传统基于生成对抗网络的语音增强算法(Speech Enhancement Algorithm Based on Generative Adversarial Networks,SEGAN)在时域上对语音进行增强处理,完全忽略了语音样本在频域上的分布情况。在低信噪比条件下,语音信号会淹没在噪声中,...传统基于生成对抗网络的语音增强算法(Speech Enhancement Algorithm Based on Generative Adversarial Networks,SEGAN)在时域上对语音进行增强处理,完全忽略了语音样本在频域上的分布情况。在低信噪比条件下,语音信号会淹没在噪声中,带噪语音的时域分布信息很难捕获,因此,SEGAN的增强性能会急剧下降,其增强语音的语音质量和语音可懂度很低。针对该问题,提出了基于时频域生成对抗网络的语音增强算法(Time-Frequency Domain SEGAN,TFSEGAN)。TFSEGAN采用了时频域双判别器的模型结构和时频域L1损失函数,时域判别器的输入为语音样本的时域特征,频域判别器的输入为语音样本的频域特征。在训练过程中,时域判别器将语音样本的时域分布信息作为判别标准,而频域判别器将语音样本的频域分布信息作为判别标准。在两个判别器的作用下,TFSEGAN的生成器能够同时学习语音样本在时域和频域中的分布规律和信息。实验证明,在低信噪比条件下,与SEGAN相比,TFSEGAN的语音质量与可懂度分别提升了约17.45%和11.75%。展开更多
目的探讨磁共振T_(2)WI信号强度(T_(2)W I Signal Intensity,SI-T_(2))、表观扩散系数(ADC)值二者联合鉴别肾透明细胞癌(CCRCC)和乏脂型肾血管平滑肌脂肪瘤(fpAML)的价值。方法回顾性分析经手术病理证实的38例肾透明细胞癌和27例乏脂型...目的探讨磁共振T_(2)WI信号强度(T_(2)W I Signal Intensity,SI-T_(2))、表观扩散系数(ADC)值二者联合鉴别肾透明细胞癌(CCRCC)和乏脂型肾血管平滑肌脂肪瘤(fpAML)的价值。方法回顾性分析经手术病理证实的38例肾透明细胞癌和27例乏脂型肾血管平滑肌脂肪瘤患者的资料。所有患者术前均行MRI平扫和扩散加权成像(DWI)检查。对压脂T_(2)WI(T_(2)WI-FS)信号强度值SI-T_(2)及ADC值进行定量测量,记录病灶与正常肾组织的测量比值,采用多因素Logistic回归分析确定SI-T_(2)、ADC值联合鉴别两种肿瘤的独立影响因素,以病理结果为金标准,绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC)并做比较。结果肾透明细胞癌的SI-T_(2)比值高于乏脂型肾血管平滑肌脂肪瘤,信号强度的ROC AUC、灵敏度及特异度分别为0.725、85.5%和71.6%(P<0.05);肾透明细胞癌的ADC比值低于乏脂型肾血管平滑肌脂肪瘤(P<0.05),ADC测量的ROC AUC、灵敏度和特异性分别为0.812、79.8%和75.4%。SI-T_(2)联合ADC值的ROC分析发现,鉴别两者的AUC、敏感性及特异性分别为0.885,89.5%、83.6%。结论MRI T_(2)信号强度比值及ADC比值均有助于鉴别肾透明细胞癌和乏脂型肾血管平滑肌脂肪瘤,两种参数结合能更好的鉴别这两种肿瘤。展开更多
Tyrosine-decahydrofluorene derivatives are a class of hybrid compounds that integrate the properties of polyketides and nonribosomal peptides.These compounds feature a[6.5.6]tricarbocyclic core and a para-cyclophane e...Tyrosine-decahydrofluorene derivatives are a class of hybrid compounds that integrate the properties of polyketides and nonribosomal peptides.These compounds feature a[6.5.6]tricarbocyclic core and a para-cyclophane ether moiety in their structures and exhibit anti-tumor and anti-microbial activities.In this study,we constructed the biosynthetic pathway of xenoacremones from Xenoacremonium sinensis ML-31 in the Aspergillus nidulans host,resulting in the identification of four novel tyrosine-decahydrofluorene analogs,xenoacremones I–L(1-4),along with two known analogs,xenoacremones A and B.Remarkably,compounds 3 and 4 contained a 12-membered para-cyclophane ring system,which is unprecedented among tyrosine-decahydrofluorene analogs in X.sinensis.The successful reconstruction of the biosynthetic pathway and the discovery of novel analogs demonstrate the utility of heterologous expression strategy for the generation of structurally diverse natural products with potential biological activities.展开更多
