Deep Learning Based Open Set Acoustic Scene Classification

Late fusion framework for Acoustic Scene Classification using LPCC, SCMC, and log-Mel band energies with Deep Neural Networks

Common acoustic properties that different classes share degrades the performance of acoustic scene classification systems. This results in a phenomenon where a few confusing pairs of acoustic scenes dominate a significant proportion of all misclassified audio segments. In this article, we propose adopting a knowledge distillation framework that trains deep neural networks using soft labels. Soft labels, extracted from another pre-trained deep neural network, are used to reflect the similarity between different classes that share similar acoustic properties. We also propose utilizing specialist models to provide additional soft labels. Each specialist model in this study refers to a deep neural network that concentrates on discriminating a single pair of acoustic scenes that are frequently misclassified. Self multi-head attention is explored for training specialist deep neural networks to further concentrate on target pairs of classes. The goal of this article is to train a single deep neural network that demonstrates performance equivalent to, or higher than, an ensemble of multiple models, by distilling the knowledge from several models. Diverse experiments conducted using the detection and classification of acoustic scenes and events 2019 task 1-a dataset demonstrate that the knowledge distillation framework is effective in acoustic scene classification. Specialist models successfully decrease the number of misclassified audio segments in the target classes. The final single model with the proposed method that is trained by the proposed knowledge distillation from several models, including specialists trained using an attention mechanism, shows a classification accuracy of 77.63 %, higher than an ensemble of the baseline and multiple specialists.

Most of the research on acoustic scene classification (ASC) focuses on classification problem with only known scene classes. In practice, scene classification problem to be solved generally is based on an open set, which contains unknown scenes. This paper proposes a two-stage method that solves the open set problem on ASC. The proposed system decomposes open set ASC problem into two stages. To mitigate the impact of unknown scenes on the subsequent recognition process of known scenes, the first stage is to identify unknown scenes. The second stage classifies defined acoustic scenes. In this case, the threshold selection strategy we proposed further sorts out unknown scenes that were not identified in the previous stage. Experiments show that the method proposed in this paper can effectively identify unknown scenes and classify known scenes, by segmenting the open set acoustic scene classification task and selecting an appropriate judgment threshold. On the development dataset released by DCASE Challenge 2019 Task 1C, the model proposed outperforms the first place.

Outcome prediction by serum calprotectin in patients with COVID-19 in the emergency department

BackgroundFibroScan® Expert 630 is a novel fibroscan device equipped with SmartExam software, which allows continuous measurement of controlled attenuation parameter (CAP) during the entire examination and automatic measurement depths adaption...

Background: Sepsis continues to impose a major healthcare burden globally. Early identification and prompt initiation of appropriate therapy, particularly antibiotics, are critical determinants of survival. Various bedside scoring systems have been developed for early detection and prognostication of sepsis. Objective: This chapter compares the predictive accuracy of Systemic Inflammatory Response Syndrome (SIRS), quick Sequential Organ Failure Assessment (qSOFA), National Early Warning Score (NEWS), and Modified Early Warning Score (MEWS) for in-hospital mortality among patients with suspected sepsis admitted to the intensive care unit (ICU). Methods: This prospective comparative study was conducted in the Department of General Medicine and Critical Care at Dhiraj General Hospital, affiliated with Sumandeep Vidyapeeth University, Waghodia, Gujarat, India. A total of 188 adult patients with suspected sepsis admitted to the ICU were included. Clinical and laboratory parameters required to calculate SIRS, qSOFA, NEWS, and MEWS were recorded at admission. The Modified EWS (MEWS) score comprises five physiologic variables: systolic blood pressure (SBP), HR, RR, temperature, and mental status. The MEWS score considers the relative deviation from a patient’s normal blood pressure and urine output to identify surgical patients who would potentially benefit from intensive care. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUROC) were calculated for each scoring system in predicting ICU mortality. Hierarchical Summary Receiver Operating Characteristic (HSROC) curve analysis was performed to compare overall prognostic accuracy. Results: Among 188 patients, 42 (22.34%) died during hospitalisation. qSOFA ≥2 demonstrated the highest specificity (73.61%) but the lowest sensitivity (36.02%). SIRS ≥2 and NEWS ≥5 showed the highest sensitivity (77.78%), although SIRS had the lowest specificity (23.88%). MEWS ≥5 demonstrated balanced performance with a sensitivity of 76.36% and specificity of 63.91%. The highest discriminatory ability for ICU mortality was observed with MEWS ≥5 (AUROC 0.76; 95% CI 0.68–0.84), compared to NEWS ≥5 (AUROC 0.61), qSOFA ≥2 (AUROC 0.56), and SIRS ≥2 (AUROC 0.49). HSROC analysis confirmed the superior overall prognostic performance of MEWS. The better predictive accuracy of NEWS is most likely due to the fact that it incorporates a larger number of physiological parameters than qSOFA and SIRS, including the majority of their constituent atoms. Any scoring system for sepsis should favour better sensitivity over specificity because the consequences of false-negative results (delayed or missed therapy) are arguably much higher than those of false-positive results (unnecessary antibiotics). The AUROC for SIRS, however, was so low that it appears to have very little clinical use. Conclusion: Among the four scoring systems evaluated, MEWS demonstrated the highest predictive accuracy for ICU mortality. However, none of the scoring systems achieved both high sensitivity and high specificity. These tools should therefore complement, rather than replace, comprehensive clinical assessment.

