• Korean Journal of Radiology
• /
• v.24 no.2
• /
• pp.155-165
• /
• 2023

Objective: Little is known about the effects of using different expert-determined reference standards when evaluating the performance of deep learning-based automatic detection (DLAD) models and their added value to radiologists. We assessed the concordance of expert-determined standards with a clinical gold standard (herein, pathological confirmation) and the effects of different expert-determined reference standards on the estimates of radiologists' diagnostic performance to detect malignant pulmonary nodules on chest radiographs with and without the assistance of a DLAD model. Materials and Methods: This study included chest radiographs from 50 patients with pathologically proven lung cancer and 50 controls. Five expert-determined standards were constructed using the interpretations of 10 experts: individual judgment by the most experienced expert, majority vote, consensus judgments of two and three experts, and a latent class analysis (LCA) model. In separate reader tests, additional 10 radiologists independently interpreted the radiographs and then assisted with the DLAD model. Their diagnostic performance was estimated using the clinical gold standard and various expert-determined standards as the reference standard, and the results were compared using the t test with Bonferroni correction. Results: The LCA model (sensitivity, 72.6%; specificity, 100%) was most similar to the clinical gold standard. When expert-determined standards were used, the sensitivities of radiologists and DLAD model alone were overestimated, and their specificities were underestimated (all p-values < 0.05). DLAD assistance diminished the overestimation of sensitivity but exaggerated the underestimation of specificity (all p-values < 0.001). The DLAD model improved sensitivity and specificity to a greater extent when using the clinical gold standard than when using the expert-determined standards (all p-values < 0.001), except for sensitivity with the LCA model (p = 0.094). Conclusion: The LCA model was most similar to the clinical gold standard for malignant pulmonary nodule detection on chest radiographs. Expert-determined standards caused bias in measuring the diagnostic performance of the artificial intelligence model.

• International Journal of Computer Science & Network Security
• /
• v.24 no.5
• /
• pp.11-20
• /
• 2024

While artificial neural networks are adept at identifying patterns, they can struggle to distinguish between actual correlations and false associations between extracted facial features and criminal behavior within the training data. These associations may not indicate causal connections. Socioeconomic factors, ethnicity, or even chance occurrences in the data can influence both facial features and criminal activity. Consequently, the artificial neural network might identify linked features without understanding the underlying cause. This raises concerns about incorrect linkages and potential misclassification of individuals based on features unrelated to criminal tendencies. To address this challenge, we propose a novel region-based training approach for artificial neural networks focused on criminal propensity detection. Instead of solely relying on overall facial recognition, the network would systematically analyze each facial feature in isolation. This fine-grained approach would enable the network to identify which specific features hold the strongest correlations with criminal activity within the training data. By focusing on these key features, the network can be optimized for more accurate and reliable criminal propensity prediction. This study examines the effectiveness of various algorithms for criminal propensity classification. We evaluate YOLO versions YOLOv5 and YOLOv8 alongside VGG-16. Our findings indicate that YOLO achieved the highest accuracy 0.93 in classifying criminal and non-criminal facial features. While these results are promising, we acknowledge the need for further research on bias and misclassification in criminal justice applications

• Journal of Institute of Control, Robotics and Systems
• /
• v.21 no.2
• /
• pp.168-172
• /
• 2015

In general, open-loop fiber-optic gyroscopes (FOG) are less stable than closed-loop FOGs but they offer simpler implementation. The typical operation time of line-of-sight (LOS) stabilization systems is a few seconds to one hour. In this paper, a open-loop fiber optic gyroscope (FOG) for LOS applications is designed and implemented. The design goal is aimed at implementing a low cost, compact FOG with low Angle Random Walk (ARW) (< $0.03deg/\sqrt{h}$) and bias instability (< 0.25deg/h). The FOG uses an open-loop all-fiber configuration with 100M PM fiber wound on a small diameter spool. In order to get the design goal, digital signal processing techniques for signal detection, modulation control and compensation are designed and implemented in FPGA.

