Enhancing pediatric pneumonia diagnosis through masked autoencoders.

Pneumonia, an inflammatory lung condition primarily triggered by bacteria, viruses, or fungi, presents distinctive challenges in pediatric cases due to the unique characteristics of the respiratory system and the potential for rapid deterioration. Timely diagnosis is crucial, particularly in children under 5, who have immature immune systems, making them more susceptible to pneumonia. While chest X-rays are indispensable for diagnosis, challenges arise from subtle radiographic findings, varied clinical presentations, and the subjectivity of interpretations, especially in pediatric cases. Deep learning, particularly transfer learning, has shown promise in improving pneumonia diagnosis by leveraging large labeled datasets. However, the scarcity of labeled data for pediatric chest X-rays presents a hurdle in effective model training. To address this challenge, we explore the potential of self-supervised learning, focusing on the Masked Autoencoder (MAE). By pretraining the MAE model on adult chest X-ray images and fine-tuning the pretrained model on a pediatric pneumonia chest X-ray dataset, we aim to overcome data scarcity issues and enhance diagnostic accuracy for pediatric pneumonia. The proposed approach demonstrated competitive performance an AUC of 0.996 and an accuracy of 95.89% in distinguishing between normal and pneumonia. Additionally, the approach exhibited high AUC values (normal: 0.997, bacterial pneumonia: 0.983, viral pneumonia: 0.956) and an accuracy of 93.86% in classifying normal, bacterial pneumonia, and viral pneumonia. This study also investigated the impact of different masking ratios during pretraining and explored the labeled data efficiency of the MAE model, presenting enhanced diagnostic capabilities for pediatric pneumonia.

Bimodal CNN for cardiovascular disease classification by co-training ECG grayscale images and scalograms.

This study aimed to develop a bimodal convolutional neural network (CNN) by co-training grayscale images and scalograms of ECG for cardiovascular disease classification. The bimodal CNN model was developed using a 12-lead ECG database collected from Chapman University and Shaoxing People's Hospital. The preprocessed database contains 10,588 ECG data and 11 heart rhythms labeled by a specialist physician. The preprocessed one-dimensional ECG signals were converted into two-dimensional grayscale images and scalograms, which are fed simultaneously to the bimodal CNN model as dual input images. The proposed model aims to improve the performance of CVDs classification by making use of ECG grayscale images and scalograms. The bimodal CNN model consists of two identical Inception-v3 backbone models, which were pre-trained on the ImageNet database. The proposed model was fine-tuned with 6780 dual-input images, validated with 1694 dual-input images, and tested on 2114 dual-input images. The bimodal CNN model using two identical Inception-v3 backbones achieved best AUC (0.992), accuracy (95.08%), sensitivity (0.942), precision (0.946) and F1-score (0.944) in lead II. Ensemble model of all leads obtained AUC (0.994), accuracy (95.74%), sensitivity (0.950), precision (0.953), and F1-score (0.952). The bimodal CNN model showed better diagnostic performance than logistic regression, XGBoost, LSTM, single CNN model training with grayscale images alone or with scalograms alone. The proposed bimodal CNN model would be of great help in diagnosing cardiovascular diseases.

Multi-Modal Stacking Ensemble for the Diagnosis of Cardiovascular Diseases.

BACKGROUND: Cardiovascular diseases (CVDs) are a leading cause of death worldwide. Deep learning methods have been widely used in the field of medical image analysis and have shown promising results in the diagnosis of CVDs. METHODS: Experiments were performed on 12-lead electrocardiogram (ECG) databases collected by Chapman University and Shaoxing People's Hospital. The ECG signal of each lead was converted into a scalogram image and an ECG grayscale image and used to fine-tune the pretrained ResNet-50 model of each lead. The ResNet-50 model was used as a base learner for the stacking ensemble method. Logistic regression, support vector machine, random forest, and XGBoost were used as a meta learner by combining the predictions of the base learner. The study introduced a method called multi-modal stacking ensemble, which involves training a meta learner through a stacking ensemble that combines predictions from two modalities: scalogram images and ECG grayscale images. RESULTS: The multi-modal stacking ensemble with a combination of ResNet-50 and logistic regression achieved an AUC of 0.995, an accuracy of 93.97%, a sensitivity of 0.940, a precision of 0.937, and an F1-score of 0.936, which are higher than those of LSTM, BiLSTM, individual base learners, simple averaging ensemble, and single-modal stacking ensemble methods. CONCLUSION: The proposed multi-modal stacking ensemble approach showed effectiveness for diagnosing CVDs.