Актуальность. Новые возможности улучшения функции внешнего дыхания (ФВД) связаны с применением высокочастотной осцилляции грудной клетки (ВЧОГК), которая позволяет восстановить дренаж бронхиального дерева и оптимизировать легочную вентиляцию. Цель исследования — изучение и оценка ФВД у детей с внебольничной пневмонией (ВП). Материалы и методы. Проведено обследование 107 детей (основная группа (ОГ) — 55 человек и контрольная группа (КГ) — 52 человека) в возрасте 6–17 лет с ВП с острым и неосложненным течением. Дети ОГ получали базисную терапию (БТ) с проведением процедур ВЧОГК, тогда как дети КГ получали исключительно БТ. Исследовали ФВД в динамике лечения, применяя спирометрию. Результаты. В начале терапии ВП показатели ФВД в исследуемых группах имели несущественные различия. При анализе параметров ФВД у детей ОГ на 10-й день БТ отмечено повышение объема форсированного выдоха за 1-ю секунду (ОФВ1) (88,36 ± 1,55 %, p = 0,02), жизненной емкости легких (ЖЕЛ) (88,18 ± 1,53 %, p = 0,02), форсированной жизненной емкости легких (ФЖЕЛ) (86,77 ± 1,37 %, p = 0,03), максимальной объемной скорости воздуха на уровне выдоха 25 % (МОШ25) (90,10 ± 2,99 %, p = 0,02) и максимальной вентиляции легких (МВЛ) (88,31 ± 1,70 %, p = 0,04). Сравнивая показатели ФВД детей КГ, выявили менее выраженную динамику, в частности, ОФВ1 (81,65 ± 2,44 %), ЖЕЛ (82,95 ± 2,56), ФЖЕЛ (80,85 ± 2,09), МОШ25 (82,63 ± 3,08 %) и МВЛ (85,65 ± 1,99 %). При проведении ROC-анализа динамика восстановления ФВД у детей ОГ более заметна за счет улучшения МВЛ, о чем свидетельствует самая большая площадь под кривой ROC (ППК) — 0,99, ОФВ1 — ППК 0,94 и ОФВ1/ФЖЕЛ — ППК 0,94. У детей КГ была менее выражена динамика исследуемых параметров ФВД, что подтверждается МВЛ — ППК 0,63, ОФВ1 — ППК 0,79 и ОФВ1/ФЖЕЛ — ППК 0,89. Выводы. Для улучшения вентиляционной функции легких у детей с ВП необходимо проводить процедуры бронходренажа, используя ВЧОГК в составе БТ, о чем свидетельствуют полученные положительные результаты спирометрии.

Acoustic scene classification contains frequently misclassified pairs of classes that share many common acoustic properties. Specific details can provide vital clues for distinguishing such pairs of classes. However, these details are generally not noticeable and are hard to generalize for different data distributions. In this study, we investigate various methods for capturing discriminative information and simultaneously improve the generalization ability. We adopt a max feature map method that replaces conventional non-linear activation functions in deep neural networks; therefore, we apply an element-wise comparison between the different filters of a convolution layer’s output. Two data augmentation methods and two deep architecture modules are further explored to reduce overfitting and sustain the system’s discriminative power. Various experiments are conducted using the “detection and classification of acoustic scenes and events 2020 task1-a” dataset to validate the proposed methods. Our results show that the proposed system consistently outperforms the baseline, where the proposed system demonstrates an accuracy of 70.4% compared to the baseline at 65.1%.

