JP3583771B2 - Chromatogram analysis method - Google Patents

Description

上式において、時刻ｔにおける検出波形をｙ（ｔ），サンプル注入時のオリジナルな長方形波形をｒ（ｔ），拡がり関数をｈ（ｔ）とする。上式においては、検出波形ｙ（ｔ）は、オリジナルな長方形波形ｒ（ｔ）と、拡がり関数ｈ（ｔ）とのコンボリュ−ションであることを示している。
【００３４】
ピ−クの形状の拡がりについては、上記説明でよいが、リテンションタイムについても考慮すると、もう一つＤｉｒａｃのデルタ関数がコンボリュ−ション演算されていると考えられる。これを（３）式に示す。
【数３】

【００３５】
上式において、ＥＭＧ（ｔ）は、Ｅｘｐｏｎｅｎｔｉａｌｌｙ Ｍｏｄｉｆｉｅｄ Ｇａｕｓｓｉｎであり、σ、τ、ｔ_Ｒは、ＥＭＧ（ｔ）のパラメータ、δ（ｔ）はデルタ関数、ｍ（ｔ）は修正関数である。
拡がりのメカニズムについては、“ＤＹＮＡＭＩＣＳ ＯＦ ＣＨＲＯＭＡＴＯＧＲＡＰＹ”，Ｊ．Ｃ．Ｇｉｄｄｉｎｇｓ著，Ｍａｒｃｅｌ Ｄｅｋｋｅｒ刊，Ｎｅｗ Ｙｏｒｋ，１９６５に詳しく説明されているのでそれに譲る。
【００３６】
拡がり関数ｈ（ｔ）は、カラム内での拡がりとカラム外での拡がりの両方からの影響により形成される。つまりカラム内での拡がり関数とカラム外での拡がり関数のコンボリュ−ションである。
【００３７】
図２における非保持ピ−ク６１に着目し、図３を参照し説明すると、注入時は長方形波形７１だったものが、カラム４１を通過し、ガウシアン波形７２に拡げられ、カラム外の拡がりの影響を受け、波形７３となり、ピ−ク６１として検出されていることになる。
【００３８】
一方、保持ピ−ク６２も同様なメカニズムで拡げられ、ピ−クが形成されているのであるが、前記カラム４１内に保持されているために、若干広めのピ−ク幅に拡げられている。
すなわち、ピ−ク６２に影響する拡がり関数は、ピ−ク６１と共通の拡がり関数とそれ以外の余剰の拡がり関数とのコンボリュ−ションになっている。
【００３９】
この場合、ピ−ク６２の拡がり関数からピ−ク６１の拡がり関数の影響を取り除き、余剰の拡がり関数だけを得ることができる。この処理をディコンボリュ−ション処理と呼ぶ。
【００４０】
前記ディコンボリュ−ション演算は、図２に示すピ−ク６４とピ−ク６５とのような重なりピ−クに対して効果がある。
図４を参照して説明する。図４は、図２に示す重なりピ−クに対しディコンボリュ−ション演算を説明している。
【００４１】
前記ピ−ク６４波形と前記ピ−ク６５波形とに対して前記ピ−ク６１波形の拡がり関数をディコンボリュ−ション演算すると、図４に示すような分離された余剰分のピ−クが得られ、より正確にそれぞれのピ−クの面積が計算できる。
実際は、ピ−ク波形６１を面積１に規格化した規格化関数をディコンボリュ−ション演算に利用する。規格化はピ−ク面積を保存するために行われる。
【００４２】
厳密に言えば、この規格化関数は、（式）２の注入時の長方形関数ｒ（ｔ）とピ−ク波形６１に影響する拡がり関数ｈ（ｔ）の規格化されたコンボリュ−ションである。
実際には、ピ−ク波形６１以外のピ−クにも長方形関数の影響があるため、ピ−ク６１波形の規格化関数でディコンボリュ−ション演算しても問題にはならないわけである。
【００４３】
以上の図４のディコンボリュ−ション手順を図１のマイクロボアＨＰＬＣシステムのブロック図と、図５に示す検出波形処理のフロ−チャ−トとを参照して説明する。
まず、非保持成分を添加した標準試料５２が注入され、工程３１にて波形処理が開始される。
【００４４】
次に、工程３２で非保持ピ−ク６１を同定する。これはクロマトグラムの最初の時刻に現われるある大きさ以上のピ−クとして自動的に同定するか、またはオペレ−タが入力部５４から時刻指定を行うか、ＣＲＴ４６に表示しているクロマトグラムからマウスを使用し、ピ−ク６１をピックアップすることで指定することができる。
【００４５】
工程３３において、非保持ピ−ク６１の面積を１に規格化する。この時、ＣＲＴ４６に規格化関数を表示することもできる。
工程３４において、オペレ−タが正確に定量したい重なりピ−クのある時間の始点と終点を設定する。前記ＣＲＴ４６には、図４の左図のように、この部分だけを拡大したクロマトグラムを表示することもできる。ここでオペレ−タはクロマトグラムの全ての時間を設定することも可能である。
【００４６】
工程３５において規格化関数を用いて、設定された時間区間内でディコンボリュ−ション処理を行う。
工程３６において処理前後のクロマトグラムをＣＲＴ４６に図４の右側と左側とのように表示する。また、図６に示す如く、プリンタ４７に重ね描きしたクロマトグラムをプロットすることもできる。
【００４７】
工程３７において、シャ−プになった図４の右側のピ−クを対象に積分を実行し、定量計算を行う。
この処理後のクロマトグラムでもまだピ−クが重なっている場合には、従来通りの垂直分割等で面積計算する。
ここでガウシアン、ＥＭＧ等の非線形関数で最小自乗法により面積分割することもできる。ここでＣＲＴ４６かプリンタ４７に定量値を出力する。
工程３８において処理を終了する。
【００４８】
ディコンボリュ−ション処理をする拡がり関数を標準試料５２中の非保持成分のピ−ク波形の規格化関数で代用したが、この方法の他にも拡がり関数を得ることができる。
標準試料に添加する代わりに、予め拡がり関数を求めるために非保持成分だけ注入し、ピ−ク波形を得ることができる。
【００４９】
また、逆相クロマトグラフィ−ならメタノ−ル，アセトニトリル，ＴＨＦ，クロロホルム等有機溶媒１００％の溶離液を流し、保持の弱い成分を注入し、非保持ピ−クを得ることができる。
イオン交換クロマトグラフィ−では、例えばアミノ酸分析計の場合、ｐＨの高い再生液を流し、保持の弱いホスホセリン等を注入し、そのピ−ク波形を得ることができる。
【００５０】
また、カラム充填剤を操作し、非保持ピ−クを得ることもできる。イオン交換クロマトグラフィ−なら、イオン交換基の修飾しない坦体だけの充填剤を作成し、非保持ピ−クを得ることもできる。
【００５１】
ポストカラム反応ＬＣの実施例としてアミノ酸分析計をあげる。このシステムもカラム外の拡がりを無視できない。カラム外の拡がりは、カラムをジョイントに交換し、非保持ピ−クの波形を拡がり関数として見積ることができる。
【００５２】
図７に示す如く検出されたクロマトグラムにおいては、スレオニン（Ｔｈｒ）とセリン（Ｓｅｒ）の重なり部分が１４％であるのに対し、ディコンボリュ−ション処理されたクロマトグラムでは４％と１／３以下に向上した。
この処理クロマトグラムは、ポストカラム反応ＬＣにおいて、カラムの直後で検出できたとした場合のクロマトグラムに相当している。
【００５３】
拡がり関数の共通部分の除去例として、３．３分間処理のグリコヘモグロビン分析計をあげる。
拡がり関数を図８の左側に示すＨｂＦのピ−ク波形から見積ることができる。これはＨｂＡ１ｃのピ−ク幅が、リテンションタイムの順序を考慮し、ＨｂＦの幅よりも若干広いと考えられるためである。
【００５４】
図８の左側の検出クロマトグラムより孤立したピ−クであるＨｂＦの波形を規格化し、拡がり関数を得る。その規格化関数を用いて、不安定型と安定型とのＨｂＡ１ｃの重なりピ−クをディコンボリュ−ション処理すれば、図８の右側に示すような分離の向上したクロマトグラムが求められるわけである。
【００５５】
ディコンボリュ−ション処理は、この場合処理クロマトグラムを直接表示する必要はないが、定量精度を向上するためには、一般的なＨＰＬＣであるグリコヘモグロビン分析計においても非常に有効な手法である。
【００５６】
次にディコンボリュ−ション演算を正しく行うために演算前に利用すべき処理について説明する。図２に示されているノイズ６７やドリフト６８がディコンボリュ−ション演算を妨害する。
これらをそのままディコンボリュ−ション処理すると、擬似的なピ−クとして増幅されたり、ディコンボリュ−ション処理されるべきピ−クが正しく処理されないといった弊害が起こる。
【００５７】
まず、ノイズ除去処理としては一般的に検出クロマトグラムのデ−タ点をいくつか足し算して低減する方法が採られる。
この他に平滑化の方法や公知の高周波成分を除去する方法も利用できる。
【００５８】
ドリフト除去処理としては一般的にピ−ク間の谷から谷を直線で結だり、強制的に水平に直線を描き、検出クロマトグラムからベ−スラインを引き算する方法が採られる。
この他に公知の低周波成分を除去する方法，微積分を利用し低次の多項式成分を除去する方法、微分の特徴点からベ−スラインを探索する方法，ニュ−ラルネットワ−クを利用する方法等がある。
【００５９】
ここまでは拡がり関数を測定デ−タより求めてきたが、拡がり関数が未知である場合でも、ガウシアンをディコンボリュ−ト演算することができる。
前述したように検出クロマトグラムのピ−ク波形は、（２）式で示されるように、拡がり関数ｈ（ｔ）と長方形波形のｒ（ｔ）とのコンボリュ−ションで表わせる。
【００６０】
この拡がり関数ｈ（ｔ）は、前述の“ＤＹＮＡＭＩＣＳ ＯＦ ＣＨＲＯＭＡＴＯＧＲＡＰＹ”Ｊ．Ｃ．Ｇｉｄｄｉｎｇｓ著により、カラム内の拡がりがどのような物理学的，化学的相互作用に起因していても、カラム外の拡がりが無視できないくらい大きくとも、ガウシアン様の形状をしている。
【００６１】
ｈ（ｔ）がリ−ディンク的かまたはテ−リング的にガウシアンから多少歪んでいたとしても、下記の（４）式のようにガウシアンＧ（ｔ）と修正関数ｍ（ｔ）とのコンボリュ−ションとして記述できる。
【数４】