Diagnostic usefulness of deep learning methods for Helicobacter pylori infection using esophagogastroduodenoscopy images.

Background and Aims: We aimed to assess the diagnostic potential of deep convolutional neural networks (DCNNs) for detecting Helicobacter pylori infection in patients who underwent esophagogastroduodenoscopy and Campylobacter-like organism tests. Methods: We categorized a total of 13,071 images of various gastric sub-areas and employed five pretrained DCNN architectures: ResNet-101, Xception, Inception-v3, InceptionResnet-v2, and DenseNet-201. Additionally, we created an ensemble model by combining the output probabilities of the best models. We used images of different sub-areas of the stomach for training and evaluated the performance of our models. The diagnostic metrics assessed included area under the curve (AUC), specificity, accuracy, positive predictive value, and negative predictive value. Results: When training included images from all sub-areas of the stomach, our ensemble model demonstrated the highest AUC (0.867), with specificity at 78.44%, accuracy at 80.28%, positive predictive value at 82.66%, and negative predictive value at 77.37%. Significant differences were observed in AUC between the ensemble model and the individual DCNN models. When training utilized images from each sub-area separately, the AUC values for the antrum, cardia and fundus, lower body greater curvature and lesser curvature, and upper body greater curvature and lesser curvature regions were 0.842, 0.826, 0.718, and 0.858, respectively, when the ensemble model was used. Conclusions: Our study demonstrates that the DCNN model, designed for automated image analysis, holds promise for the evaluation and diagnosis of Helicobacter pylori infection.

A convolutional deep learning model for improving mammographic breast-microcalcification diagnosis.

This study aimed to assess the diagnostic performance of deep convolutional neural networks (DCNNs) in classifying breast microcalcification in screening mammograms. To this end, 1579 mammographic images were collected retrospectively from patients exhibiting suspicious microcalcification in screening mammograms between July 2007 and December 2019. Five pre-trained DCNN models and an ensemble model were used to classify the microcalcifications as either malignant or benign. Approximately one million images from the ImageNet database had been used to train the five DCNN models. Herein, 1121 mammographic images were used for individual model fine-tuning, 198 for validation, and 260 for testing. Gradient-weighted class activation mapping (Grad-CAM) was used to confirm the validity of the DCNN models in highlighting the microcalcification regions most critical for determining the final class. The ensemble model yielded the best AUC (0.856). The DenseNet-201 model achieved the best sensitivity (82.47%) and negative predictive value (NPV; 86.92%). The ResNet-101 model yielded the best accuracy (81.54%), specificity (91.41%), and positive predictive value (PPV; 81.82%). The high PPV and specificity achieved by the ResNet-101 model, in particular, demonstrated the model effectiveness in microcalcification diagnosis, which, in turn, may considerably help reduce unnecessary biopsies.

Texture analysis using machine learning-based 3-T magnetic resonance imaging for predicting recurrence in breast cancer patients treated with neoadjuvant chemotherapy.

OBJECTIVES: To determine whether texture analysis for magnetic resonance imaging (MRI) can predict recurrence in patients with breast cancer treated with neoadjuvant chemotherapy (NAC). METHODS: This retrospective study included 130 women who received NAC and underwent subsequent surgery for breast cancer between January 2012 and August 2017. We assessed common features, including standard morphologic MRI features and clinicopathologic features. We used a  commercial software and analyzed texture features from pretreatment and midtreatment MRI. A random forest (RF) method was performed to build a model for predicting recurrence. The diagnostic performance of this model for predicting recurrence was assessed and compared with those of five other machine learning classifiers using the Wald test. RESULTS: Of the 130 women, 21 (16.2%) developed recurrence at a median follow-up of 35.4 months. The RF classifier with common features including clinicopathologic and morphologic MRI features showed the lowest diagnostic performance (area under the receiver operating characteristic curve [AUC], 0.83). The texture analysis with the RF method showed the highest diagnostic performances for pretreatment T2-weighted images and midtreatment DWI and ADC maps showed better diagnostic performance than that of an analysis of common features (AUC, 0.94 vs. 0.83, p < 0.05). The RF model based on all sequences showed a better diagnostic performance for predicting recurrence than did the five other machine learning classifiers. CONCLUSIONS: Texture analysis using an RF model for pretreatment and midtreatment MRI may provide valuable prognostic information for predicting recurrence in patients with breast cancer treated with NAC and surgery. KEY POINTS: • RF model-based texture analysis showed a superior diagnostic performance than traditional MRI and clinicopathologic features (AUC, 0.94 vs.0.83, p < 0.05) for predicting recurrence in breast cancer after NAC. • Texture analysis using RF classifier showed the highest diagnostic performances (AUC, 0.94) for pretreatment T2-weighted images and midtreatment DWI and ADC maps. • RF model showed a better diagnostic performance for predicting recurrence than did the five other machine learning classifiers.