Hierarchical classification for acoustic scenes using deep learning

Restriction spectrum imaging (RSI), as an advanced quantitative diffusion-weighted magnetic resonance imaging technique, has the potential to distinguish primary benign and malignant lung lesions. To explore how well the tri-compartmental RSIperforms in distinguishing primary benign from malignant lung lesions compared with diffusion-weighted imaging (DWI), and to further explore whether positron emission tomography/magnetic resonance imaging (PET/MRI) can improve diagnostic efficacy. Prospective. 137 patients, including 108 malignant and 29 benign lesions (85 males, 52 females; average age = 60.0 ± 10.0 years). T2WI, T1WI, multi-b value DWI, MR-based attenuation correction, and PET imaging on a 3.0 T whole-body PET/MR system. The apparent diffusion coefficient (ADC), RSI-derived parameters (restricted diffusion , hindered diffusion , and free diffusion ) and the maximum standardized uptake value (SUVmax) were calculated and analyzed for diagnostic efficacy individually or in combination. Student's t-test, Mann-Whitney U test, receiver operating characteristic (ROC) curves, Delong test, Spearman's correlation analysis. P < 0.05 was considered statistically significant. The , SUVmax were significantly higher, and , ADC were significantly lower in the malignant group [0.717 ± 0.131, 9.125 (5.753, 13.058), 0.194 ± 0.099, 1.240 (0.972, 1.407)] compared to the benign group [0.504 ± 0.236, 3.390 (1.673, 6.030), 0.398 ± 0.195, 1.485 ± 0.382]. The area under the ROC curve (AUC) values ranked from highest to lowest as follows: AUC (SUVmax) > AUC ( ) > AUC ( ) > AUC (ADC) > AUC ( ) (AUC = 0.819, 0.811, 0.770, 0.745, 0549). The AUC (AUC = 0.900) of the combined model of RSI with PET was significantly higher than that of either single-modality imaging. RSI-derived parameters ( , ) might help to distinguish primary benign and malignant lung lesions and the discriminatory utility of was not observed. The RSI exhibits comparable or potentially enhanced performance compared with DWI, and the combined RSI and PET model might improve diagnostic efficacy. Stage 2.

BackgroundThermal imaging (TI) is a portable, low cost imaging tool with high feasibility for use. Clinical joint assessment.Is routinely performed in rheumatoid arthritis (RA) patient care.ObjectivesTo assess a combined TI...

Circulating miRNAs previously associated with androgen excess in women might be used as diagnostic biomarkers for polycystic ovary syndrome (PCOS). Models based on circulating miR-142-3p and miR-598-3p expression show good discrimination among women with and without PCOS, particularly when coupled with easily available measurements such as waist-to-hip ratio (WHR) and circulating LH-to-FSH (LH/FSH) ratios. The lack of standardization of the signs, methods, and threshold values used to establish the presence of the diagnostic criteria (hyperandrogenism, ovulatory dysfunction, and polycystic ovarian morphology) complicates the diagnosis of PCOS. Certain biomarkers may help with such a diagnosis. We conducted a validation study to check the diagnostic accuracy for PCOS of several miRNAs that were associated with the syndrome in a small pilot study that had been previously carried out by our research group. This was a diagnostic test study involving 140 premenopausal women. We included 71 women with PCOS and 69 healthy control women in the study. Both groups were selected as to be similar in terms of body mass index. We used miRCURY LNA™ Universal RT microRNA PCR to analyse the five miRNAs that had shown the strongest associations with PCOS in a much smaller pilot study previously conducted by our group. We studied diagnostic accuracy using receiver operating characteristics (ROC) curve analysis. Only the expression of two miRNAs, miR-142-3p and miR-598-3p, of the five studied, was different between the women with PCOS and the non-hyperandrogenic controls. The diagnostic accuracy of the combination of these circulating miRNAs was good (area under the ROC curve (AUC) 0.801; 95% CI: 0.72-0.88) and was further improved when adding WHR (AUC 0.834, 95% CI: 0.756-0.912), LH/FSH ratio (AUC = 0.869, 95% CI: 0.804-0.934) or both (AUC = 0.895, 95% CI: 0.835-0.954). We developed several models by selecting different threshold values for these variables favouring either sensitivity or specificity, with positive and negative predictive values as high as 88% or 85%, respectively. Patients included here had the classic PCOS phenotype, consisting of hyperandrogenism and ovulatory dysfunction; hence, the present results might not apply to milder phenotypes lacking androgen excess. If confirmed in larger studies addressing different populations and PCOS phenotypes, these biomarkers may be useful to simplify the clinical diagnosis of this prevalent syndrome. This research was funded by Instituto de Salud Carlos III, Spanish Ministry of Science and Innovation (grants PI15/01686, PIE16/00050, PI18/01122 & PI21/00116) and co-funded by European Regional Development Fund 'A way to make Europe'. Centro de Investigación Biomédica en Red Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM) and Instituto Ramón y Cajal de Investigación Sanitaria (IRYCIS) are also initiatives of the Instituto de Salud Carlos III. The authors have no competing interests to disclose. N/A.