【００６２】
上式において、ガウシアンをＧ（ｔ）、修正関数をｍ（ｔ）、σは標準偏差である。
例として、ｈ（ｔ）がＥＭＧ（ｔ）により表わせるテ−リングピ−クであるなら、前述の（３）式のように解析数学的に修正関数が求められる。
【００６３】
また、重なりピ−クは、（５）式のように表わせる。
【数５】

ここで、σ_１、σ_２はそれぞれ標準偏差である。
【００６４】
ガウシアンは、上記の如く、加成性があるため、２つのガウシアンのコンボリュ−ションに分解できるため、下記（６）式のような関係を持っている。
【数６】

【００６５】
つまり、（５）式で表わされている２つのピ−ク波形はガウシアンを用いて、ディコンボリュ−ション演算ができる。
この時そのガウシアンは、２つのピ−クの持つ標準偏差σ_１とσ_２よりも小さな標準偏差σ０を持つ幅の狭い波形でなければならない。
【００６６】
こうして、検出された重なりピークよりも幅の狭いピークを得ることができる。
最終的には、ｙ（ｔ）をガウシアンＧ（ｔ，σ_０）によりディコンボリューション演算すると、クロマトグラムは、幅のより狭いガウシアンともとのままの修正関数とのコンボリューションにより表わすことができるわけである。
【００６７】
【数７】