Texture Analysis with 3.0-T MRI for Association of Response to Neoadjuvant Chemotherapy in Breast Cancer.

Background Previous studies have suggested that texture analysis is a promising tool in the diagnosis, characterization, and assessment of treatment response in various cancer types. Therefore, application of texture analysis may be helpful for early prediction of pathologic response in breast cancer. Purpose To investigate whether texture analysis of features from MRI is associated with pathologic complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer. Materials and Methods This retrospective study included 136 women (mean age, 47.9 years; range, 31-70 years) who underwent NAC and subsequent surgery for breast cancer between January 2012 and August 2017. Patients were monitored with 3.0-T MRI before (pretreatment) and after (midtreatment) three or four cycles of NAC. Texture analysis was performed at pre- and midtreatment T2-weighted MRI, contrast material-enhanced T1-weighted MRI, diffusion-weighted MRI, and apparent diffusion coefficient (ADC) mapping by using commercial software. A random forest method was applied to build a predictive model for classifying those with pCR with use of texture parameters. Diagnostic performance for predicting pCR was assessed and compared with that of six other machine learning classifiers (adaptive boosting, decision tree, k-nearest neighbor, linear support vector machine, naive Bayes, and linear discriminant analysis) by using the Wald test and DeLong method. Results Forty of the 136 patients (29%) achieved pCR after NAC. In the prediction of pCR, the random forest classifier showed the lowest diagnostic performance with pretreatment ADC (area under the receiver operating characteristic curve [AUC], 0.53; 95% confidence interval: 0.44, 0.61) and the highest diagnostic performance with midtreatment contrast-enhanced T1-weighted MRI (AUC, 0.82; 95% confidence interval: 0.74, 0.88) among pre- and midtreatment T2-weighted MRI, contrast-enhanced T1-weighted MRI, diffusion-weighted MRI, and ADC mapping. Conclusion Texture parameters using a random forest method of contrast-enhanced T1-weighted MRI at midtreatment of neoadjuvant chemotherapy were valuable and associated with pathologic complete response in breast cancer. © RSNA, 2019 Online supplemental material is available for this article.

High efficiency ultraviolet GaN-based vertical light emitting diodes on 6-inch sapphire substrate using ex-situ sputtered AlN nucleation layer.

We demonstrated the growth of crack-free high-quality GaN-based UV vertical LEDs (VLEDs) (λ = 365 nm) on 6-inch sapphire substrates by using an ex-situ sputtered AlN nucleation layer (NL) and compared their performance with that of UV VLEDs with an in situ low temperature (LT) AlGaN NL. The X-ray diffraction (XRD) results showed that the ex-situ AlN sample contained lower densities of screw-type and edge-type threading dislocations than the in situ AlGaN NL sample. The micro-Raman results revealed that the ex-situ AlN sample was under more compressive stress than the in situ AlGaN sample. As the current was increased, the electroluminescence peaks of both of the samples blue-shifted, reached a minimum wavelength at 1000 mA, and then slightly red-shifted. Packaged VLEDs with the ex-situ AlN NL yielded 6.5% higher light output power at 500 mA than that with the in situ AlGaN NL. The maximum EQEs of the VLED with the in situ AlGaN and ex-situ AlN NLs were 43.7% and 48.2%, respectively. Based on the XRD and Raman results, the improved light output power of the ex-situ AlN sample is attributed to the lower density of TDs.

Diffusion radiomics as a diagnostic model for atypical manifestation of primary central nervous system lymphoma: development and multicenter external validation.