To determine the diagnostic value of preoperative inflammatory biomarkers and CA199, alone or in combination, in diagnosing pancreatic cancer (PCC). This retrospective study was comprised of 75 PCC patients and 83 healthy controls (HC). The participant's medical data was mined from the electronic records of the First Affiliated Hospital of Guangxi Medical University. The data included the preoperative circulating albumin/fibrinogen ratio (AFR), the platelet/lymphocyte ratio (PLR), the lymphocyte/monocyte ratio (LMR), the neutrophil/lymphocyte ratio (NLR), and the derived NLR (dNLR). The receiver operating characteristic (ROC) curve and the area under the ROC curve (AUROC) were used to evaluate the diagnostic efficacy of these candidate biomarkers for PCC. A single AFR significantly distinguished PCC from the healthy controls (AUROC: 0.903, 95% CI: 0.846 - 0.945) and had a significantly higher sensitivity and larger AUROC than CA199 (AUROC: 0.814, 95% CI: 0.774 - 0.871). The combinations of AFR with CA199 (AUROC: 0.932, 95% CI: 0.881 - 0.966), RDW with CA199 (AUROC: 0.905, 95% CI: 0.849 - 0.946), Alb with CA199 (AUROC: 0.869, 95% CI: 0.806 - 0.917), and Fib with CA199 (AUROC: 0.921, 95% CI: 0.868 - 0.958) also yielded higher sensitivities and larger AUROCs than CA199 alone. Circulating AFR was an effective biomarker in diagnosing PCC. Combining AFR, RDW, Alb, and Fib with CA199 could improve the diagnostic efficacy for PCC.

Objective To explore the value of quantitative T2 weighted magnetic resonance imaging (T2WI) and diffusion tensor imaging (DTI) parameters for predicting functional outcomes after surgery for cervical spondylotic myelopathy (CSM). Methods One hundred and forty CSM patients received T2WI and DTI before surgery at Tianjin Hospital between April 2014 and April 2016. They were then given systematic rehabilitation treatment after the surgery. The Japanese Orthopaedics Association (JOA) scoring system was applied to evaluate their neurological function before and after the surgery. According to the JOA recovery rate at 1-year follow-up, the patients were divided into a good recovery group (recovery rate≥50%) and a poor recovery group (recovery rate <50%). Maximum spinal cord compression (MSCC), signal change ratio (SCR), transverse area (TA), apparent diffusion coefficient (ADC) and fractional anisotropy (FA) were compared between the two groups. Receiver operating characteristics (ROC) curves were used to measure the prognostic ability and determine the best cut-off value for each variable. The independent predictors of a poor recovery were estimated using univariate and multivariate analysis. Results ROC analysis showed that the area under the ROC curve (AUC) of MSCC was 0.593. For SCR it was 0.682. For the TA it was 0.706. For the ADC it was 0.719 and for FA it was 0.749. The respective cut-off values were 44% compression, 1.561, 0.46 cm2, 1.339×10-3 mm2/s and 0.386. FA had the largest AUC, followed by ADC, TA, SCR and MSCC. MSCC and SCR had low discrimination power (AUC<0.7) in predicting a poor recovery, whereas TA, ADC and FA had moderate discrimination power (AUC 0.7-0.9). Logistic multivariate regression showed that a low JOA score, TA≤0.46 cm2 or FA≤0.386 were independent risk factors for a poor recovery. A predicting model built according to the results of the logistic regression analysis gave an AUC of 0.87, significantly better than that of the JOA score. With a cut-off value of 0.36, the sensitivity and specificity were 80% and 77% respectively. Conclusions Combining T2WI and DTI parameters with the JOA score may better predict the recovery of patients with CSM. The values can also provide references for making up rehabilitation plans. Key words: T2 weighted magnetic resonance imaging; Diffusion tensor imaging; Cervical vertebrae; Spinal cord compression; Prognosis

In Acoustic Scene Classification (ASC) two major approaches have been followed. While one utilizes engineered features such as mel-frequency-cepstral-coefficients (MFCCs), the other uses learned features that are the outcome of an optimization algorithm. I-vectors are the result of a modeling technique that usually takes engineered features as input. It has been shown that standard MFCCs extracted from monaural audio signals lead to i-vectors that exhibit poor performance, especially on indoor acoustic scenes. At the same time, Convolutional Neural Networks (CNNs) are well known for their ability to learn features by optimizing their filters. They have been applied on ASC and have shown promising results. In this paper, we first propose a novel multi-channel i-vector extraction and scoring scheme for ASC, improving their performance on indoor and outdoor scenes. Second, we propose a CNN architecture that achieves promising ASC results. Further, we show that i-vectors and CNNs capture complementary information from acoustic scenes. Finally, we propose a hybrid system for ASC using multi-channel i-vectors and CNNs by utilizing a score fusion technique. Using our method, we participated in the ASC task of the DCASE-2016 challenge. Our hybrid approach achieved 1st rank among 49 submissions, substantially improving the previous state of the art.