适于多尺度宫颈癌细胞检测的改进算法北大核心CSCD

深度学习技术因其强大的特征提取能力而被广泛应用于目标检测任务中。针对多尺度宫颈癌细胞的识别准确率不均衡、检测效率低等问题,本文提出一种基于YOLO v3模型的改进识别算法mo-YOLO v3(mini-object-YOLO v3)。选用20倍数字扫描仪下采集的宫颈细胞图像作为数据集,为提高算法的鲁棒性,引入对比度增强、灰度图、旋转和翻转等多种数据增强策略扩充数据集;模型以Darknet53网络结合注意力机制作为主干模块,针对宫颈癌细胞尺寸差异大的特点,提出一种多尺度特征融合算法来优化模型结构;针对小目标检测精度低的问题,提出一种改进的损失函数,采用相对位置信息的方法减弱物体框对检测结果的影响。测试结果表明,本文所提的mo-YOLO v3模型不仅在总体识别精度上有明显的优势,同时大大提高了小尺寸宫颈癌细胞的定位精度。该模型对宫颈癌细胞识别的准确率达到90.42%,查准率达到96.20%,查全率达到93.77%,相似指数ZSI为94.97%,高于同类算法。     

Inhibition of human T-cell responses to house dust mite allergens by a T-cell receptor peptide

Recent analysis of the usage of T-cell receptor (TcR) beta chain variable region (V beta) gene elements by house dust mite (HDM)-reactive T cells from an atopic donor suggested that TcR-V beta 3 gene products may form a major component of the human T-cell repertoire reactive to this common allergen. In this study a peptide analog of the TcR-V beta 3 complementarity determining region 2 (CDR2) is shown to inhibit the polyclonal human T-cell response to HDM; this effect is specific because inhibition is dependent on the presence of V beta 3 + T cells. This experimental approach has been used to determine whether the pattern seen in T-cell clones derived from one atopic donor reflects TcR-V beta usage in the polyclonal response to allergen in the general population. Inhibition of more than 50% of the polyclonal response to allergen by V beta 3-CDR2 peptide was observed in 16 of 21 donors tested, suggesting that TcR-V beta 3 gene usage may form a major component of the human HDM repertoire and as such offer a suitable target for T cell-directed specific immunotherapy in HDM-allergic individuals. Depletion of CD8+ T cells abolishes peptide-mediated inhibition of CD4+ T-cell proliferation to HDM, suggesting that induction of a CD8+ regulatory T-cell subset by the CDR2 peptide may modulate HDM-specific allergic T-cell responses.     

Ovarian metastases from cervical carcinomas other than pure adenocarcinomas. A report of 12 cases

Twelve cases of ovarian metastases from cervical carcinomas, most with clinical manifestations of ovarian involvement, are reported. The patients were 23-73 years of age (average, 43 years). The ovarian and cervical tumors were synchronous in eight patients; in three, ovarian tumors were discovered 10 months, 2.5, and 3 years after the detection of a cervical neoplasm. In one patient, the cervical tumor was not discovered until autopsy 7 months after presentation. Four patients had abdominal swelling or distention, three had vaginal bleeding, three had an abnormal Papanicolaou smear, and two had masses discovered during pelvic examination. The ovarian tumors, six of which were bilateral, ranged from 5-17 cm (average, 9.5 cm) in maximal dimension in 11 patients; in the 12th patient, the involved ovary was not enlarged. The cervical tumors were grossly evident in 10 patients. They were usually deeply invasive, often with extracervical extension. Four were squamous cell carcinomas; two, small cell carcinomas; one, a mixed small cell carcinoma and adenocarcinoma; one, a mixed poorly differentiated carcinoid and adenocarcinoma; two, adenosquamous carcinomas; one, a transitional cell carcinoma; and one, an undifferentiated carcinoma. Various features, including bilaterality of the ovarian tumors, the finding that the histologic features of the ovarian tumors typically were unusual for a primary ovarian neoplasm, and the presence of extensive extracervical disease, led to the conclusion that the ovarian tumors were metastatic from the cervix. Although ovarian metastases of cervical carcinoma are uncommon, this series illustrates that, occasionally striking examples with clinical manifestations of ovarian involvement occur.     

Assessment of antifungal activities of fluconazole and amphotericin B administered alone and in combination against Candida albicans by using a dynamic in vitro mycotic infection model

We evaluated the pharmacodynamic activities of fluconazole and amphotericin B given alone and in combination against Candida albicans by using an in vitro model of bloodstream infection that simulates human serum pharmacokinetic parameters for these antifungals. Fluconazole was administered as a bolus into the model to simulate regimens of 200 mg every 24 h (q24 h) and 400 mg q24 h. Amphotericin B was administered at doses producing the peak concentration (2.4 micrograms/ml) observed with a regimen of 1 mg/kg of body weight q24 h. A combination regimen of fluconazole (400 mg q24 h) and amphotericin B (1 mg/kg q24 h) administered simultaneously and as a staggered regimen (amphotericin B bolus given 8 h after fluconazole bolus) was also simulated in the model to characterize possible antagonism between these agents. Fluconazole alone and amphotericin B alone demonstrated fungistatic (< 99.9% reduction in numbers of CFU per milliliter from the starting inoculum) and fungicidal (> 99.9% reduction) activity, respectively. When fluconazole and amphotericin B were administered simultaneously, fungicidal activity similar to that observed with amphotericin B alone was observed. Staggered administration of fluconazole and amphotericin B, however, resulted in a substantial reduction of the fungicidal activity of amphotericin B, producing fungistatic activity similar to that observed with noncombination fluconazole regimens. These results demonstrate the usefulness of this model for comparing the in vitro pharmacodynamic characteristics of different antifungal regimens and support the theory of azole-polyene antagonism. The effects of this antagonism on the in vivo activity and clinical usefulness of combination antifungal therapy, however, remain to be determined.     