Background: Radiomics is a rapidly growing field in neuro-oncology, but studies have been limited to conventional MRI, and external validation is critically lacking. We evaluated technical feasibility, diagnostic performance, and generalizability of a diffusion radiomics model for identifying atypical primary central nervous system lymphoma (PCNSL) mimicking glioblastoma. Methods: A total of 1618 radiomics features were extracted from diffusion and conventional MRI from 112 patients (training set, 70 glioblastomas and 42 PCNSLs). Feature selection and classification were optimized using a machine-learning algorithm. The diagnostic performance was tested in 42 patients of internal and external validation sets. The performance was compared with that of human readers (2 neuroimaging experts), cerebral blood volume (90% histogram cutoff, CBV90), and apparent diffusion coefficient (10% histogram, ADC10) using the area under the receiver operating characteristic curve (AUC). Results: The diffusion radiomics was optimized with the combination of recursive feature elimination and a random forest classifier (AUC 0.983, stability 2.52%). In internal validation, the diffusion model (AUC 0.984) showed similar performance with conventional (AUC 0.968) or combined diffusion and conventional radiomics (AUC 0.984) and better than human readers (AUC 0.825-0.908), CBV90 (AUC 0.905), or ADC10 (AUC 0.787) in atypical PCNSL diagnosis. In external validation, the diffusion radiomics showed robustness (AUC 0.944) and performed better than conventional radiomics (AUC 0.819) and similar to combined radiomics (AUC 0.946) or human readers (AUC 0.896-0.930). Conclusion: The diffusion radiomics model had good generalizability and yielded a better diagnostic performance than conventional radiomics or single advanced MRI in identifying atypical PCNSL mimicking glioblastoma.

Use of a patterned current blocking layer to enhance the light output power of InGaN-based light-emitting diodes.

We employed a patterned current blocking layer (CBL) to enhance light output power of GaN-based light-emitting diodes (LEDs). Nanoimprint lithography (NIL) was used to form patterned CBLs (a diameter of 260 nm, a period of 600, and a height of 180 nm). LEDs (chip size: 300 × 800 µm2) fabricated with no CBL, a conventional SiO2 CBL, and a patterned SiO2 CBL, respectively, exhibited forward-bias voltages of 3.02, 3.1 and 3.1 V at an injection current of 20 mA. The LEDs without and with CBLs gave series resistances of 9.8 and 11.0 Ω, respectively. The LEDs with a patterned SiO2 CBL yielded 39.6 and 11.9% higher light output powers at 20 mA, respectively, than the LEDs with no CBL and conventional SiO2 CBL. On the basis of emission images and angular transmittance results, the patterned CBL-induced output enhancement is attributed to the enhanced light extraction and current spreading.

Reciprocal Interactions between Medial Septum and Hippocampus in Theta Generation: Granger Causality Decomposition of Mixed Spike-Field Recordings.

The medial septum (MS) plays an essential role in rhythmogenesis in the hippocampus (HIPP); theta-rhythmic bursts of MS neurons are believed to drive theta oscillations in rats' HIPP. The MS theta pacemaker hypothesis has solid foundation but the MS-hippocampal interactions during different behavioral states are poorly understood. The MS and the HIPP have reciprocal connections and it is not clear in particular what role, if any, the strong HIPP to MS projection plays in theta generation. To study the functional interactions between MS and HIPP during different behavioral states, this study investigated the relationship between MS single-unit activity and HIPP field potential oscillations during theta states of active waking and REM sleep and non-theta states of slow wave sleep (SWS) and quiet waking (QW), i.e., sleep-wake states that comprise the full behavioral repertoire of undisturbed, freely moving rats. We used non-parametric Granger causality (GC) to decompose the MS-HIPP synchrony into its directional components, MS→HIPP and HIPP→MS, and to examine the causal interactions between them within the theta frequency band. We found a significant unidirectional MS→HIPP influence in non-theta states which switches to bidirectional theta drive during theta states with MS→HIPP and HIPP→MS GC being of equal magnitude. In non-theta states, unidirectional MS→HIPP influence was accompanied by significant MS-HIPP coherence, but no signs of theta oscillations in the HIPP. In theta states of active waking and REM sleep, sharp theta coherence and strong theta power in both structures was associated with a rise in HIPP→MS to the level of the MS→HIPP drive. Thus, striking differences between waking and REM sleep theta states and non-theta states of SWS and QW were primarily observed in activation of theta influence carried by the descending HIPP→MS pathway associated with more regular rhythmic bursts in the MS and sharper MS→HIPP GC spectra without a significant increase in MS→HIPP GC magnitude. The results of this study suggest an essential role of descending HIPP to MS projections in theta generation.

Formation of an indium tin oxide nanodot/Ag nanowire electrode as a current spreader for near ultraviolet AlGaN-based light-emitting diodes.

