage Classification Based on Deep Learning for Multi-channel Signals
No full textSleep plays a crucial role in restoring physical and mental health, making it essential to monitor sleep patterns objectively. Polysomnography is a standard method used to classify sleep stages, but it tends to be costly and requires specialist involvement. In this respect, many automatic sleep classification algorithms have recently been developed through practical and easy-to-measure wearable devices. However, further research is needed on how to combine biosignals from multichannels with different sampling rates measured through wearable devices. In this study, we proposed a sleep stage classification algorithm in multi-channel signals using an electrocardiogram, accelerometer, and gyroscope. Specifically, convolutional neural networks were used to compare the sleep stage classification performance according to the sampling rate. In the 4-class sleep stage classification for the wake, light sleep, deep sleep, and rapid eye movement, an accuracy of 80.23%, an F1-score of 0.8097, and a kappa value of 0.6711 were achieved when the sampling rate was adjusted based on electrocardiogram. On the other hand, when the sampling rate was based on an accelerometer and gyroscope, the accuracy was 64.33%, the F1-score was 0.6389, and the kappa value was 0.4708 in the 4-class sleep stage classification. These results could provide great insight into developing a sleep stage classification model using multi-channel signals based on wearable devices, and would also be available in other applications such as sleep apnea