JP3661797B2 - Method for determining corresponding points for alignment of radiographic images - Google Patents

Description

【００４４】
となるが、クロス型フィルタとは、要素ａ＝Ａi,j(i=j or i=n-j-1),ｂ＝Ａi,j(i ≠j and i ≠n-j-1)としたときにａ≠ｂ（ａ＞ｂ）であるようなマトリクスをいうものであり、本実施例においては、例えば
【００４５】
【数２】

【００４６】
というマトリクスを用いる。
【００４７】
このようなクロス型のフィルタを用いることによって、図４に示すように、ろっ骨とろっ骨とが交わるクロスエッジのような構造の複雑な部分、すなわち、その周囲の領域からみて極端に濃度が変化する点を検出することができ、本実施例においては４つの点40Ａ，40Ｂ，40Ｃ，40Ｄを検出することができる。なお、このようなクロスエッジは４つの領域14Ａ〜14Ｄ中に複数存在するが、本実施例においては各領域においてフィルタの出力が最大値となった点40Ａ，40Ｂ，40Ｃ，40Ｄを検出するものとする。
【００４８】
このようにして４つの点40Ａ〜40Ｄが検出されると、この４つの点を基準対応点とし、各基準対応点を中心としたテンプレート領域を放射線画像4a上に設定する。すなわち、図５に示すように４等分した各領域14Ａ〜14Ｄに基準対応点40Ａ〜40Ｄを中心としたテンプレート領域41Ａ〜41Ｄを設定する。
【００４９】
次いで各テンプレート領域41Ａ〜41Ｄを放射線画像4b上において所定範囲内で移動させてテンプレートマッチングを行う。ここで、テンプレートマッチングは上述した相関法またはＳＳＤＡを用いて行う。相関法においては前述したように標準化値が最大となる点が以下に記載する対応点の座標を与える。また、ＳＳＤＡにおいても上述したように残差の和が最小となる点が対応点の座標を与える。
【００５０】
このようにしてテンプレートマッチングを行うことにより、図６に示すように放射線画像4bに複数の対応点が求められる。
【００５１】
次いで、このようにして求められた複数の対応点について、以下の条件を満たすか否かの評価を行う。すなわち、
（１）テンプレートマッチングにより得られた、テンプレート領域41Ａ〜41Ｄと放射線画像4bの一部との相関値が所定値以上であること、
（２）対応点近傍の相関値の分布が尾根状とならないこと、
（３）任意の２基準対応点間の距離と、２基準対応点に対応する放射線画像4b上の２対応点の距離との比が一定値の範囲内にあること、
の評価を行う。以下この評価について詳細に説明する。
【００５２】
まず、（１）の評価について説明する。
【００５３】
図６に示す放射線画像4b上において求められた複数の対応点について、各対応点におけるテンプレートマッチングを行った際の相関値を求め、各対応点における相関値の値が0.84未満となる対応点を排除する。すなわち、各対応点における相関値をＣＯＲＲ（ｉ，ｊ）（ｉ，ｊは対応点の座標）としたときに、
ＣＯＲＲ（ｉ，ｊ）＜0.84 …(3)
となった対応点を図６に示す複数の対応点から排除する。これにより放射線画像4b中に求められた複数の対応点のうち、ノイズのように位置合せに不適な対応点が排除される。
【００５４】
次いで、前述した評価（２）について説明する。
【００５５】
まず、図６に示す放射線画像4b上において求められた複数の対応点について、各対応点を中心として上下左右および45度斜め方向の８点における相関値の傾きを求める。例えば、図７に示す対応点である画素50を中心として８つの画素51Ａ〜51Ｈにおけるテンプレートマッチングの際の相関値の傾きを求める。
【００５６】
すなわち、
Ａｋ＝（ＣＯＲＲ（ｉ，ｊ）−ＣＯＲＲｋ（ｉ，ｊ））／10 …(4)
但し、ＣＯＲＲ（ｉ，ｊ）：対応点における相関値
ＣＯＲＲｋ（ｉ，ｊ）：ｋ方向における10画素先の相関値
Ａｋ：ｋ方向における傾き
により傾きを求め、角傾きＡｋについて
Ａｋ＜Ｔｈ（Ｔｈ＝0.00075 ） …(5)
ならば、その画素を図６に示す複数の対応点から排除する。
【００５７】
このように、ある方向について相関値の傾きが小さい場合、その画素は図８に示すような相関値の分布が尾根状となっている部分に存在するものと判断される。このように、相関値の分布が尾根状となっている部分に対応点が存在すると、位置合せを行ったときこの尾根に沿って画像がずれてしまうこととなる。したがって、図８に示すようにある画素52の近傍の相関値の分布が尾根状となり、対応点の位置が矢印Ａ方向にずれ易いような場合は、（２）の条件を満たすか否かの評価によりその対応点は位置合せのためには不適なものとして、複数の対応点から排除される。
【００５８】
次いで、前述した評価（３）について説明する。
【００５９】
まず、放射線画像4a上の４つの基準対応点40Ａ〜40Ｄのうちの任意の２点間の距離と、この２点に対応する放射線画像4b上の２対応点間の距離とを求める。例えば、図９に示すように、放射線画像4a上における基準対応点40Ａ，40Ｂ間の距離ＡＢと、放射線画像4b上における対応点40Ａ′，40Ｂ′間の距離Ａ′Ｂ′を求める。すなわち、基準対応点40Ａ，40Ｂの座標を（Ｕｉ，Ｖｉ），（Ｕｊ，Ｖｊ）とし、対応点40Ａ′，40Ｂ′の座標を（Ｘｉ，Ｙｉ），（Ｘｊ，Ｙｊ）としたときに
【００６０】
【数３】

