| 若山 諒大 | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, | 松本 健一, | 小野 直亮 | |
|
title: Visualization of Nutritional Value Based on Data Science
abstract: Given that all foods contain multiple nutrients, it is challenging to accurately grasp and understand their nutritional value. For a comprehensive understanding of nutritional value in foods, it is necessary to evaluate foods based on data science using the Standards Tables of Food Composition in Japan. This study focused on visualization as one method of evaluating the nutritional value of foods, specifically two-dimensional mapping and scoring. In two-dimensional visualization, food classification was performed based on their nutritional composition using a two-dimensional map by t-SNE (t-distributed Stochastic Neighbor Embedding) and k-NN (k-nearest neighbor). As a results, most foods formed clusters according to the food groups in the food composition table. This revealed that many food groups have common nutrient patterns and can be classified based on nutrient similarity. In visualization by scoring, we focused on the Nutritional Profiling System (NPS), a method of evaluating the nutritional value of foods by scoring the amount of nutrients and other components contained in foods based on science. Since the issues related to nutrition and health differ depending on the region or life stage, in the development of the Meiji NPS, we focused on different health issues in Japan depending on the life stage and developed two types of NPS: the Meiji NPS for adults considering lifestyle-related diseases among adults and thinness among young women, and the Meiji NPS for older adults considering frailty. The convergent validity of the Meiji NPS was confirmed by the correlation coefficients with the Nutrient-Rich Foods Index 9.3 (NRF9.3), an NPS that has been shown to be related to the quality of the diet. As a result, it was indicated that the Meiji NPS is a valid NPS that can sufficiently evaluate the nutritional value of foods. language of the presentation: 日本語（Japanese） 発表題目: データサイエンスに基づいた栄養価値の可視化 発表概要:全ての食品には複数の栄養素が含まれていることを考えると、栄養価値を正しく把握し、理解することは困難である。食品における栄養価値の体系的な理解のためには、日本食品成分標準表をもとにデータサイエンスに基づいて食品の評価を行う必要がある。食品の栄養価値を評価する方法の1つとして可視化が挙げられ、本研究ではマッピングによる可視化とスコアによる可視化を対象とした。マッピングによる可視化では、t-SNE（t-distributed Stochastic Neighbor Embedding）による2次元マップとk-NN（k-nearest neighbor）から、栄養組成に基づいて食品を分類した。その結果、ほとんどの食品が食品成分表の食品群に従ってクラスターを形成していた。これにより、多くの食品群が共通の栄養素パターンを持ち、栄養素の類似性に基づいて分類できることが明らかになった。スコアによる可視化では、食品に含まれる栄養素等の量を科学的な根拠に基づきスコア化するなどして、食品の栄養価値を評価する手法であるNPS（Nutritional Profiling System）に着目した。地域あるいはライフステージにより栄養や健康に関する課題は異なるため、Meiji NPSの設計においてはライフステージの違いによって異なる日本の健康課題に着目し、生活習慣病と若年女性のやせ対策を考慮した成人NPSとフレイル対策を考慮した高齢者NPSの2種類を開発した。食事の質の高さとの関連が立証されているNPSであるNutrient-Rich Foods Index 9.3（NRF9.3）との相関を検討することにより、成人NPSおよび高齢者NPSの妥当性を確認した。その結果、Meiji NPSは十分に食品の栄養価値を評価できる尺度であることが示された。 | ||||||
| HOSSAIN MD SHAKHAOUT | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, | 松本 健一, | 小野 直亮, | MD.Altaf-Ul-Amin |
|
title: Medical Imaging: Blur Detection in Colonoscopy and Deblarring abstract: Colorectal cancer (CRC) is the third most diagnosed cancer in the world and according to recent studies, CRC will increase by about 60% by the year 2030, which is more than 2.2 million new cases and 1.1 million cancer deaths. To tackle CRC, colonoscopy is a proven technique for early detection. Colonoscopy images are often corrupted by blur, opaqueness, halation etc. Noisy frames increase the possibility of miss detection. Detecting blurry frames in colonoscopy and restoration with the help of image processing and deep learning will help correctly detect the polyps, even possibly may reduce the time of colonoscopy. I will present our noble image processing algorithm to detect blurry frames from colonoscopy videos which can be further extended in other medical images. I will also discuss about recent state-of-the-art image restoration diffusion architectures which can be implemented for blur colonoscopy image restoration. language of the presentation: English | ||||||