上式の〔 〕内のように、σ_１，σ_２よりピーク幅の狭いピークが得られる。より詳しくいえば、注入時の長方形波形はこの修正関数によりコンボリュ−ション演算されている。
【００６８】
この拡がり関数ｈ（ｔ）が未知の場合の処理を図９を参照して説明する。
工程７１で処理を開始する。
工程７２で検出クロマトグラムからノイズ・ドリフトを除去し、ディコンボリュ−ション処理の準備をする。
工程７３でディコンボリュ−ション処理を手動で行なうか、自動で行なうのかを入力する。
【００６９】
自動の場合、工程８１で、検出クロマトグラムのピ−ク幅を全体的に見て、最小の幅よりもやや小さな標準偏差を持つガウシアンを拡がり関数として設定する。これは面積１に規格化されている。
ここで、リテンションタイムの短いピ−クと長いピ−クでピ−ク幅が２倍以上異なる場合は、時間を分割してディコンボリュ−ション演算することもできる。
【００７０】
ここで自動的にピ−ク幅を見つけ出すためにフ−リエ変換を利用することもできる。工程７６でディコンボリュ−ション処理を実行する。工程７７で処理クロマトグラムが発散していないか、負の下限値を下回っていないか判定する。また不分離ピ−クがピ−ク間の谷を形成しているか等を判定する。
【００７１】
ここで判定が合格となれば、工程７９へ進む。判定が不合格の場合、もう一度工程８１へ戻り、判定を参照し標準偏差を加減し、ディコンボリュ−ション処理を行なう。
判定合格の場合、工程７９で処理クロマトグラムの表示，定量結果の表示等を行なう。工程８０で終了する。
【００７２】
手動の場合、工程７４で入力操作を行う。一般的にはディコンボリュ−ション処理するガウシアンの標準偏差を入力する。他に処理区間の始・終点の入力もできる。また特定の重なりピ−クの分離を向上するために、ガウシアンの最適な標準偏差を探索することもできる。
【００７３】
この場合、標準偏差の初期値，推定されるピ−クの個数，大きさ，幅，保持時間等を参考値として入力することもできる。また制限値として、ピ−ク個数は３個までとか、重なりピ−クの幅が０．２分以上０．４分以下のように入力することもできる。
【００７４】
工程７５で工程７４にて入力された情報を演算し、ディコンボリュ−ション処理の設定値を求め、判定の基準値を算出する。工程７６でディコンボリュ−ション処理を実行する。工程７７で判定を行う。
不合格の場合、その理由を工程７８で表示し、工程７４で再入力を促す。合格の場合、工程７９で処理クロマトグラムと定量結果を表示する。
工程８０で終了する。
【００７５】
カテコ−ルアミン分析計により得られた生の１０分間のクロマトグラムを図１０の上段（ａ）に示す。
これは、血漿や尿のような生体試料の分析例である。生体試料を分析すると数多くのピ−クがしばしば現われる。これを適切なガウシアンでディコンボリュ−ション処理すると、中段（ｂ）のようなクロマトグラムが得られる。
各ピ−クがベ−スラインまで完全に分離している。
【００７６】
しかし、これをそのまま表示しても、処理クロマトグラムの意味するところは理解困難である。
むしろ、これを下段の（ｃ）に示した公知の表示法にしても差し支えない。
また、定量値のみを出力するだけで、ディコンボリュ−ション処理をオペレ−タ−に意識させないのがよいと考えられる。
【００７７】
以上、本発明の一実施例について説明したが、本発明は、上記実施例に限定されるものでなく、種々の変形例が考えられる。
本発明に係るクロマトグラム解析方法の他の一方法を説明する。
図１１は、本発明に係る解析方法におけるディコンボリュ−ション、リィコンボリュ−ションによる重ね描きの説明したものである。
【００７８】
図１１（１）は、検出されたオリジナルの波形である。このオリジナルの波形にディコンボリュ−ション処理し、各波形ピークを分離・独立させる。この状態を図１１（２）に示す。
分離・独立させた波形ピークをそれぞれディコンボリュ−ションしたときと同じ拡がり関数ｈ（ｔ）によりリコンボリュ−ションする。
リコンボリュ−ション演算された波形を検出されたオリジナルの波形に重ね描きをする。図１１（３）は、この状態を示したものである。
【００７９】
〔実施例 ２〕
本発明に係るクロマトグラフ解析装置の他の一実施例を説明する。図１２は本発明に係る他の一実施例であるクロマトグラフ解析装置の構成を示すブロック図である。
本発明に係るクロマトグラフ解析装置の全体構成については、〔実施例 １〕において、図１のマイクロボアＨＰＬＣシステムについて説明したので、煩瑣となるので再度の説明を省略し、データ解析部４５のみを説明する。
【００８０】
前記データ解析部４５は、ノイズ・ドリフト除去機構と、拡がり関数解析機構と、ディコンボリューション演算機構と、定量計算機構と、判定機構とから構成されている。図中、図示されてないが、Ａ／Ｄ変換器、記憶装置、インタフェース部等も備えていることはいうまでもない。
前記構成の内、ノイズ・ドリフト除去機構、定量計算機構、判定機構は、公知の技術であるが順序としてあわせて説明をする。なお、前記各機構部は、バスによって接続されている。
【００８１】
前記ノイズ・ドリフト除去機構は、平滑化機構と、ベースライン算出機構と、高速フーリエ変換機構と、ニューラル・ネットワーク機構と、微分・積分機構と、ファジイ推論機構とを備えており、前記平滑化機構は、検出器４３からのクロマトグラムの測定データ群を移動平均法、Ｓａｖｉｔｚｋｙ−Ｇｏｌａｙ法、Ｋａｗａｔａ−Ｍｉｎａｍｉ法、周波数領域法のいずれかにより滑らかな曲線にフイットし、前記ベースライン算出機構は、適切なアルゴリズムによりベースラインを決定する。
【００８２】
前記高速フーリエ変換機構は、高周波成分であるノイズおよび低周波成分であるベースライン変動をそれぞれ周波数分布に基づき除去する。
また、前記ニューラル・ネットワーク機構は、ドリフト形状を予め学習しておき、測定データから所定の結合定数に従い、ドリフトを除去する。
【００８３】
前記微分・積分機構は、微分・積分により低次の多項式で表現されるベースライン成分を除去する。この整理された測定データ点群をファジイ推論機構が所定の結合定数に従い平滑化する。このようにして検出器４３からのクロマトグラムのノイズおよびドリフトは除去される。
【００８４】
前記拡がり関数解析機構は、Ｇａｕｓｓｉａｎ発生機構と、面積１規格化機構と、検出ピーク波形設定機構とを備え、解析数学的拡がり関数の場合は、前記Ｇａｕｓｓｉａｎ発生機構によりＧａｕｓｓｉａｎを発生する。実測的拡がり関数の場合は、前記面積１規格化機構により実測されたピーク波形を規格化し、前記検出ピーク波形設定機構により前記規格化された実測波形を拡がり関数としている。このようにして実測的ないし解析数学的拡がり関数を発生する。
【００８５】
前記ディコンボリューション演算機構は、時間区間設定機構と、拡がり関数設定機構と、ディコンボリューション処理機構とを備え、前記時間区間設定機構は、ディコンボリューション処理する区間を設定し、前記拡がり関数設定機構は、上記の発生させた実測的ないし解析数学的拡がり関数から拡がり関数を設定する。
前記ディコンボリューション処理機構は、設定されたディコンボリューション処理する区間の重なり波形を前記拡がり関数によりディコンボリューション処理し、重なり波形を分離・独立させる。
【００８６】
前記定量計算機構は、積分機構を備え、前記積分機構は、前記分離・独立させたピーク波形の面積をそれぞれ算出する。このようにしてピーク面積ないし高さに基づき定量を計算する。前記定量結果は、ＣＲＴ４６、プリンタ部４７に出力される。
【００８７】
各種の判定を行う判定機構は、ディコンボリューション処理の成功を判定するために利用する判別式の値を計算する判別式計算機構、ディコンボリューション処理するために利用するＧａｕｓｓｉｎを探索するときに制限となる値を算出する制限値算出機構と、所定の結合係数に従いその最適なＧａｕｓｓｉｎを探索する拡がり関数探索機構を具備し、上記各処理段階において、各機構が所定の働きが可能となるようサポートしている。
【００８８】
図１２における制御部４４、データ解析部４５、ＣＲＴ部４６、プリンタ部４７等の機能を一体化して、Ａ／Ｄ変換部と、メモリ部と、制御部と、インタフェース部と、出力機器とからなり、クマトグラムの検出ピ−ク波形から拡がりを表わす関数ｈ（ｔ）を算出・出力し、検出ピ−ク波形および前記関数ｈ（ｔ）を用いて前記検出ピ−ク波形をディコンボリューション処理し、ディコンボリューション処理後のクロマトグラムを出力することを特徴とするクロマトグラム解析方法とクロマトグラフ装置とに用いられるデ−タ処理装置を構成しても差し支ええない。
【００８９】
【発明の効果】
以上詳細に説明したように、本発明の構成によれば、クロマトグラムの波形の歪を補正し、重なりのピ−ク波形をシャ−プにすることにより分離を向上させ、ピ−ク面積をより正確に求められ、定量分析の精度が改良されたクロマトグラム解析方法を提供することができる。
【図面の簡単な説明】
【図１】本発明における一実施例に係る解析方法に用いるマイクロボアＨＰＬＣシステムの構成を示すブロック図である。
【図２】図１の解析方法におけるクロマトグラムを示す線図である。
【図３】図１の解析方法におけるクロマトグラムの拡がりメカニズムの略示説明図である。
【図４】図１の解析方法におけるクロマトグラムの重なりピークに対するディコンボリュ−ション演算説明図である。
【図５】図１の解析方法に係る処理工程のフロ−チヤ−トである。
【図６】図１の解析方法におけるクロマトグラムの重ね描きプロット図である。
【図７】図１の解析方法によるアミノ酸分析のクロマトグラムを示す線図である。
【図８】図１の解析方法によるグリコヘモグロビン分析のクロマトグラムを示す線図である。
【図９】図１の解析方法において拡がり関数が未知の場合の処理フロ−チャ−トである。
【図１０】図１の解析方法におけるカテコ−ルアミン分析のクロマトグラムを示す線図である。
【図１１】図１の解析方法におけるリィコンボリューション演算による重ね描き説明図である。
【図１２】本発明に係る他の一実施例に係るクロマトグラフ解析装置の構成を示すブロック図である。
【符号の説明】
４０…ポンプ
４１…カラム
４２…検出器
４３…サンプラ
４４…制御部
４５…デ−タ解析部
４６…ＣＲＴ表示部
４７…プリンタ
４９…溶離液
５１…未知試料
５２…標準試料
５４…入力部[0001]
[Industrial applications]
The present invention relates to chromatography techniques such as liquid chromatography and gas chromatography, and more particularly to a chromatogram analysis method (hereinafter referred to as an analysis method) and a chromatographic apparatus used therein, and data used in the analytical method and the chromatographic apparatus. The present invention relates to a processing device.
[0002]
[Prior art]
Prior to the description of the conventional technology, technical terms are defined.
In the present application, the convolution is defined as a convolution integral of two functions x (t) and y (t). In general, an output waveform from a measuring instrument is an input waveform and an apparatus function (an impulse response in an electric circuit). , And a convolution with a spectroscope).
Deconvolution is defined as the inverse operation of convolution. Specifically, this corresponds to solving an integral equation represented by convolution.
The reconvolution means that a waveform deconvolved using a device function, that is, a function h (t) representing a spread is convoluted again using the function h (t).
[0003]
Conventionally, generally, output waveform data from a measuring instrument is distorted by the dynamic characteristics of the measuring instrument, that is, the influence of the device function.
The same applies to a chromatographic apparatus, in which distortion occurs on the chromatogram and peaks overlap on the chromatogram due to the apparatus function, that is, the function h (t) representing the spread.
In such a case, it has been difficult to accurately determine the peak area of each chromatogram.
[0004]
Conventionally, when the above-mentioned distortion is applied, the overlap of the peaks is decomposed by a deconvolution process using a synthesis separation method based on a normal curve fitting method, and is subjected to data processing. It has been proposed to quantify.
A general solution to this problem is described in "Waveform Data Processing for Scientific Measurement", edited by Shigeo Minami, published by CQ Publishing Co., Ltd. (1986).
Also, many improvement techniques have been proposed, for example, the technique described in JP-A-62-17465 and the technique described in JP-A-63-151851.
[0005]
In the technique described in Japanese Patent Application Laid-Open No. Sho 62-17465, the half-width on the rising side is approximated by Gaussian, which is a predetermined multiple of the distance from the rising point to the end point. In the case of peak waveforms, there are few cases where such approximation can be used for replacement.