【００６１】
により距離ＡＢ，Ａ′Ｂ′を求める。
【００６２】
次いで距離ＡＢ，Ａ′Ｂ′の比を求め、この比が所定のしきい値の範囲内にあるか否かをみる。すなわち
Ｔｈ１＜Ａ′Ｂ′／ＡＢ＜Ｔｈ２ …(7)
なる評価を行い、式(7) の条件を満たす対応点のみが位置合せに適した対応点として用いられる。
【００６３】
ここで、式(7) における比Ａ′Ｂ′／ＡＢは各点の位置が真に対応する場合、本来ならば１となるはずであるが、図１に示す放射線画像の撮影の際には、蓄積性蛍光体シート７に蓄積記録される放射線画像は、蓄積性蛍光体シート５に蓄積記録される放射線画像と比較して若干大きい画像となる。これは、蓄積性蛍光体シート７は放射線源２からみてシート５よりも離れた位置にあるため、放射線が広がって照射されることにより生ずる現象である。したがって、式(7) におけるＡ′Ｂ′／ＡＢの条件を１と定めずに、１より若干大きいしきい値Ｔｈ１からＴｈ２の範囲にあるようにして、条件（３）の評価を行うことが望ましい。
【００６４】
このように、図６に示す複数の対応点の評価を行い、最終的に位置合せに適した対応点として図10に示すように最大４つの対応点44Ａ〜44Ｄが４つの領域43Ａ〜43Ｄ内において選択される。
【００６５】
次いで、放射線画像4a上における各基準対応点40Ａ〜40Ｄの座標を（ｕi ，ｖi ）(i=1〜4)として、アフィン変換
【００６６】
【数４】

【００６７】
但し、ｍは拡大，縮小率、θは回転量、Ａ，Ｂは回転移動補正および拡大または縮小率補正を示す係数、Ｃ，Ｄは平行移動補正を示す係数
にしたがって、各対応点（ｘi ，ｙi ）の座標を変換することにより第１の放射線画像4aと第２の放射線画像4bとを重ね合せる。ここで式(3) に基づく座標変換では、第２の放射線画像全体をＸ方向とＹ方向とで互いに独立に拡大もしくは縮小すること、第２の放射線画像全体を回転移動すること、および第２の放射線画像をＸ方向，Ｙ方向に平行移動すること、のすべてが同時に行われる。
【００６８】
ここで、式(8) に含まれる係数Ａ，Ｂ，Ｃ，Ｄの求め方について説明する。
【００６９】
本発明においては、最小二乗法により係数Ａ，Ｂ，Ｃ，Ｄを決定するものであり、まず、基準対応点（ｕi ，ｖi ）と対応点（ｘi ，ｙi ）との位置的な誤差Ｅを、