文摘传统基于生成对抗网络的语音增强算法(Speech Enhancement Algorithm Based on Generative Adversarial Networks,SEGAN)在时域上对语音进行增强处理,完全忽略了语音样本在频域上的分布情况。在低信噪比条件下,语音信号会淹没在噪声中,带噪语音的时域分布信息很难捕获,因此,SEGAN的增强性能会急剧下降,其增强语音的语音质量和语音可懂度很低。针对该问题,提出了基于时频域生成对抗网络的语音增强算法(Time-Frequency Domain SEGAN,TFSEGAN)。TFSEGAN采用了时频域双判别器的模型结构和时频域L1损失函数,时域判别器的输入为语音样本的时域特征,频域判别器的输入为语音样本的频域特征。在训练过程中,时域判别器将语音样本的时域分布信息作为判别标准,而频域判别器将语音样本的频域分布信息作为判别标准。在两个判别器的作用下,TFSEGAN的生成器能够同时学习语音样本在时域和频域中的分布规律和信息。实验证明,在低信噪比条件下,与SEGAN相比,TFSEGAN的语音质量与可懂度分别提升了约17.45%和11.75%。
文摘目的探讨磁共振T_(2)WI信号强度(T_(2)W I Signal Intensity,SI-T_(2))、表观扩散系数(ADC)值二者联合鉴别肾透明细胞癌(CCRCC)和乏脂型肾血管平滑肌脂肪瘤(fpAML)的价值。方法回顾性分析经手术病理证实的38例肾透明细胞癌和27例乏脂型肾血管平滑肌脂肪瘤患者的资料。所有患者术前均行MRI平扫和扩散加权成像(DWI)检查。对压脂T_(2)WI(T_(2)WI-FS)信号强度值SI-T_(2)及ADC值进行定量测量,记录病灶与正常肾组织的测量比值,采用多因素Logistic回归分析确定SI-T_(2)、ADC值联合鉴别两种肿瘤的独立影响因素,以病理结果为金标准,绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC)并做比较。结果肾透明细胞癌的SI-T_(2)比值高于乏脂型肾血管平滑肌脂肪瘤,信号强度的ROC AUC、灵敏度及特异度分别为0.725、85.5%和71.6%(P<0.05);肾透明细胞癌的ADC比值低于乏脂型肾血管平滑肌脂肪瘤(P<0.05),ADC测量的ROC AUC、灵敏度和特异性分别为0.812、79.8%和75.4%。SI-T_(2)联合ADC值的ROC分析发现,鉴别两者的AUC、敏感性及特异性分别为0.885,89.5%、83.6%。结论MRI T_(2)信号强度比值及ADC比值均有助于鉴别肾透明细胞癌和乏脂型肾血管平滑肌脂肪瘤,两种参数结合能更好的鉴别这两种肿瘤。
基金supported by the the National Key Research and Development Program of China(No.2021YFC2300400)the Biological Resources Program,Chinese Academy of Sciences(No.KFJ-BRP-009)the Key Research Program of Frontier Sciences,the Chinese Academy of Sciences(No.ZDBS-LY-SM016).
文摘Tyrosine-decahydrofluorene derivatives are a class of hybrid compounds that integrate the properties of polyketides and nonribosomal peptides.These compounds feature a[6.5.6]tricarbocyclic core and a para-cyclophane ether moiety in their structures and exhibit anti-tumor and anti-microbial activities.In this study,we constructed the biosynthetic pathway of xenoacremones from Xenoacremonium sinensis ML-31 in the Aspergillus nidulans host,resulting in the identification of four novel tyrosine-decahydrofluorene analogs,xenoacremones I–L(1-4),along with two known analogs,xenoacremones A and B.Remarkably,compounds 3 and 4 contained a 12-membered para-cyclophane ring system,which is unprecedented among tyrosine-decahydrofluorene analogs in X.sinensis.The successful reconstruction of the biosynthetic pathway and the discovery of novel analogs demonstrate the utility of heterologous expression strategy for the generation of structurally diverse natural products with potential biological activities.
