Search

PURPOSE: To introduce wave-encoded acquisition and reconstruction techniques for highly accelerated EPI with reduced g-factor penalty and image artifacts.THEORY AND METHODS: Wave-EPI involves application of sinusoidal gradients during the EPI readout, which spreads the aliasing in all spatial directions, thereby taking better advantage of 3D coil sensitivity profiles. The amount of voxel spreading that can be achieved by the wave gradients during the short EPI readout period is constrained by the slew rate of the gradient coils and peripheral nerve stimulation monitor. We propose to use a "half-cycle" sinusoidal gradient to increase the amount of voxel spreading that can be achieved while respecting the slew and stimulation constraints. Extending wave-EPI to multi-shot acquisition minimizes geometric distortion and voxel blurring at high in-plane resolutions, while structured low-rank regularization mitigates shot-to-shot phase variations. To address gradient imperfections, we propose to use different point spread functions for the k-space lines with positive and negative polarities, which are calibrated with a FLEET-based reference scan.RESULTS: Wave-EPI enabled whole-brain single-shot gradient-echo (GE) and multi-shot spin-echo (SE) EPI acquisitions at high acceleration factors at 3T and was combined with g-Slider encoding to boost the SNR level in 1 mm isotropic diffusion imaging. Relative to blipped-CAIPI, wave-EPI reduced average and maximum g-factors by up to 1.21- and 1.37-fold at Rin × Rsms = 3 × 3, respectively.CONCLUSION: Wave-EPI allows highly accelerated single- and multi-shot EPI with reduced g-factor and artifacts and may facilitate clinical and neuroscientific applications of EPI by improving the spatial and temporal resolution in functional and diffusion imaging.

In recent studies, self-supervised pre-trained models tend to outperform supervised pre-trained models in transfer learning. In particular, self-supervised learning (SSL) of utterance-level speech representation can be used in speech applications that require discriminative representation of consistent attributes within an utterance: speaker, language, emotion, and age. Existing frame-level self-supervised speech representation, e.g., wav2vec, can be used as utterance-level representation with pooling, but the models are usually large. There are also SSL techniques to learn utterance-level representation. One of the most successful is a contrastive method, which requires negative sampling: selecting alternative samples to contrast with the current sample (anchor). However, this does not ensure that all the negative samples belong to classes different from the anchor class without labels. This paper applies a non-contrastive self-supervised method to learn utterance-level embeddings. We adapted DIstillation with NO labels (DINO) from computer vision to speech. Unlike contrastive methods, DINO does not require negative sampling. We compared DINO to x-vector trained in a supervised manner. When transferred to down-stream tasks (speaker verification, speech emotion recognition (SER), and Alzheimer's disease detection), DINO outperformed x-vector. We studied the influence of several aspects during transfer learning such as dividing the fine-tuning process into steps, chunk lengths, or augmentation. During fine-tuning, tuning the last affine layers first and then the whole network surpassed fine-tuning all at once. Using shorter chunk lengths, although they generate more diverse inputs, did not necessarily improve performance, implying speech segments at least with a specific length are required for better performance per application. Augmentation was helpful in SER., Comment: EARLY ACCESS of IEEE JSTSP Special Issue on Self-Supervised Learning for Speech and Audio Processing

Text-to-speech (TTS) models trained to minimize the spectrogram reconstruction loss can learn speaker embeddings without explicit speaker identity supervision, unlike x-vector speaker identification (SID) systems. Leveraging this way of speaker embedding learning can be useful in unsupervised or semi-supervised scenarios where non, or only some, of the training data have speaker labels. Thus, in this paper, we evaluate speaker embeddings learned by training the spectrogram prediction network under unsupervised and semi-supervised scenarios. We experimented with different data sampling strategies. The best one was sampling two different segments from the same utterance, namely A and B, where the spectrogram of B is predicted given the B phone sequence and the speaker embedding extracted from A. This method improved by 3.4% relative in EER, compared to using the same utterance for both A and B without segmenting. In the unsupervised scenario, the best speaker embedding outperformed i-vectors, the state-of-the-art unsupervised speaker embedding, in speaker verification by 12.9% relative in EER. We observed high correlation between reconstruction loss and speaker embedding quality. In the semi-supervised scenario, having more unlabeled data in training led to a better performance in speaker verification. Adding 5314 unlabeled speakers to 800 labeled speakers improved EER by 10.8 % relative.

