• Journal of Internet Computing and Services
• /
• v.24 no.5
• /
• pp.17-27
• /
• 2023

Today, as AI (Artificial Intelligence) technology develops and its practicality increases, it is widely used in various application fields in real life. At this time, the AI model is basically learned based on various statistical properties of the learning data and then distributed to the system, but unexpected changes in the data in a rapidly changing data situation cause a decrease in the model's performance. In particular, as it becomes important to find drift signals of deployed models in order to respond to new and unknown attacks that are constantly created in the security field, the need for lifecycle management of the entire model is gradually emerging. In general, it can be detected through performance changes in the model's accuracy and error rate (loss), but there are limitations in the usage environment in that an actual label for the model prediction result is required, and the detection of the point where the actual drift occurs is uncertain. there is. This is because the model's error rate is greatly influenced by various external environmental factors, model selection and parameter settings, and new input data, so it is necessary to precisely determine when actual drift in the data occurs based only on the corresponding value. There are limits to this. Therefore, this paper proposes a method to detect when actual drift occurs through an Anomaly analysis technique based on XAI (eXplainable Artificial Intelligence). As a result of testing a classification model that detects DGA (Domain Generation Algorithm), anomaly scores were extracted through the SHAP(Shapley Additive exPlanations) Value of the data after distribution, and as a result, it was confirmed that efficient drift point detection was possible.

• Nuclear Medicine and Molecular Imaging
• /
• v.42 no.3
• /
• pp.192-200
• /
• 2008

To evaluate the hemodynamic changes and the predictive factors of the clinical outcome in pediatric patients with moyamoya disease, we analyzed pre/post basal/acetazolamide stress brain perfusion SPECT with automated volume of interest (VOIs) method. Methods: Total fifty six (M:F = 33:24, age $6.7{\pm}3.2$ years) pediatric patients with moyamoya disease, who underwent basal/acetazolamide stress brain perfusion SPECT within 6 before and after revascularization surgery (encephalo-duro-arterio-synangiosis (EDAS) with frontal encephalo-galeo-synangiosis (EGS) and EDAS only followed on contralateral hemisphere), and followed-up more than 6 months after post-operative SPECT, were included. A mean follow-up period after post-operative SPECT was $33{\pm}21$ months. Each patient's SPECT image was spatially normalized to Korean template with the SPM2. For the regional count normalization, the count of pons was used as a reference region. The basal/acetazolamide-stressed cerebral blood flow (CBF), the cerebral vascular reserve index (CVRI), and the extent of area with significantly decreased basal/acetazolamide- stressed rCBF than age-matched normal control were evaluated on both medial frontal, frontal, parietal, occipital lobes, and whole brain in each patient's images. The post-operative clinical outcome was assigned as good, poor according to the presence of transient ischemic attacks and/or fixed neurological deficits by pediatric neurosurgeon. Results: In a paired t-test, basal/acetazolamide-stressed rCBF and the CVRI were significantly improved after revascularization (p<0.05). The significant difference in the pre-operative basal/acetazolamide-stressed rCBF and the CVRI between the hemispheres where EDAS with frontal EGS was performed and their contralateral counterparts where EDAS only was done disappeared after operation (p<0.05). In an independent student t-test, the pre-operative basal rCBF in the medial frontal gyrus, the post-operative CVRI in the frontal lobe and the parietal lobe of the hemispheres with EDAS and frontal EGS, the post-operative CVRI, and ${\Delta}CVRI$ showed a significant difference between patients with a good and poor clinical outcome (p<0.05). In a multivariate logistic regression analysis, the ${\Delta}CVRI$ and the post-operative CVRI of medial frontal gyrus on the hemispheres where EDAS with frontal EGS was performed were the significant predictive factors for the clinical outcome (p =0.002, p =0.015), Conclusion: With probabilistic map, we could objectively evaluate pre/post-operative hemodynamic changes of pediatric patients with moyamoya disease. Specifically the post-operative CVRI and the post-operative CVRI of medial frontal gyrus where EDAS with frontal EGS was done were the significant predictive factors for further clinical outcomes.