Developing a brief measure of smoking in the home: description and preliminary evaluation

The insulin-like growth factors (IGF-I and IGF-II) play a key role in cellular proliferation and are involved in cellular transformation. The expression of the IGF-I receptor has been demonstrated in a variety of human tumor cell lines including ovarian cancer cells. Phosphorothioate antisense oligodeoxynucleotides (S-ODNs) were analyzed for their potential to suppress the IGF-I receptor in the NIH: OVCAR-3 ovarian cancer cell line. The application of the antisense S-ODN reduced potently the cell growth of unstimulated NIH:OVCAR-3 cells, whereas sense and mismatch S-ODNs were without any effect. This effect resembled that of the monoclonal antibody (MAb) alphaIR-3. In contrast to the antisense compound, this MAb only partially inhibited the IGF-I-induced proliferation of ovarian cancer cells. The concentration of the antisense S-ODN to exhibit an identical inhibition of cell proliferation was reduced to 50 nM when the oligonucleotides were delivered by the cationic lipid formulation lipofectin. The specificity of the antisense S-ODN action was confirmed by reduction of the receptor protein and of the receptor mRNA, as assayed by flow cytometry and by Northern blot hybridizations. Our data demonstrate the potency of antisense S-ODNs to target the IGF-I receptor message and show that antisense strategies against the IGF-I receptor may provide new strategies for the therapy of ovarian cancer.     

Semiconductor‐Based Photoelectrochemical Water Splitting at the Limit of Very Wide Depletion Region

In semiconductor‐based photoelectrochemical (PEC) water splitting, carrier separation and delivery largely relies on the depletion region formed at the semiconductor/water interface. As a Schottky junction device, the trade‐off between photon collection and minority carrier delivery remains a persistent obstacle for maximizing the performance of a water splitting photoelectrode. Here, it is demonstrated that the PEC water splitting efficiency for an n‐SrTiO₃ (n‐STO) photoanode is improved very significantly despite its weak indirect band gap optical absorption (α < 10⁴ cm^?1), by widening the depletion region through engineering its doping density and profile. Graded doped n‐SrTiO₃ photoanodes are fabricated with their bulk heavily doped with oxygen vacancies but their surface lightly doped over a tunable depth of a few hundred nanometers, through a simple low temperature reoxidation technique. The graded doping profile widens the depletion region to over 500 nm, thus leading to very efficient charge carrier separation and high quantum efficiency (>70%) for the weak indirect transition. This simultaneous optimization of the light absorption, minority carrier (hole) delivery, and majority carrier (electron) transport by means of a graded doping architecture may be useful for other indirect band gap photocatalysts that suffer from a similar problem of weak optical absorption.     

Spatial and radiometric resolution of the Landsat-3 RBV system

Abstract

Major constraints on the RBV system spatial and radiometric resolution are the optical limitations of the camera lens, the sensitivity of the vidicon, the system noise and the video sampling rate. The Landsat-3 RBV lens resolution limit is 21.45 m ground equivalent half-cycle. The Earthnet video sampling rate at 7.5 megasamples, gives an across track ground pixel size of 18 m. The along track pixel size is controlled by the number of scan lines in one frame (i.e. 4125), which gives a ground equivalent pixel size of 24 m. The MTF curve calculated from the RCA manual shows an EIFOV of 45.5 m. At 21.45 m ground equivalent half-cycle, the modulation transfer factor is 5.4 per cent which with a dynamic range of 64 quantizing levels gives a two or three level contrast between black and white linear ground targets (i.e. at the extremes of the dynamic range). In the panchromatic waveband with a white level saturation at 2.013 mW/cm²-sr, most ground cover types will occur within the first 50 per cent of the dynamic range. Given the Earthnet quantization step size to r.m.s. noise ratio of 0.51, there is a 56 per cent probability of assigning a level to within ± 1 value of its correct level. This is assuming a Gaussian distribution for the system noise. The noise is equivalent to the MTF sensitivity of the system at the lens resolution limit. In this study no attempt has been made to quantify the additional noise introduced by the radiometric defects observed in the RBV data or atmospheric attenuation.     

Evidence and a computational explanation of cultural differences in facial expression recognition.

Facial expressions are crucial to human social communication, but the extent to which they are innate and universal versus learned and culture dependent is a subject of debate. Two studies explored the effect of culture and learning on facial expression understanding. In Experiment 1, Japanese and U.S. participants interpreted facial expressions of emotion. Each group was better than the other at classifying facial expressions posed by members of the same culture. In Experiment 2, this reciprocal in-group advantage was reproduced by a neurocomputational model trained in either a Japanese cultural context or an American cultural context. The model demonstrates how each of us, interacting with others in a particular cultural context, learns to recognize a culture-specific facial expression dialect. (PsycINFO Database Record (c) 2010 APA, all rights reserved)