Although this technique is effective in reducing the storage capacity, it has a drawback that it is not necessarily effective in accurately obtaining the peak area when the peak waveforms overlap.
[0006]
In the technique described in Japanese Patent Application Laid-Open No. 63-151851, the overlapped peak is divided into two Gaussians, which has been so-called vertical division and skimming division, and is approximated by a method of least squares. Although the deconvolution process is performed by the technique, there is a disadvantage that a physically meaningful solution cannot be obtained.
[0007]
Further, in the techniques described in U.S. Pat. Nos. 4,807,148 and 4,941,101, a numerical analysis technique is used in the deconvolution processing, and there are Jansson's method, Jacobi method, Gauss-Seidel method, and the like. Although the number of stored words is advantageous, there is a disadvantage in that the number of discontinuous points in the solution is increased.
[0008]
[Patent Document 1]
JP-A-62-17465
[Patent Document 2]
JP-A-63-151851
[Patent Document 3]
U.S. Pat. No. 4,807,148
[Patent Document 4]
U.S. Pat. No. 4,941,101
[Non-patent document 1]
Minami Shigeo, "Waveform Data Processing for Scientific Measurement", CQ Publishing Co., Ltd., (1986)
[0009]
[Problems to be solved by the invention]
The various conventional deconvolution processes described above provide respective solutions for the deconvolution process when the chromatogram is distorted and the synthesis and separation when the chromatogram has an overlapping peak waveform. However, in the case where an overlapping peak waveform is generated due to the distortion, it has been studied in detail up to the point of separating the overlapping waveform in consideration of the distortion and has not been dealt with.
[0010]
Further, the above-mentioned conventional synthesis separation method for the overlap of chromatogram peak waveforms has the following problem.
1. Assuming a non-linear function such as Gaussian, the validity of selecting that function is questionable.
2. Introducing multi-channel information such as the UV spectrum requires an expensive detector such as a photodiode array detector.
3. In the peak decomposition, since the least squares method is used, a long calculation time is required until a solution is obtained, and a physically meaningful solution may not be obtained in some cases.
[0011]
SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art, and improves the separation by correcting waveform distortion of a chromatogram and sharpening overlapping peak waveforms. The present invention aims to provide a chromatogram analysis method in which the work area is more accurately determined and the accuracy of quantitative analysis is improved.
[0012]
[Means for Solving the Problems]
In order to achieve the above object, a configuration of the present invention relating to a chromatogram analysis method includes a chromatogram analysis method in which a sample to be analyzed having passed through a separation column is detected by a detector, and a chromatogram obtained by the detection is analyzed. At
A step of introducing a standard sample not retained by the separation column into the separation column, separately from introducing the sample to be analyzed into the separation column,
Identifying a peak of the standard sample from the chromatogram;
Setting a start point and an end point of a certain time of the peak of the sample to be analyzed from the chromatogram,
Performing a deconvolution process on the data within the set time interval using a peak waveform of the identified standard sample.
In the chromatogram analysis method described above,
When the peaks after performing the deconvolution processing are a plurality of peaks and remain in an overlapping state, perform vertical division or area division using a nonlinear function on the overlapping peaks. It is characterized by.
[0013]
In order to achieve the above object, another configuration of the present invention according to the chromatogram analysis method includes detecting a sample to be analyzed that has passed through a separation column with a detector, and standard deviation of a peak in the chromatogram obtained by the detection. In a chromatogram analysis method of performing analysis using a Gaussian function having σ as a coefficient,
Setting an arbitrary value to the standard deviation σ of the Gaussian function;
Performing a deconvolution process on the chromatogram;
Regarding the chromatogram after the deconvolution process, the chromatogram does not diverge, does not fall below the negative lower limit, or an unseparated peak forms a valley between peaks, and satisfies one of the following criteria. Determining whether or not
Adjusting the standard deviation value of the Gaussian function when the determination is unsuccessful, and performing deconvolution processing again.
In the chromatogram analysis method described above,
Using the spreading function used in the deconvolution processing, a waveform after the reconvolution processing is displayed on the peak separated by the deconvolution processing so as to overlap the peak waveform before the deconvolution processing. .
[0014]
More specifically, the chromatogram peaks are spread for various reasons. For example, when a sample flows through a flow path such as a pipe or a flow cell of a detector, the sample zone is expanded. This is called extra-column spreading.
The contribution due to the spread typified by the spread outside the column is convoluted to the original peak waveform and appears in the detected waveform.
That is, if the spread function is known, the original peak waveform can be calculated by performing a deconvolution operation on the detected waveform.
[0015]
There are various methods for determining the spread function h (t).
When the methods are roughly classified into two, there are a method of experimentally obtaining from detected data and a method of assuming an analytical mathematical h (t) having some parameters characterizing a waveform.
The former can be considered as a method in which h (t) is known, and the latter can be considered as a method in which h (t) is unknown.
[0016]
As a method for obtaining the h (t) from the measurement, for example, a method of exchanging the column with a joint, injecting a sample, and regarding the waveform measured by the detector as h (t) in order to obtain the spread outside the column is available. is there.
If the extension in the column that does not depend on the type of components in the sample is also added, a sample not retained can be injected, and the peak waveform can be regarded as h (t).
[0017]
Other methods for obtaining peaks that are not retained include the use of a solvent having an extremely high elution ability as an eluent, and the preparation of a column packing material that does not modify a functional group so as not to retain it.