Bone marrow mesenchymal stem cells (BM MSCs) can differentiate into multi-lineage tissues. However, obtaining BM MSCs by aspiration is difficult and can be painful; therefore peripheral blood (PB) MSCs might provide an easier alternative for clinical applications. Here, we show that circulating PB MSCs proliferate as efficiently as BM MSCs in the presence of extracellular matrix (ECM) and that differentiation potential into osteoblast in vitro and in vivo. Both BM MSCs and PB MSCs developed into new bone when subcutaneously transplanted into immune-compromised mice using hydroxyapatite/tricalcium phosphate as a carrier. Furthermore, LY294002 and Wortmannin blocked mesenchymal stem cell attachment in a dose-dependent manner, suggesting a role of phosphatidylinositol 3-kinase in MSC attachment. Our data showed that the growth of PB MSCs could be regulated by interaction with the ECM and that these cells could differentiate into osteoblasts, suggesting their potential for clinical applications.

Endothelin-1 (ET-1) induces contraction, proliferation, and collagen synthesis of activated hepatic stellate cells and is a potent mediator of portal hypertension. Endothelin-converting enzyme-1 (ECE-1) generates ET-1 from the inactive precursor big-endothelin-1. The cellular distribution and activity of ECE-1 in the liver is unknown. Hepatic fibrogenesis was induced in rats by CCl4 administration and secondary biliary cirrhosis after 6 weeks of complete bile duct occlusion (BDO). The tissue ET-1 and ET receptor protein levels were quantified, the ECE-1 isoform mRNAs were measured by RNase protection assay and ECE-1 activity was analyzed. ECE-1a and -b mRNA were upregulated in biliary cirrhosis and in CCl4-injured livers, whereas ECE-1c mRNA remained unchanged. ECE-1 activity was increased after BDO and peaked at 12 h after acute CCl4-intoxication. Tissue levels of ET-1, ETA- and ETB receptors were elevated 7-, 5-, and 4.6-fold in cirrhotic rats, respectively. ECE-1 activity increased following BDO and acute CCl4-intoxication. In conclusion, ECE-1a and -b RNAs are upregulated in fibrogenesis, indicating that these isoforms play a central role in ET-1 generation during fibrogenesis and portal hypertension.

Purpose To develop a scan-specific model that estimates and corrects k-space errors made when reconstructing accelerated MRI data. Methods Scan-specific artifact reduction in k-space (SPARK) trains a convolutional-neural-network to estimate and correct k-space errors made by an input reconstruction technique by back-propagating from the mean-squared-error loss between an auto-calibration signal (ACS) and the input technique's reconstructed ACS. First, SPARK is applied to generalized autocalibrating partially parallel acquisitions (GRAPPA) and demonstrates improved robustness over other scan-specific models, such as robust artificial-neural-networks for k-space interpolation (RAKI) and residual-RAKI. Subsequent experiments demonstrate that SPARK synergizes with residual-RAKI to improve reconstruction performance. SPARK also improves reconstruction quality when applied to advanced acquisition and reconstruction techniques like 2D virtual coil (VC-) GRAPPA, 2D LORAKS, 3D GRAPPA without an integrated ACS region, and 2D/3D wave-encoded imaging. Results SPARK yields SSIM improvement and 1.5 - 2× root mean squared error (RMSE) reduction when applied to GRAPPA and improves robustness to ACS size for various acceleration rates in comparison to other scan-specific techniques. When applied to advanced reconstruction techniques such as residual-RAKI, 2D VC-GRAPPA and LORAKS, SPARK achieves up to 20% RMSE improvement. SPARK with 3D GRAPPA also improves RMSE performance by ~2×, SSIM performance, and perceived image quality without a fully sampled ACS region. Finally, SPARK synergizes with non-Cartesian, 2D and 3D wave-encoding imaging by reducing RMSE between 20% and 25% and providing qualitative improvements. Conclusion SPARK synergizes with physics-based acquisition and reconstruction techniques to improve accelerated MRI by training scan-specific models to estimate and correct reconstruction errors in k-space.

