Performance assessment through bootstrap

A new performance evaluation paradigm for computer vision systems is proposed. In real situation, the complexity of the input data and/or of the computational procedure can make traditional error propagation methods infeasible. The new approach exploits a resampling technique recently introduced in statistics, the bootstrap. Distributions for the output variables are obtained by perturbing the nuisance properties of the input, i.e., properties with no relevance for the output under ideal conditions. From these bootstrap distributions, the confidence in the adequacy of the assumptions embedded into the computational procedure for the given input is derived. As an example, the new paradigm is applied to the task of edge detection. The performance of several edge detection methods is compared both for synthetic data and real images. The confidence in the output can be used to obtain an edgemap independent of the gradient magnitude     

Characteristics of body-tied triple-gate pMOSFETs

Body-tied triple-gate pMOSFETs were fabricated using bulk Si wafers and characterized. Process steps to implement the devices are explained briefly. Device characteristics of the triple-gate pMOSFETs were compared with those of the conventional planar channel device. While maintaining low off-leakage currents and threshold voltages similar to those of planar pMOSFETs in the parallel arrayed 30 000 transistors, the body-tied triple-gate MOSFETs showed about 74 mV/dec of subthreshold swing (92 mV/dec for conventional devices) and a drain-induced barrier lowering of 34 mV/V (92 mV/V for conventional devices). It was also addressed that I/sub SUB//I/sub D/ of the body-tied triple-gate is lower than that of the planar channel device.     

A study on the temperature dependence of epoxy/BaTiO<Subscript>3</Subscript> composite embedded capacitor films

In this study, the temperature dependence of capacitance, one of the most important properties of embedded capacitor films (ECFs), was investigated. The temperature dependence of the capacitance of ECFs was determined by the temperature dependence of the dielectric constant and thickness, and among these, the main factor was the dielectric constant of ECFs. The dielectric constant of ECFs was determined by that of epoxy and BaTiO₃ powders. Below 130°C, the dielectric constant of ECFs increased as temperature increased, and was mainly affected by an epoxy matrix. However, above 130°C (the Curie temperature of BaTiO₃), the increased rate of the dielectric constant of ECFs started decreasing. This was due to the fact that BaTiO₃ powder undergoes a phase transition from a tetragonal to a cubic structure, and its dielectric constant decreases at 130°C. The dielectric constant of BaTiO₃ powder was obtained from measured dielectric constants of ECF and application of the Lichtenecker logarithmic rule.     

A 1 Gbit synchronous dynamic random access memory with anindependent subarray-controlled scheme and a hierarchical decodingscheme

A prototype 1 Gbit synchronous DRAM with independent subarray-controlled isolation and hierarchical decoding schemes is demonstrated to alleviate the difficulties encountered in high-density devices with regard to failure analysis and performance optimization. The scheme to isolate memory arrays from “hard” defects and to overcome the dc leakages of “soft” defects with external sources allows monitoring of the leakage current for the defect analysis and testing of the device without being limited by the capabilities of on-chip voltage sources. A hierarchical decoding scheme with a dynamic CMOS series logic predecoder achieves improvements in circuit speed, power, and complexity. As a result, evaluation of the prototype devices can be facilitated, and the optimized circuit schemes achieve enhanced circuit performance. A fully working 1 Gbit synchronous DRAM with a chip size of 570 mm² was fabricated using a 0.16 μm CMOS process and tested for excellent functionality up to 143 MHz     

Effects of information on the reinforcing, subjective, and psychomotor effects of nitrous oxide in healthy volunteers