It is also effective to add a component not retained in the standard sample to obtain a non-retained peak from the chromatogram of the standard sample.
[0018]
If h (t) is unknown, the spread function h (t) corresponding to a part of the peak width is deconvoluted under the following conditions, and a sharper peak is obtained. It can be accurately quantified based on the peak area from the processed chromatogram.
[0019]
The precondition is that the detected peak waveform is originally a convolution operation of a spread function with respect to a delta function or a very sharp rectangular density distribution function. Can be
Here, the spread function is almost Gaussian, which has a waveform slightly modified in a tailing or reading manner.
In other words, it can be considered that the spreading function itself is a convolution operation of the tailing or reading correction function for Gaussian.
[0020]
Further, Gaussian has an additive property that when two Gaussians are subjected to a convolution operation, they are also Gaussians.
As described above, by performing the deconvolution operation on the Gaussian constituting a part of the spread function, it is possible to calculate a peak approaching a sharper original waveform.
[0021]
[Action]
The function of each of the above technical means is as follows.
According to the configuration of the present invention, a known h (t) or Gaussian is rationally selected from the detected chromatogram, and deconvolution operation is performed to obtain a peak sharper than the overlapping peak of the detected chromatogram. When the peaks are separated and separated into chromatograms having peaks, and the separated and independent peaks are integrated, the peak area is accurately obtained, and accurate quantification can be performed.
In the deconvolution operation, unlike the least square method of a nonlinear function, for example, a physically meaningful solution can always be obtained in a short calculation time.
[0022]
In the deconvolution operation, when the overlapping peaks are separated and separated, the chromatogram is observed to judge the validity of the operation, but it requires skill. Therefore, the independent peaks can be subjected to the reconvolution operation using the spread function h (t) used for the deconvolution operation, and each peak constituting the first overlapping peak can be reproduced. By overlaying each peak after the reconvolution operation and the first overlapping peak, the validity of the deconvolution operation can be accurately determined.
[0023]
Further, according to the configuration of the present invention, an expensive detector such as a photodiode array detector is omitted, and the chromatogram analysis method can be performed with a simple configuration.
[0024]
【Example】
[Example 1]
Hereinafter, an embodiment according to the present invention will be described with reference to FIGS.
FIG. 1 is a block diagram showing a configuration of a microbore HPLC system used in a chromatogram analysis method (hereinafter, referred to as an analysis method) according to one embodiment of the present invention, and FIG. 2 shows a chromatogram in the analysis method of FIG. FIG. 3 is a schematic explanatory view of a spread mechanism of a chromatogram in the analysis method of FIG. 1, FIG. 4 is an explanatory view of deconvolution calculation in the analysis method of FIG. 1, and FIG. 5 is an analysis method of FIG. FIG. 6 is a plot plot of the chromatogram obtained by the analysis method shown in FIG.
[0025]
FIG. 7 is a diagram showing a chromatogram of amino acid analysis by the analysis method of FIG. 1, FIG. 8 is a diagram showing a chromatogram of glycohemoglobin analysis by the analysis method of FIG. 1, and FIG. 9 is a diagram of FIG. FIG. 10 is a processing flowchart in the case where the spread function is unknown in the analysis method, FIG. 10 is a diagram showing a chromatogram of catecholamine analysis in the analysis method of FIG. 1, and FIG. 11 is a reconvolution in the analysis method of FIG. FIG. 18 is an explanatory diagram of superposition by a calculation operation.
[0026]
In order to carry out the chromatogram analysis method according to the present invention, a microbore HPLC system is used. FIG. 1 shows the overall configuration of the microbore HPLC system.
In FIG. 1, 40 is a pump, 41 is a column, 42 is a detector, 43 is a sampler, 44 is a controller, 45 is a data analyzer, 46 is a CRT display, 47 is a printer, 49 is an eluent, and 51 is an eluent. Is an unknown sample, 52 is a standard sample, and 54 is an input unit.
[0027]
Except for the data analysis unit 45 (described later in [Embodiment 2]), each unit constituting the present system uses a normal chromatogram analysis method. Therefore, detailed description is omitted because it is complicated, and will be described briefly.
[0028]
In the microbore HPLC system, the spread outside the column greatly affects the separation and cannot be ignored. The pump 40 sends the eluent 49 according to a command from the control unit 44. The sampler 43 having a movable needle valve injects the unknown sample 51 or the standard sample 52 into the flow path to the column 41.
The sample is sent to the column 41 together with the eluent 49, and the contained components are separated and developed, and detected by the detector 42.
[0029]
The detected chromatogram is stored in the data analyzer 45.
FIG. 2 is the detected chromatogram. Since a non-retained component is added to the standard sample 52, it is detected as a non-retained peak 61. The peak 62 to be eluted next is a component of the standard sample, and unlike the peak 61, is retained in the column and has a slightly wider peak width.
[0030]
With reference to FIG. 3, a mechanism for forming a peak waveform will be described.
FIG. 3 is an explanatory diagram of a mechanism for expanding a chromatogram in the analysis method according to the embodiment of FIG.
In FIG. 3, the same reference numerals as those in FIG. 1 denote the same parts, and a detailed description thereof will be omitted.
The upper diagram (a) is a block diagram showing a sample flow channel, and the lower (b) diagram shows the sample concentration in the sample flow channel of the upper (a) diagram.
[0031]
A sample containing a single component is injected into the channel by the sampler 43. The spatial distribution diagram of the sample concentration C in the channel at this time is a rectangular distribution diagram 71, where x is the flow direction.
When this rectangular distribution diagram 71 passes through the column 41, it expands. The stage after passing through the column 41 has a Gaussian shape as shown in a distribution diagram 72.
[0032]
Further, when the concentration distribution of Gaussian is detected by the detector 42, the concentration distribution is expanded by a flow path pipe, a flow cell of the detector, or the like. line The distribution shape is wider than that in FIG.
[0033]
If this is described by a mathematical expression, it can be represented by a model as shown in Expression (2).
(Equation 2)