Zero-shot multi-speaker Text-to-Speech (TTS) generates target speaker voices given an input text and the corresponding speaker embedding. In this work, we investigate the effectiveness of the TTS reconstruction objective to improve representation learning for speaker verification. We jointly trained end-to-end Tacotron 2 TTS and speaker embedding networks in a self-supervised fashion. We hypothesize that the embeddings will contain minimal phonetic information since the TTS decoder will obtain that information from the textual input. TTS reconstruction can also be combined with speaker classification to enhance these embeddings further. Once trained, the speaker encoder computes representations for the speaker verification task, while the rest of the TTS blocks are discarded. We investigated training TTS from either manual or ASR-generated transcripts. The latter allows us to train embeddings on datasets without manual transcripts. We compared ASR transcripts and Kaldi phone alignments as TTS inputs, showing that the latter performed better due to their finer resolution. Unsupervised TTS embeddings improved EER by 2.06\% absolute with regard to i-vectors for the LibriTTS dataset. TTS with speaker classification loss improved EER by 0.28\% and 0.73\% absolutely from a model using only speaker classification loss in LibriTTS and Voxceleb1 respectively.

Purpose To develop a new non-contrast-enhanced peripheral MR angiography that provides a high contrast angiogram without using electrocardiography triggering and saturation radiofrequency pulses. Methods A velocity-selective excitation technique is used in conjunction with the golden-angle radial sampling scheme. The signal amplitude varies according to the velocity of the flow by the velocity-selective excitation technique. Because the arterial blood velocity varies depending on the cardiac phase, the acquired data can be classified into systolic and diastolic phase based on the signal amplitude of the artery. Two images are then reconstructed from the systolic and diastolic phase data, respectively, and an image reflecting the differences between the two images is obtained to eliminate background and vein signals. The performance of the proposed method was compared with the quiescent-interval single shot (QISS) in eight healthy subjects and an elderly subject. Results The proposed method generated fewer residual venous and background signals than the QISS. Furthermore, the maximum intensity projection images, the relative contrast, and the apparent contrast-to-noise ratio results showed that the proposed method produced a better contrast than the QISS. Conclusions The proposed non-contrast-enhanced peripheral MR angiography technique can provide a high contrast angiogram without the use of electrocardiography triggering and saturation radiofrequency pulses. Magn Reson Med 79:779–788, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

In this paper, we propose to improve emotion recognition by combining acoustic information and conversation transcripts. On the one hand, an LSTM network was used to detect emotion from acoustic features like f0, shimmer, jitter, MFCC, etc. On the other hand, a multi-resolution CNN was used to detect emotion from word sequences. This CNN consists of several parallel convolutions with different kernel sizes to exploit contextual information at different levels. A temporal pooling layer aggregates the hidden representations of different words into a unique sequence level embedding, from which we computed the emotion posteriors. We optimized a weighted sum of classification and verification losses. The verification loss tries to bring embeddings from the same emotions closer while separating embeddings from different emotions. We also compared our CNN with state-of-the-art text-based hand-crafted features (e-vector). We evaluated our approach on the USC-IEMOCAP dataset as well as the dataset consisting of US English telephone speech. In the former, we used human-annotated transcripts while in the latter, we used ASR transcripts. The results showed fusing audio and transcript information improved unweighted accuracy by relative 24% for IEMOCAP and relative 3.4% for the telephone data compared to a single acoustic system.

Multi-shot echo planar imaging (msEPI) is a promising approach to achieve high in-plane resolution with high sampling efficiency and low T2* blurring. However, due to the geometric distortion, shot-to-shot phase variations and potential subject motion, msEPI continues to be a challenge in MRI. In this work, we introduce acquisition and reconstruction strategies for robust, high-quality msEPI without phase navigators. We propose Blip Up-Down Acquisition (BUDA) using interleaved blip-up and -down phase encoding, and incorporate B0 forward-modeling into Hankel structured low-rank model to enable distortion- and navigator-free msEPI. We improve the acquisition efficiency and reconstruction quality by incorporating simultaneous multi-slice acquisition and virtual-coil reconstruction into the BUDA technique. We further combine BUDA with the novel RF-encoded gSlider acquisition, dubbed BUDA-gSlider, to achieve rapid high isotropic-resolution MRI. Deploying BUDA-gSlider with model-based reconstruction allows for distortion-free whole-brain 1mm isotropic T2 mapping in about 1 minute. It also provides whole-brain 1mm isotropic diffusion imaging with high geometric fidelity and SNR efficiency. We finally incorporate sinusoidal wave gradients during the EPI readout to better use coil sensitivity encoding with controlled aliasing., Comment: 13 pages, 10 figures

