• Journal of Digital Convergence
• /
• v.19 no.12
• /
• pp.347-352
• /
• 2021

In this paper, we propose a method for finding feature subsets that are effective for classification in an input dataset by using a multi-agent reinforcement learning method. In the field of machine learning, it is crucial to find features suitable for classification. A dataset may have numerous features; while some features may be effective for classification or prediction, others may have little or rather negative effects on results. In machine learning problems, feature selection for increasing classification or prediction accuracy is a critical problem. To solve this problem, we proposed a feature selection method based on reinforced learning. Each feature has one agent, which determines whether the feature is selected. After obtaining corresponding rewards for each feature that is selected, but not by the agents, the Q-value of each agent is updated by comparing the rewards. The reward comparison of the two subsets helps agents determine whether their actions were right. These processes are performed as many times as the number of episodes, and finally, features are selected. As a result of applying this method to the Wisconsin Breast Cancer, Spambase, Musk, and Colon Cancer datasets, accuracy improvements of 0.0385, 0.0904, 0.1252 and 0.2055 were shown, respectively, and finally, classification accuracies of 0.9789, 0.9311, 0.9691 and 0.9474 were achieved, respectively. It was proved that our proposed method could properly select features that were effective for classification and increase classification accuracy.

• Journal of Marine Life Science
• /
• v.7 no.1
• /
• pp.15-20
• /
• 2022

Glioblastoma is one of the highly aggressive central nervous system tumors and it is difficult to treat owing its anatomical location. Peptides are novel class of drugs which has the potential to cross the blood brain barrier and exerts its anti-tumor activity. Here, we discovered a novel peptide from abalone (Haliotis discus hannai) next generation sequencing (NGS) data and tested its anticancer effect on glioblastoma cell line SNU-489. The anticancer activity was measured using a cytotoxicity assay in a time and dose-dependent manner. A concentration and time dependent increase in the cytotoxicity was seen in cells treated with the novel peptide. The highest cytotoxicity rate of about 67% was observed in SNU-489 cells treated with 200 µM peptide for 48 hrs. However, the cytotoxic effect was not or less observed in a normal skin cell line HaCaT at similar concentration, thus, evident of peptide's cell specific anticancer activity. In addition, the gene expression level of necroptosis-related genes was analyzed by qRT-PCR to elucidate the anticancer mechanism of the novel peptide. RIPK3 expression was significantly increased by 9.6-fold in 200 µM of novel peptide treatment group, and MLKL expression level was significantly elevated by 2-fold in 100 µM treated group compared to the control group. Therefore, this study confirmed that the novel abalone-derived peptide has anticancer potency, and it causes cancer cell death through the necroptosis mechanism. Collectively, these results suggest that the novel peptide could be candidate anticancer agent for the treatment of glioblastoma in the future.

• Korean Journal of Radiology
• /
• v.21 no.9
• /
• pp.1036-1044
• /
• 2020

Objective: To investigate the clinical feasibility of synthetic diffusion-weighted imaging (sDWI) at different b-values in patients with breast cancer by assessing the diagnostic image quality and the quantitative measurements compared with conventional diffusion-weighted imaging (cDWI). Materials and Methods: Fifty patients with breast cancer were assessed using cDWI at b-values of 800 and 1500 s/mm² (cDWI₈₀₀ and cDWI₁₅₀₀) and sDWI at b-values of 1000 and 1500 s/mm² (sDWI₁₀₀₀ and sDWI₁₅₀₀). Qualitative analysis (normal glandular tissue suppression, overall image quality, and lesion conspicuity) was performed using a 4-point Likert-scale for all DWI sets and the cancer detection rate (CDR) was calculated. We also evaluated cancer-to-parenchyma contrast ratios for each DWI set in 45 patients with the lesion identified on any of the DWI sets. Statistical comparisons were performed using Friedman test, one-way analysis of variance, and Cochran's Q test. Results: All parameters of qualitative analysis, cancer-to-parenchyma contrast ratios, and CDR increased with increasing b-values, regardless of the type of imaging (synthetic or conventional) (p < 0.001). Additionally, sDWI₁₅₀₀ provided better lesion conspicuity than cDWI₁₅₀₀ (3.52 ± 0.92 vs. 3.39 ± 0.90, p < 0.05). Although cDWI₁₅₀₀ showed better normal glandular tissue suppression and overall image quality than sDWI₁₅₀₀ (3.66 ± 0.78 and 3.73 ± 0.62 vs. 3.32 ± 0.90 and 3.35 ± 0.81, respectively; p < 0.05), there was no significant difference in their CDR (90.0%). Cancer-to-parenchyma contrast ratios were greater in sDWI₁₅₀₀ than in cDWI₁₅₀₀ (0.63 ± 0.17 vs. 0.55 ± 0.18, p < 0.001). Conclusion: sDWI₁₅₀₀ can be feasible for evaluating breast cancers in clinical practice. It provides higher tumor conspicuity, better cancer-to-parenchyma contrast ratio, and comparable CDR when compared with cDWI₁₅₀₀.