In the above equation, the detected waveform at time t is y (t), the original rectangular waveform at the time of sample injection is r (t), and the spread function is h (t). In the above equation, the detected waveform y (t) is a convolution of the original rectangular waveform r (t) and the spread function h (t).
[0034]
The spread of the peak shape may be described above. However, in consideration of the retention time, it is considered that another Dirac delta function is subjected to the convolution operation. This is shown in equation (3).
(Equation 3)

[0035]
In the above equation, EMG (t) is Exponentially Modified Gaussin, and σ, τ, t _R Is a parameter of EMG (t), δ (t) is a delta function, and m (t) is a correction function.
The mechanism of spreading is described in "DYNAMICS OF CHROMATOGRAPHY", J. Amer. C. Giddings, published by Marcel Dekker, New York, 1965;
[0036]
The spreading function h (t) is formed by the influence from both the spreading inside the column and the spreading outside the column. That is, a convolution of the spread function inside the column and the spread function outside the column.
[0037]
Focusing on the non-holding peak 61 in FIG. 2 and referring to FIG. 3, a rectangular waveform 71 at the time of injection passes through the column 41, is expanded into a Gaussian waveform 72, and spread outside the column. As a result, the waveform becomes a waveform 73 and is detected as a peak 61.
[0038]
On the other hand, the holding peak 62 is also expanded by a similar mechanism to form a peak. However, since the holding peak 62 is held in the column 41, it is expanded to a slightly wider peak width. I have.
That is, the spread function affecting the peak 62 is a convolution of the spread function common to the peak 61 and the other surplus spread functions.
[0039]
In this case, it is possible to remove the influence of the spread function of the peak 61 from the spread function of the peak 62 and obtain only the surplus spread function. This process is called a deconvolution process.
[0040]
The deconvolution operation is effective for overlapping peaks such as peak 64 and peak 65 shown in FIG.
This will be described with reference to FIG. FIG. 4 illustrates the deconvolution operation for the overlapping peaks shown in FIG.
[0041]
When the spread function of the peak 61 waveform is deconvoluted with respect to the peak 64 waveform and the peak 65 waveform, a separated surplus peak as shown in FIG. 4 is obtained. Thus, the area of each peak can be calculated more accurately.
Actually, a normalization function obtained by normalizing the peak waveform 61 to the area 1 is used for the deconvolution operation. Normalization is performed to conserve peak area.
[0042]
Strictly speaking, this normalization function is a standardized convolution of the rectangular function r (t) at the time of injection of (Equation 2) and the spread function h (t) affecting the peak waveform 61. .
Actually, since the peak function other than the peak waveform 61 is also affected by the rectangular function, it does not matter if the deconvolution operation is performed by the normalization function of the peak 61 waveform.
[0043]
The above-described deconvolution procedure of FIG. 4 will be described with reference to the block diagram of the micro-bore HPLC system of FIG. 1 and the flowchart of the detection waveform processing shown in FIG.
First, the standard sample 52 to which the non-retained component is added is injected, and the waveform processing is started in step 31.
[0044]
Next, in step 32, the non-holding peak 61 is identified. This is automatically identified as a peak of a certain size or more that appears at the first time of the chromatogram, or the operator specifies the time from the input unit 54, or from the chromatogram displayed on the CRT 46. It can be designated by picking up the peak 61 using a mouse.
[0045]
In step 33, the area of the non-holding peak 61 is normalized to one. At this time, the standardization function can be displayed on the CRT 46.
In step 34, the operator sets the starting point and the ending point of the time at which there is an overlapped peak to be accurately quantified. On the CRT 46, a chromatogram in which only this portion is enlarged can be displayed as shown in the left diagram of FIG. Here, the operator can also set all the times of the chromatogram.
[0046]
In step 35, a deconvolution process is performed within a set time interval using a normalization function.
In step 36, the chromatograms before and after the processing are displayed on the CRT 46 as shown on the right and left sides of FIG. In addition, as shown in FIG. 6, a chromatogram overlaid on the printer 47 can be plotted.
[0047]
In step 37, integration is performed on the sharp peak on the right side of FIG. 4 to perform quantitative calculation.
If peaks still overlap in the chromatogram after this processing, the area is calculated by a conventional vertical division or the like.
Here, it is also possible to divide the area by the least square method using a nonlinear function such as Gaussian or EMG. Here, the quantitative value is output to the CRT 46 or the printer 47.
In step 38, the process ends.
[0048]
Although the spread function for the deconvolution processing is substituted by the normalization function of the peak waveform of the non-retained component in the standard sample 52, a spread function can be obtained in addition to this method.
Instead of adding to the standard sample, only the non-retained component is injected in order to obtain a spread function in advance, and a peak waveform can be obtained.
[0049]
In the case of reversed-phase chromatography, an eluent containing 100% of an organic solvent such as methanol, acetonitrile, THF, chloroform or the like is flowed, and a component having a low retention is injected to obtain a non-retention peak.
In ion exchange chromatography, for example, in the case of an amino acid analyzer, a regenerated solution having a high pH is allowed to flow, and phosphoserine or the like having a low retention is injected to obtain a peak waveform.
[0050]
In addition, a non-retained peak can be obtained by manipulating the column packing material. In the case of ion-exchange chromatography, it is also possible to prepare a packing consisting only of a carrier without modification of the ion-exchange group to obtain a non-retained peak.
[0051]
An example of the post-column reaction LC is an amino acid analyzer. This system cannot ignore the spread outside the column. Spread outside the column can be estimated by exchanging the column with a joint and estimating the waveform of the unretained peak as a spread function.
[0052]
In the chromatogram detected as shown in FIG. 7, the overlapping portion of threonine (Thr) and serine (Ser) is 14%, whereas in the chromatogram subjected to the deconvolution treatment, it is 4% or less than 1/3. Improved.
This processed chromatogram corresponds to a chromatogram in the case where detection was possible immediately after the column in the post-column reaction LC.
[0053]
As an example of removing the common part of the spread function, a glycated hemoglobin analyzer processed for 3.3 minutes will be described.
The spread function can be estimated from the peak waveform of HbF shown on the left side of FIG. This is because the peak width of HbA1c is considered to be slightly wider than the width of HbF in consideration of the order of the retention times.
[0054]
The waveform of HbF, which is an isolated peak, is normalized from the detection chromatogram on the left side of FIG. 8 to obtain a spread function. If the overlapping peaks of the unstable type and the stable type HbA1c are subjected to deconvolution processing using the standardized function, a chromatogram with improved separation as shown on the right side of FIG. 8 is obtained.
[0055]
In this case, it is not necessary to directly display the processed chromatogram, but the deconvolution treatment is a very effective technique in a glycohemoglobin analyzer which is a general HPLC in order to improve the quantitative accuracy.
[0056]
Next, a description will be given of processing to be used before the operation in order to correctly perform the deconvolution operation. Noise 67 and drift 68 shown in FIG. 2 interfere with the deconvolution operation.
If these are deconvoluted as they are, there will be problems such as amplification as pseudo peaks and improper processing of peaks to be deconvoluted.
[0057]
First, as the noise removal processing, a method of reducing the number of data points of a detected chromatogram by adding some data points is generally adopted.
In addition, a smoothing method and a known method for removing high-frequency components can be used.
[0058]
As the drift removal processing, generally, a method of connecting a valley between peaks with a straight line or forcibly drawing a straight line horizontally and subtracting a base line from a detected chromatogram is employed.
Other known methods for removing low-frequency components, methods for removing low-order polynomial components using calculus, methods for searching for a base line from differential feature points, methods using a neural network, etc. There is.
[0059]
Up to this point, the spread function has been obtained from the measured data. However, even when the spread function is unknown, the Gaussian can be deconvoluted.
As described above, the peak waveform of the detected chromatogram can be represented by the convolution of the spread function h (t) and the rectangular waveform r (t), as shown in equation (2).
[0060]
This spreading function h (t) is described in “DYNAMICS OF CHROMATOGRAPHY” J. C. According to Giddings, a Gaussian-like shape is obtained regardless of the extent of physical or chemical interaction within the column, even if the extent outside the column is not negligible.
[0061]
Even if h (t) is slightly distorted from Gaussian in reading or tailing, the convolution of Gaussian G (t) and correction function m (t) as shown in the following equation (4). Can be described as
(Equation 4)

