| Japanese |
|---|
| Title | FDG動態解析における速度定数推定値の分布とノイズの関係に関する基礎的検討 |
|---|
| Subtitle | 原著 |
|---|
| Authors | 木村裕一*, 外山比南子**, 千田道雄** |
|---|
| Authors(kana) | |
|---|
| Organization | *東京医科歯科大学医用器材研究所, **東京都老人総合研究所ポジトロン医学研究部門 |
|---|
| Journal | 核医学 |
|---|
| Volume | 33 |
|---|
| Number | 2 |
|---|
| Page | 107-114 |
|---|
| Year/Month | 1996/2 |
|---|
| Article | 原著 |
|---|
| Publisher | 日本核医学会 |
|---|
| Abstract | 「要旨」動態解析において, 速度定数の正確な推定には, 評価関数Φ, すなわち最小二乗法では, 残差二乗和を未知パラメータ値のセットの関数として表したものの, 形状の把握が必要である. 本論文では, 組織時間放射能曲線(tTAC)中の雑音のレベルとΦの形状との関係を, シミュレーションによって求め, Φの形状および速度定数の真値が分布する範囲を検討した. 同時に, 臨床データのΦの形状を求めた. その結果, tTACにノイズが含まれていない場合には, Φにおいて真値は周囲から独立した最小点となっており, その発見は容易である. しかしノイズが増加するに伴い, 真値はその周囲に発生した極小点群に埋没していく. シミュレーションの結果, 真値付近のΦの形状は, 真値を含む椀状となっており, 非線形計画法による収束点を中心とした, 周囲20%ほどの領域を再探索することによって, より正確な速度定数の推定が可能であることが示唆された. |
|---|
| Practice | 臨床医学：一般 |
|---|
| Keywords | Dynamic study, Compartment model, Noise, FDG. |
|---|
| English |
|---|
| Title | Basic Study on Relationship between Estimated Rate Constants and Noise in FDG Kinetic Analysis |
|---|
| Subtitle | Original Articles |
|---|
| Authors | Yuichi KIMURA*, Hinako TOYAMA**, Michio SENDA** |
|---|
| Authors(kana) | |
|---|
| Organization | *Institute for Medical & Dental Engineering, Tokyo Medical & Dental University, **Positron Medical Center, Tokyo Metropolitan Institute of Gerontology |
|---|
| Journal | The Japanese Journal of nuclear medicine |
|---|
| Volume | 33 |
|---|
| Number | 2 |
|---|
| Page | 107-114 |
|---|
| Year/Month | 1996/2 |
|---|
| Article | Original article |
|---|
| Publisher | THE JAPANESE SOCIETY OF NUCLEAR MEDICINE |
|---|
| Abstract | [Summary]For accurate estimation of the rate constants in 18F-FDG dynamic study, the shape of the estimation function(Φ)is crucial. In this investigation, the relationship between the noise level in tissue time activity curve and the shape of the least squared estimation function which is the sum of squarred error between a function of model parameters and a measured data is calculated in 3 parameter model of 18F-FDG. In the first simulation, by using actual plasma time activity curve, the true tissue curve was generated from known sets of rate constants ranging 0.05<=k1<=0.15, 0.1<=k2<=0.2 and 0.01<=k3<=0.1 in 0.01 step. This procedure was repeated under various noise levels in the tissue time activity curve from 1 to 8% of the maximum value in the tissue activity. In the second simulation, plasma and tissue time activity curves from clinical 18F-FDG dynamic study were used to calculate the Φ. Results are : (1)In the noise-free case, because the global minima is separated from neighboring local minimums, it was easy to find out the optimum point. However, with increasing noise level, the optimum point was buried in many neighboring local minima. Making it difficult to find out the optimum point. (2)The optimum point was found within 20% of the convergence point by standard non-linear optimization method. (3)The shape of Φ for the clinical data was similar to that with the noise level of 3 or 5% in the first simulation. Therefore direct search within the area extending 20% from the result of usual non-linear curve fitting procedure is recommended for accurate estimation of the constants. |
|---|
| Practice | Clinical medicine |
|---|
| Keywords | Dynamic study, Compartment model, Noise, FDG. |
|---|