• Progress in Superconductivity and Cryogenics
• /
• v.19 no.1
• /
• pp.26-29
• /
• 2017

We studied the distribution of the current density and its magnetic-field dependence in GdBCO coated conductors with AC bias currents using low temperature scanning Hall probe microscopy. We selectively measured magnetic field profiles from AC signal obtained by Lock-in technique and calculated current distributions by inversion calculation. In order to confirm the AC measurement results, we applied DC current corresponding to RMS value of AC current and compared distribution of AC and DC transport current. We carried out the same measurements at various external DC magnetic fields, and investigated field dependence of AC current distribution. We notice that the AC current distribution unaffected by external magnetic fields and preserved their own path on the contrary to DC current.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2002.05b
• /
• pp.27-30
• /
• 2002

Photoconductor for direct detection flat-panel imager present a great materials challenge, since their requirements include high X -ray absorption, ionization and charge collection, low leakage current and large area deposition. Selenium is practical material. But it needs high thickness and high voltage in selenium for high ionization rate. We report comparative studies of detector sensitivity. One is an a-Se with $70{\mu}m$ thickness on glass. The other has hybrid layer of depositting ZnS phosphor with $100{\mu}m$ on a-Se. The leakage current of hybrid is similar to it of a-Se, but photocurrent is lager than a-Se. Both of them have high spatial resolution, but hybrid has higher sensitivity than a-Se at comparable bias voltage.

• Nuclear Engineering and Technology
• /
• v.28 no.4
• /
• pp.397-404
• /
• 1996

The hydrogenated amorphous silicon (a-Si : H) holds good promise for radiation detection from its inherent merits over crystalline counterpart. For the application to alpha spectroscopy, the induced charge collection in a-Si : H pin detector diodes ons simulated based on a relevant non-uniform charge generation model. The simulation was peformed for the initial energy and the range of incident alpha particles, detector thickness and the operational parameters such as the applied reverse bias voltage and shaping time. From the simulation, the total charge collection was strongly affected by hole collection as expected. To get a reasonable signal generation, therefore, the hole collection should be seriously considered for detector operational parameters such as shaping time and reverse voltage etc. For the spectroscopy of alpha particle from common alpha sources, the amorphous silicon should have about 70${\mu}{\textrm}{m}$ thickness.

• Journal of the Korean Statistical Society
• /
• v.33 no.2
• /
• pp.219-231
• /
• 2004

Process adjustment is a complimentary tool to process monitoring in process control. Although original intention of process adjustment is not identifying a special cause, detection and elimination of special causes may lead to significant process improvement. In this paper, we examine the impact of special causes on process adjustment. The bias in the adjusted output process is derived for each type of special causes, and average run length (ARL) of the Shewhart chart applied to the adjusted output is computed for each special cause types. Numerical results are illustrated for the ARL of the Shewhart chart, thereupon seriousness of special causes on process adjustment is evaluated for each type of special causes.

• Annual Conference on Human and Language Technology
• /
• 2012.10a
• /
• pp.58-61
• /
• 2012

본 논문에서는 트위터 사용자들의 의견을 바이어스 탐지 하기 위해, 핵심 자질 추출 방법으로 HITS 그래프를 이용한 방법을 제안한다. 제안하는 핵심 자질 추출 방법은 사람이 직접 추출하지 못하는 자질도 추출할 수 있는 장점을 보였다. 제안한 핵심 자질 추출이 바이어스 탐지에 유효함을 검증하기 위해 4개의 토픽에 대해 평가 했을 때 제안 모델이 기존 모델보다 우수한 성능을 보였다.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.7
• /
• pp.640-645
• /
• 2013

This paper presents a step length estimation algorithm for Pedestrian Dead Reckoning using linear calibrated ZUPT (zero velocity update) with a foot mounted IMU. The IMU consists of 3 axis accelerometer, gyro and magnetometer. Attitude of IMU is estimated using an inertial navigation algorithm. To increase accuracy of step length estimation algorithm, we propose a stance detection algorithm and an enhanced ZUPT. The enhanced ZUPT calculates firefighter's step length considering velocity error caused by sensor bias during one step. This algorithm also works efficiently at various motions, such as crawling, sideways and stair stepping. Through experiments, the step length estimation performance of the proposed algorithm is verified.

• Proceedings of the KIEE Conference
• /
• 1992.07a
• /
• pp.434-436
• /
• 1992

In this paper, we propose a new multilevel nonlinear filter for simultaneous edge detection and noise suppression, which we call a order statistic-median hybrid(OMH) filler. The median-related filters cause an edge shift in the presence of an impulse near the edge. The proposed filter reduces such edge shifting while suppressing impulsive as well as nonimpulsive noise. We show that at the noisy edge point the OMH filter is substantially superior to the median filter, the $\alpha$-TM filter and the STM filter[I] in two respects: (a) the output bias error and (b) the output mean square error. Test results confirm that the OMH filter is robust in preserving sharp edges, inhibiting edge shifting, and suppressing a wide variety of noise. The structure for the OMH filter integrated circuit is also described.