[0062]
In the above equation, G (t) is a Gaussian, m (t) is a correction function, and σ is a standard deviation.
As an example, if h (t) is a tailing peak that can be represented by EMG (t), a correction function is obtained analytically and mathematically as shown in the above equation (3).
[0063]
The overlapping peak can be expressed as in equation (5).
(Equation 5)

Where σ ₁ , Σ ₂ Is the standard deviation.
[0064]
Since Gaussian is additive as described above, it can be decomposed into two Gaussian convolutions, and therefore has the relationship as shown in the following equation (6).
(Equation 6)

[0065]
That is, the two peak waveforms expressed by the equation (5) can be subjected to deconvolution operation using Gaussian.
At this time, the Gaussian is the standard deviation σ of the two peaks. ₁ And σ ₂ It must be a narrow waveform with a smaller standard deviation σ0.
[0066]
Thus, a peak narrower than the detected overlapping peak can be obtained.
Finally, y (t) is converted to Gaussian G (t, σ ₀ ), The chromatogram is converted to a narrower Gaussian with the original width. Osamu Mano It can be expressed by convolution with a positive function.
[0067]
(Equation 7)

As shown in [] above, σ ₁ , Σ ₂ A narrower peak is obtained. More specifically, a rectangular waveform at the time of injection is subjected to a convolution operation by this correction function.
[0068]
The processing when the spread function h (t) is unknown will be described with reference to FIG.
The process starts at step 71.
Step 72 removes noise drift from the detected chromatogram and prepares for deconvolution processing.
In step 73, it is input whether the deconvolution process is performed manually or automatically.
[0069]
In the automatic case, in step 81, a Gaussian having a standard deviation slightly smaller than the minimum width is set as a spread function by looking at the peak width of the detected chromatogram as a whole. This is standardized to area 1.
Here, when the peak width differs between the peak having a short retention time and the peak having a long retention time more than twice, the deconvolution operation can be performed by dividing the time.
[0070]
Here, Fourier transform can be used to automatically find the peak width. In step 76, a deconvolution process is performed. In step 77, it is determined whether the processed chromatogram has diverged or has not fallen below a negative lower limit. Further, it is determined whether or not the non-separable peak forms a valley between the peaks.
[0071]
If the determination is successful, the process proceeds to step 79. If the determination is unsuccessful, the process returns to step 81 again, and the deconvolution processing is performed by referring to the determination and adjusting the standard deviation.
If the judgment is successful, display of the processing chromatogram, display of the quantitative result, and the like are performed in step 79. The process ends at step 80.
[0072]
In the case of manual operation, an input operation is performed in step 74. Generally, the standard deviation of the Gaussian to be deconvolved is input. In addition, the start and end points of the processing section can be input. It is also possible to search for an optimal Gaussian standard deviation in order to improve the separation of specific overlapping peaks.
[0073]
In this case, the initial value of the standard deviation, the number, the size, the width, and the holding time of the estimated peak can be input as reference values. As the limit value, the number of peaks can be input up to three, or the width of the overlapping peaks can be set to 0.2 to 0.4 minutes.
[0074]
In a step 75, the information inputted in the step 74 is calculated, a set value of the deconvolution process is obtained, and a reference value for determination is calculated. In step 76, a deconvolution process is performed. A determination is made in step 77.
If not, the reason is displayed at step 78 and a re-entry is prompted at step 74. If it passes, the process chromatogram and quantitative results are displayed in step 79.
The process ends at step 80.
[0075]
The raw 10-minute chromatogram obtained by the catecholamine analyzer is shown in the upper part (a) of FIG.
This is an example of analyzing a biological sample such as plasma or urine. Many peaks often appear when analyzing biological samples. When this is deconvoluted with an appropriate Gaussian, a chromatogram as shown in the middle part (b) is obtained.
Each peak is completely separated up to the base line.
[0076]
However, even if this is displayed as it is, it is difficult to understand the meaning of the processed chromatogram.
Rather, this may be changed to a known display method shown in (c) below.
It is also preferable that only the quantitative value is output and the operator is not conscious of the deconvolution processing.
[0077]
As mentioned above, although one Example of this invention was described, this invention is not limited to the said Example, Various modifications are possible.
Another method for analyzing a chromatogram according to the present invention will be described.
FIG. 11 is a view for explaining the overdrawing by the deconvolution and the reconvolution in the analysis method according to the present invention.
[0078]
FIG. 11A shows the detected original waveform. The original waveform is subjected to deconvolution processing to separate and independent each waveform peak. This state is shown in FIG.
The separated and independent waveform peaks are reconvolved by the same spread function h (t) as when deconvolved.
The reconvolution-computed waveform is overlaid on the detected original waveform. FIG. 11 (3) shows this state.
[0079]
[Example 2]
Another embodiment of the chromatographic analyzer according to the present invention will be described. FIG. 12 is a block diagram showing a configuration of a chromatographic analyzer according to another embodiment of the present invention.
The overall configuration of the chromatographic analyzer according to the present invention has been described in [Example 1] with respect to the micro-bore HPLC system of FIG. explain.
[0080]
The data analysis unit 45 includes a noise drift elimination mechanism, a spread function analysis mechanism, a deconvolution operation mechanism, a quantitative calculation mechanism, and a determination mechanism. Although not shown in the figure, it is needless to say that an A / D converter, a storage device, an interface unit and the like are also provided.
The noise drift elimination mechanism, the quantitative calculation mechanism, and the determination mechanism among the above-described configurations are well-known technologies, but will be described together in order. In addition, each said mechanism part is connected by the bus.
[0081]
The noise drift elimination mechanism includes a smoothing mechanism, a baseline calculation mechanism, a fast Fourier transform mechanism, a neural network mechanism, a differentiation / integration mechanism, and a fuzzy inference mechanism. Fits the measured data group of the chromatogram from the detector 43 into a smooth curve by any one of the moving average method, the Savitzky-Golay method, the Kawata-Minami method, and the frequency domain method. The baseline is determined by a simple algorithm.
[0082]
The fast Fourier transform mechanism removes noise that is a high-frequency component and baseline fluctuation that is a low-frequency component, respectively, based on the frequency distribution.