In this paper, we explore several new schemes to train a seq2seq model to integrate a pre-trained LM. Our proposed fusion methods focus on the memory cell state and the hidden state in the seq2seq decoder long short-term memory (LSTM), and the memory cell state is updated by the LM unlike the prior studies. This means the memory retained by the main seq2seq would be adjusted by the external LM. These fusion methods have several variants depending on the architecture of this memory cell update and the use of memory cell and hidden states which directly affects the final label inference. We performed the experiments to show the effectiveness of the proposed methods in a mono-lingual ASR setup on the Librispeech corpus and in a transfer learning setup from a multilingual ASR (MLASR) base model to a low-resourced language. In Librispeech, our best model improved WER by 3.7%, 2.4% for test clean, test other relatively to the shallow fusion baseline, with multi-level decoding. In transfer learning from an MLASR base model to the IARPA Babel Swahili model, the best scheme improved the transferred model on eval set by 9.9%, 9.8% in CER, WER relatively to the 2-stage transfer baseline., 4 pages, 1 figure, 5 tables, submitted to ICASSP 2019

Zirconia is worth studying as an alternative to dental titanium implants to overcome the disadvantages of titanium. This study investigated the surface characteristics of the zirconia implant material and osteoclastogenesis responses on the surface compared with titanium. Yttrium oxide-stabilized 5% tetragonal zirconia polycrystalline specimens were manufactured, and osteoclast-precursor cells were cultured and differentiated into osteoclasts on the specimens. Surface shape, roughness, and chemical composition were evaluated. After culturing, cell morphologies and differentiation capacity were analyzed using tartrate-resistant acid phosphatase activity (TRACP). mRNA of two critical transcription factors, nuclear factor of activated T-cells 1 (NFATc1) and c-Fos were measured, and protein levels of NFATc1 and c-Fos were investigated. The zirconia specimens had rhomboid-like shapes with smooth surfaces and exhibited no difference in surface roughness compared to the titanium specimens. Morphologies of differentiated osteoclasts on both materials were similar. TRACP activity on the zirconia showed comparable results to that on the titanium. The mRNA value of NFATc1 on the zirconia was higher than that on the titanium at day four. The protein level of c-Fos was expressed thicker on the zirconia when compared to the titanium at day two. The results of this study suggest that zirconia material provides adequate osteoclastogenesis behaviors for dental implant use.

Toll-like receptors (TLRs) are usually expressed on immune cells such as macrophages, dendritic cells, mast cells, as well as on eosinophils and some epithelial cells. They play a central role in the recognition of harmful molecules that belong to invading microorganisms or internal damaged tissues, which lead to inflammation. Among the hallmarks of cancer, there is immune evasion and inflammation. Summing this with the discovery that a majority of cancers also seem to express TLRs, made researchers realize these receptors might also be linked with cancer progression. This review will cover some of the effects of TLR engagement in cancer cells that might induce the promotion or inhibition of cancer with mechanisms involved. The differences of TLR expression in cancer progression and its possible relation with patient prognosis, TLR genetic disorders found in cancer, and new strategies to cancer therapy will be discussed to target TLRs in cancer cells.

This work explores better adaptation methods to low-resource languages using an external language model (LM) under the framework of transfer learning. We first build a language-independent ASR system in a unified sequence-to-sequence (S2S) architecture with a shared vocabulary among all languages. During adaptation, we perform LM fusion transfer, where an external LM is integrated into the decoder network of the attention-based S2S model in the whole adaptation stage, to effectively incorporate linguistic context of the target language. We also investigate various seed models for transfer learning. Experimental evaluations using the IARPA BABEL data set show that LM fusion transfer improves performances on all target five languages compared with simple transfer learning when the external text data is available. Our final system drastically reduces the performance gap from the hybrid systems., Accepted at ICASSP2019