Indium tin oxide (ITO) nanodots (NDs) were combined with Ag nanowires (Ag NWs) as a p-type electrode in near ultraviolet AlGaN-based light-emitting diodes (LEDs) to increase light output power. The Ag NWs were 30 ± 5 nm in diameter and 25 ± 5 µm in length. The transmittance of 10 nm-thick ITO-only was 98% at 385 nm, while the values for ITO ND/Ag NW were 83%-88%. ITO ND/Ag NW films showed lower sheet resistances (32-51 Ω sq-1) than the ITO-only film (950 Ω sq-1). LEDs (chip size: 300 × 800 µm2) fabricated using the ITO NDs/Ag NW electrodes exhibited higher forward-bias voltages (3.52-3.75 V at 20 mA) than the LEDs with the 10 nm-thick ITO-only electrode (3.5 V). The LEDs with ITO ND/Ag NW electrodes yielded a 24%-62% higher light output power (at 20 mA) than those with the 10 nm-thick ITO-only electrode. Furthermore, finite-difference time-domain (FDTD) simulations were performed to investigate the extraction efficiency. Based on the emission images and FDTD simulations, the enhanced light output with the ITO ND/Ag NW electrodes is attributed to improved current spreading and better extraction efficiency.

Large-scale functional brain connectivity during emotional engagement as revealed by beta-series correlation analysis.

It has been hypothesized that the medial prefrontal cortex (mPFC) is a hub in the network that mediates appetitive responses whereas the amygdala is thought to mediate both aversive and appetitive processing. Both structures may facilitate adaptive responses to emotional challenge by linking perception, attention, memory, and motor circuits. We provide an initial exploration of these hypotheses by recording simultaneous EEG-fMRI in eleven participants viewing affective pictures. MPFC- and amygdala-seeded functional connectivity maps were generated by applying the beta-series correlation method. The mPFC-seeded correlation map encompassed visual regions, sensorimotor areas, prefrontal cortex, and medial temporal lobe structures, exclusively for pleasant content. For the amygdala-seeded correlation map, a similar set of distributed brain areas appeared in the unpleasant-neutral contrast, with the addition of structures such as the insula and thalamus. A substantially sparser network was recruited for the pleasant-neutral contrast. Using the late positive potential (LPP) to index the intensity of emotional engagement, functional connectivity was found to be stronger in trials with larger LPP. These results demonstrate that mPFC-mediated functional interactions are engaged specifically during appetitive processing, whereas the amygdala is coupled to distinct sets of brain regions during both aversive and appetitive processing. The strength of these interactions varies as a function of the intensity of emotional engagement.

Assessing Granger Causality in Electrophysiological Data: Removing the Adverse Effects of Common Signals via Bipolar Derivations.

Multielectrode voltage data are usually recorded against a common reference. Such data are frequently used without further treatment to assess patterns of functional connectivity between neuronal populations and between brain areas. It is important to note from the outset that such an approach is valid only when the reference electrode is nearly electrically silent. In practice, however, the reference electrode is generally not electrically silent, thereby adding a common signal to the recorded data. Volume conduction further complicates the problem. In this study we demonstrate the adverse effects of common signals on the estimation of Granger causality, which is a statistical measure used to infer synaptic transmission and information flow in neural circuits from multielectrode data. We further test the hypothesis that the problem can be overcome by utilizing bipolar derivations where the difference between two nearby electrodes is taken and treated as a representation of local neural activity. Simulated data generated by a neuronal network model where the connectivity pattern is known were considered first. This was followed by analyzing data from three experimental preparations where a priori predictions regarding the patterns of causal interactions can be made: (1) laminar recordings from the hippocampus of an anesthetized rat during theta rhythm, (2) laminar recordings from V4 of an awake-behaving macaque monkey during alpha rhythm, and (3) ECoG recordings from electrode arrays implanted in the middle temporal lobe and prefrontal cortex of an epilepsy patient during fixation. For both simulation and experimental analysis the results show that bipolar derivations yield the expected connectivity patterns whereas the untreated data (referred to as unipolar signals) do not. In addition, current source density signals, where applicable, yield results that are close to the expected connectivity patterns, whereas the commonly practiced average re-reference method leads to erroneous results.

SVDD-based pattern denoising.

The support vector data description (SVDD) is one of the best-known one-class support vector learning methods, in which one tries the strategy of using balls defined on the feature space in order to distinguish a set of normal data from all other possible abnormal objects. The major concern of this letter is to extend the main idea of SVDD to pattern denoising. Combining the geodesic projection to the spherical decision boundary resulting from the SVDD, together with solving the preimage problem, we propose a new method for pattern denoising. We first solve SVDD for the training data and then for each noisy test pattern, obtain its denoised feature by moving its feature vector along the geodesic on the manifold to the nearest decision boundary of the SVDD ball. Finally we find the location of the denoised pattern by obtaining the pre-image of the denoised feature. The applicability of the proposed method is illustrated by a number of toy and real-world data sets.