The neural network mechanism learns the drift shape in advance and removes the drift from the measured data according to a predetermined coupling constant.
[0083]
The differentiation / integration mechanism removes a baseline component represented by a low-order polynomial by differentiation / integration. The arranged measurement data points are smoothed by the fuzzy inference mechanism according to a predetermined coupling constant. Thus, the noise and drift of the chromatogram from the detector 43 are removed.
[0084]
The spread function analysis mechanism includes a Gaussian generation mechanism, an area 1 normalization mechanism, and a detected peak waveform setting mechanism. In the case of an analysis mathematical spread function, Gaussian is generated by the Gaussian generation mechanism. In the case of the measured spread function, the peak waveform actually measured by the area 1 normalization mechanism is normalized, and the measured waveform normalized by the detected peak waveform setting mechanism is used as the spread function. In this way, an actual or analytical mathematical spreading function is generated.
[0085]
The deconvolution operation mechanism includes a time section setting mechanism, a spreading function setting mechanism, and a deconvolution processing mechanism.The time section setting mechanism sets a section for deconvolution processing, and the spreading function setting mechanism includes: A spread function is set from the generated actual or analytical mathematical spread function.
The deconvolution processing mechanism performs deconvolution processing on the set overlapping waveform in the section to be subjected to the deconvolution processing using the spreading function, and separates and separates the overlapping waveform.
[0086]
The quantitative calculation mechanism includes an integration mechanism, and the integration mechanism calculates an area of each of the separated and independent peak waveforms. In this way, the quantification is calculated based on the peak area or height. The quantitative result is output to the CRT 46 and the printer unit 47.
[0087]
The determination mechanism that performs various determinations is a discriminant calculation mechanism that calculates the value of a discriminant used to determine the success of deconvolution processing, and is limited when searching for a Gaussin used for deconvolution processing. A limit value calculation mechanism for calculating a value, and a spread function search mechanism for searching for an optimum Gaussin according to a predetermined coupling coefficient, and in each of the processing steps, support for enabling each mechanism to perform a predetermined function. I have.
[0088]
The functions of the control unit 44, the data analysis unit 45, the CRT unit 46, the printer unit 47, and the like in FIG. 12 are integrated, and the A / D conversion unit, the memory unit, the control unit, the interface unit, and the output device are integrated. A function h (t) representing the spread is calculated and output from the detected peak waveform of the chromatogram, and the detected peak waveform is deconvoluted using the detected peak waveform and the function h (t). Alternatively, a data processing device used for a chromatogram analyzing method and a chromatographic device, which output a chromatogram after deconvolution processing, may be used.
[0089]
【The invention's effect】
As described above in detail, according to the configuration of the present invention, the distortion of the chromatogram waveform is corrected, the overlapping peak waveform is sharpened, and the separation is improved, and the peak area is reduced. It is possible to provide a chromatogram analysis method that is more accurately obtained and has improved quantitative analysis accuracy.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a microbore HPLC system used for an analysis method according to one embodiment of the present invention.
FIG. 2 is a diagram showing a chromatogram in the analysis method of FIG. 1;
FIG. 3 is a schematic explanatory view of a spreading mechanism of a chromatogram in the analysis method of FIG. 1;
FIG. 4 is an explanatory diagram of a deconvolution operation for an overlapping peak of a chromatogram in the analysis method of FIG. 1;
FIG. 5 is a flowchart of a processing step according to the analysis method of FIG. 1;
FIG. 6 is an overlaid plot of a chromatogram in the analysis method of FIG. 1;
FIG. 7 is a diagram showing a chromatogram of amino acid analysis by the analysis method of FIG. 1;
8 is a diagram showing a chromatogram of glycohemoglobin analysis by the analysis method of FIG.
FIG. 9 is a processing flowchart in the case where a spread function is unknown in the analysis method of FIG. 1;
10 is a diagram showing a chromatogram of catecholamine analysis in the analysis method of FIG.
FIG. 11 is an explanatory diagram of superposition by a reconvolution operation in the analysis method of FIG. 1;
FIG. 12 is a block diagram showing a configuration of a chromatographic analyzer according to another embodiment of the present invention.
[Explanation of symbols]
40 ... pump
41 ... column
42 ... Detector
43 ... Sampler
44 ... Control unit
45 Data analysis unit
46 CRT display
47… Printer
49 ... Eluent
51 ... unknown sample
52: Standard sample
54 ... input section

Claims (6)

A chromatogram analysis method in which a sample to be analyzed that has passed through a separation column is detected by a detector, and a peak in a chromatogram obtained by the detection is represented and analyzed using a Gaussian having a standard deviation σ as a coefficient as a spreading function. ,
Setting an arbitrary value to the standard deviation σ of the Gaussian;
Performing a deconvolution process on the chromatogram;
For the chromatogram after the deconvolution processing,
Is the chromatogram diverging?
Is not below the negative lower bound,
Whether the unseparated peaks form valleys between the peaks,
Determining whether any of the criteria is satisfied; and
And a step of adjusting the standard deviation value of the Gaussian and performing the deconvolution process again if the judgment is unsuccessful. In the chromatogram analysis method according to claim 1,
The peak separated by the deconvolution process is subjected to a reconvolution process using the spread function used for the deconvolution process, and a waveform after the reconvolution process is displayed by being superimposed on a peak waveform before the deconvolution process. A chromatogram analysis method. In the chromatogram analysis method according to claim 1,
A step of removing noise and drift from the chromatogram before setting the Gaussian standard deviation σ. In the chromatogram analysis method according to claim 1,
The spreading function h (t) is
h (t) = G (t) * m (t)
(Where G (t) is the Gaussian and m (t) is a correction function)
The said Gaussian is represented by Formula (1), The chromatogram analysis method characterized by the above-mentioned.

In the chromatogram analysis method according to claim 1,
A method for analyzing a chromatogram, wherein the standard deviation σ is set to a standard deviation smaller than a peak width of a detected chromatogram. In the chromatogram analysis method according to claim 1,
A chromatogram analysis method, wherein the chromatogram has a spread function whose function is unknown.

Images