Canine mesenchymal stem cells (cMSCs) are gaining popularity in the veterinary field as a regenerative therapy. But, their limited culture lifespan makes it an obstacle for preclinical investigation and therapeutic use. In this study, primary canine adipose tissue-derived MSCs (PCAT-MSCs) were isolated from adipose tissue and were transfected with the SV40-T retrovirus resulting in a life-extended immortalized canine adipose tissue-derived MSCs (ICAT-MSCs). A comparison was made through the characterization of both PCAT-MSCs and ICAT-MSCs. Both showed a fibroblastic morphology; ICAT-MSCs showed a higher potential of colony formation compared with PCAT-MSCs and a reduced population doubling time; stem cell markers SOX2 and NANOG were expressed in both cell lines; karyotyping analysis showed no abnormalities in both PCAT-MSCs and ICAT-MSCs; both cell lines were CD90+ , CD44 + , and CD45 - ; both generated chondrogenic pellet; in osteogenic differentiation both showed upregulation of Osterix, a master transcriptome of osteogenesis, but in PCAT-MSCs, an upregulation of SOX2 was also observed. In conclusion, ICAT-MSCs showed similar characteristics with PCAT-MSCs, thus established as an easy to access platform for studies on better understanding about cMSCs nature.

PURPOSE To accurately separate water and fat signals for bipolar multi-echo gradient-recalled echo sequence using a convolutional neural network (CNN). METHODS A CNN architecture was designed and trained using the relationship between multi-echo images from the bipolar multi-echo gradient-recalled echo sequence and artifact-free water-fat-separated images. The artifact-free water-fat-separated images for training the CNN were obtained from multiple signals with different TEs by using iterative decomposition of water and fat with echo asymmetry and the least-squares estimation method, in which multiple signals at different TEs were acquired using a single-echo gradient-recalled echo sequence. We also proposed a data augmentation method using a synthetic field inhomogeneity to generate multi-echo signals, including various bipolar multi-echo gradient-recalled echo artifacts so that the CNN could prevent overfitting and increase the separation accuracy. We trained the CNN using in vivo knee images and tested it using in vivo knee, head, and ankle images. RESULTS In vivo imaging results showed that the proposed CNN could separate water-fat images accurately. Although the proposed CNN was trained using only in vivo knee images, the proposed CNN could also separate water-fat images of different imaging regions. The proposed data augmentation method could prevent overfitting even with a limited number of training data sets and make the method robust to magnetic field inhomogeneities. CONCLUSION The proposed CNN could obtain water-fat-separated images from the multi-echo images acquired from the bipolar multi-echo gradient-recalled echo sequence, which included artifacts from the bipolar gradients.

Sequence-to-sequence (seq2seq) approach for low-resource ASR is a relatively new direction in speech research. The approach benefits by performing model training without using lexicon and alignments. However, this poses a new problem of requiring more data compared to conventional DNN-HMM systems. In this work, we attempt to use data from 10 BABEL languages to build a multi-lingual seq2seq model as a prior model, and then port them towards 4 other BABEL languages using transfer learning approach. We also explore different architectures for improving the prior multilingual seq2seq model. The paper also discusses the effect of integrating a recurrent neural network language model (RNNLM) with a seq2seq model during decoding. Experimental results show that the transfer learning approach from the multilingual model shows substantial gains over monolingual models across all 4 BABEL languages. Incorporating an RNNLM also brings significant improvements in terms of %WER, and achieves recognition performance comparable to the models trained with twice more training data.

This paper investigates the impact of word-based RNN language models (RNN-LMs) on the performance of end-to-end automatic speech recognition (ASR). In our prior work, we have proposed a multi-level LM, in which character-based and word-based RNN-LMs are combined in hybrid CTC/attention-based ASR. Although this multi-level approach achieves significant error reduction in the Wall Street Journal (WSJ) task, two different LMs need to be trained and used for decoding, which increase the computational cost and memory usage. In this paper, we further propose a novel word-based RNN-LM, which allows us to decode with only the word-based LM, where it provides look-ahead word probabilities to predict next characters instead of the character-based LM, leading competitive accuracy with less computation compared to the multi-level LM. We demonstrate the efficacy of the word-based RNN-LMs using a larger corpus, LibriSpeech, in addition to WSJ we used in the prior work. Furthermore, we show that the proposed model achieves 5.1 %WER for WSJ Eval’92 test set when the vocabulary size is increased, which is the best WER reported for end-to-end ASR systems on this benchmark.