One method of quality control which has recently been recommended by professional bodies in the UK is the 'rapid review' method. This involves the microscopic 30 s review of all negative cervical smears with the intention of flagging potential missed abnormalities. Although it has been suggested that rapid review is better than 10% random rescreening of negative smears, the efficiency and efficacy of this method of quality control have not been thoroughly evaluated. We have used the AxioHOME system, which can record the area of a slide covered and the screening time, to investigate slide coverage during rapid review quality control, as performed by 15 cytoscreeners and MLSOs reviewing a test set of 22 slides each. The test set comprised 18 negative slides, three positive slides, and one unsatisfactory slide. We have recorded two distinct methods of rapid review in use amongst cytotechnologists, the step method and the whole slide method. The data show that rapid review takes longer on average than the recommended 30 s, the mean screening times being 76 s and 82 s for the step and whole slide methods, respectively. Abnormal smears were missed on three of 15 occasions by the step method (sensitivity 80%, positive predictive value 85%), and on seven of 30 occasions by the whole slide method (sensitivity 76.6%, positive predictive value 45%). However, the 95% confidence intervals were wide (57.7-90.7% for the step method, and 51.9-95.7% for the whole slide method). Analysis of scanning tracks and screening rates shows significant flaws in the methodology of rapid review. Abnormal cells were not identified, although dyskaryotic cells were included in the scanning track on nine occasions, seven using the whole slide method and two using the step method. On one occasion (using the step method) abnormal cells were not identified because they were not included in the scanning track. Further research is in progress to determine optimal methods of rapid review, and whether the rapid review technique is as effective as automated screening systems for quality assurance in cytology.     

Retrospective comparison of infusional 5-fluorouracil, doxorubicin, and mitomycin-C (modified FAM) combination chemotherapy versus palliative therapy in treatment of advanced gastric cancer

About one-third of patients with gastric cancer are unresectable at the time of diagnosis. Their median survival is < 6 months, with a grave prognosis. The purpose of this study was to assess the efficacy of a modified FAM (mFAM) regimen in advanced gastric cancer. We retrospectively reviewed the clinical records of 409 advanced gastric cancer patients who had not received curative surgery. Among 409 patients, 202 patients were treated with an mFAM regimen (infusional 5-FU + doxorubocin + mitomycin-C), and 207 patients received no chemotherapy (control group). No differences were found in clinical parameters between the two groups. The 1-year survival rates were 34.1% for the mFAM-treated group and 22.5% for the control group (p = 0.0135). In subset analysis, a higher 1-year survival rate was demonstrated in patients with mFAM and palliative surgery. Of the 154 evaluable patients in the mFAM-treated group, the response rate was 17.5%. In these patients, median response duration was 30 weeks, and progression-free survival was 23 weeks. Overall toxicity of mFAM regimen was relatively tolerable and reversible. In conclusion, FAM combination chemotherapy, which has been used as a standard therapy, prolonged survival after modification of the administration schedule and combination with palliative surgery. A prospective randomized study is warranted to confirm this conclusion from our retrospective study.     

Numerical accuracy of multipole expansion for 2D MLFMA

A numerical study of the multipole expansion for the multilevel fast multipole algorithm (MLFMA) is presented. In the numerical implementation of MLFMA, the error comes from three sources: the truncation of the addition theorem; the approximation of the integration; the aggregation and disaggregation process. These errors are due to the factorization of the Green's function which is the mathematical core of the algorithm. Among the three error sources, we focus on the truncation error and a new approach of selecting truncation numbers for the addition theorem is proposed. Using this approach, the error prediction and control can be improved for the small buffer sizes and high accuracy requirements.     

RELATIONSHIPS BETWEEN CONSTANTS OF THE FREUNDLICH EQUATION AND TEMPERATURE FOR GASEOUS ADSORPTION

For the Freundlich equation, the linear dependences of In k and n on T and In k on n were derived theoretically and examined by experimental data. In addition, the influence of the adsorbed amount on the isosteric heat of adsorption was also studied. The result shows that the isosteric heat of adsorption is varied linearly with the logarithm of adsorbed amount, ft is also found from experimental data that the plots of In k vs. n for various adsorbates adsorbed on the same adsorbent can be correlated by a straight line for adsorbates with the similar values of the parameter A' of Clapeyron equation.     

Over-erase phenomenon in SONOS-type flash memory and its minimization using a hafnium oxide charge storage Layer

The over-erase phenomenon in the polysilicon-oxide-silicon nitride-oxide-silicon (SONOS) memory structure is minimized by using hafnium oxide or hafnium aluminum oxide to replace silicon nitride as the charge storage layer (the resulting structures are termed SOHOS devices, where the "H" denotes the high dielectric constant material instead of silicon nitride). Unlike SONOS devices, SOHOS structures show a reduced over-erase phenomenon and self-limiting charge storage behavior under both erase and program operations. These are attributed to the differences in band offset and the crystallinity of the charge storage layer.