とし、誤差Ｅの二乗誤差を最小にするために式(9) を解くと、
【００７０】
【数５】

【００７１】
となる。よって、式(10)から１次方程式を解くと、係数Ａ，Ｂ，Ｃ，Ｄは以下のように求められる。
【００７２】
Ａ＝［ｄ・（ｅ＋ｆ）−ｂ・i −ｃ・j ］／Δ
Ｂ＝［ｄ・（ｇ＋ｈ）＋ｃ・ｉ−ｂ・ｊ］／Δ
Ｃ＝［−ｂ・（ｅ＋ｆ）＋ｃ・（−ｇ＋ｈ）＋ａ・ｉ］／Δ
Ｄ＝［−ｃ・（ｅ＋ｆ）−ｂ・（−ｇ＋ｈ）＋ａ・ｊ］／Δ
但し、Δ＝ａ・ｄ−ｂ² −ｃ²
ａ＝Σ（ｘi ² ＋ｙi ² ），ｂ＝Σｘi ，ｃ＝Σｙi ，ｄ＝Σ
ｅ＝Σｘi ・ｕi ，ｆ＝Σｙi ・ｖi ，
ｇ＝Σｙi ・ｕi ，ｈ＝Σｘi ・ｖi ，ｉ＝Σｕi ，ｊ＝Σｖi …(11)
このようにして求められた係数、Ａ，Ｂ，Ｃ，Ｄを用いて式(8) により座標変換を行うことにより、各対応点44Ａ〜44Ｄと基準対応点40Ａ〜40Ｄとの誤差が最小となるように第２の放射線画像4bを第１の放射線画像4aに略等しく合せることができる。すなわち、図11に示すように、本発明の位置合せ方法により55組の放射線画像の位置合せを行ったデータでみてみると、基準レベルを｜Ｒ｜＝4.5 、 max｜Ｒi ｜＝1.00とした場合、すべてのデータが基準レベル以内、しかも｜Ｒ｜＝3.15、 max｜Ｒi ｜＝0.7 内に入っており、図12に示す従来の方法による位置合せ結果と比較して、非常に高精度に位置合せが行われていることがわかる。
【００７３】
このようにしてアフィン変換が行われた後、サブトラクション処理、すなわち画像信号ＳＯ₂ の位置合せ後の画像信号をＳＯ₂ ′としたとき、
Ｓ１＝Ｗａ・ＳＯ₁ −Ｗｂ・ＳＯ₂ ′＋Ｃ …(12)
但し、Ｗａ，Ｗｂは重み付け係数、
Ｃはバイアス分を表す。
【００７４】
により重み付け引き算が行われ、これにより２つの放射線画像の差の画像に対応する画像信号Ｓ１が生成される。この画像信号Ｓ１は画像処理表示装置30のＣＲＴディスプレイ32に送られ、この画像信号Ｓ１に基づく可視画像（エネルギーサブトラクション画像）がＣＲＴディスプレイ32に再生表示される。なお、本体部34で実行される上記サブトラクション処理を行う機能（ハードウェアとソフトウェアとの組合せ）が本発明の演算ユニットの一例と観念される。
【００７５】
なお、上記実施例においては、図１における放射線源２からみて遠い位置にある蓄積性蛍光体シート７より得られる放射線画像からテンプレートマッチングにより検出された対応点をアフィン変換するようにしているが、これは以下の理由によるものである。すなわち、図１に示すようないわゆるワンショットで２枚の蓄積性蛍光体シートに放射線画像を蓄積記録するような場合は、２枚のシート５，７はフィルタ６を挟んで空間的に互いに異なる位置に配置されることとなり、これにより２枚のシートの放射線源２および被写体４からの距離が異なり、このため各シート５，７に記録される放射線画像の寸法が異なり放射線源２からみて遠い位置にあるシート７に蓄積記録される放射線画像4bはシート５に蓄積記録される放射線画像4aと比較してボケたり、散乱線が多い画像となる。また、上述したアフィン変換を行った場合、変換された画像については、各画素間は何らかの補間をする必要があるため多少なりとも画質が劣化するものである。したがって、画質のよい放射線画像4aをアフィン変換し画質を劣化させるよりも、放射線画像4aと比較して画質が劣る放射線画像4bを変換した方が、結果として得られるサブトラクション画像の画質が担保されることとなる。したがって、放射線源２からみて遠い位置にある蓄積性蛍光体シート７から得られた放射線画像4bをアフィン変換することとしたものである。
【００７６】
また、上述した実施例においては、前述した３つの条件（１），（２）および（３）のすべての条件を満たした対応点の位置合せのために用いるようにしているが、３つの条件のうち少なくとも１つの条件を満たす対応点を位置合せのために用いるようにしてもよい。但し、条件が多いほどより位置合せに適した対応点が選択されるため、条件の数は求められる位置合せの精度により決定すればよい。
【００７７】
さらに、上述した実施例においては、条件（１）における相関値のしきい値として0.84を用いているが、このしきい値はいかなる値でもよく、さらに、図13に示すように放射線画像の線量が大きい部分ほどしきい値を大きくするようにしてもよい。このように線量に応じてしきい値を変化させるのは、放射線量が低いとノイズが多くなりフィルタの出力値が小さくなって特徴点を求め難くなるからである。
【００７８】
また、上述した実施例においては、条件（２）についての評価を行う際に、対応点である画素50の上下左右および斜め45度方向の８つの画素51Ａ〜51Ｈにおける相関値の傾きを求めるようにしているが、画素50の上下左右方向の４つの画素51Ａ，51Ｃ，51Ｅ，51Ｇにおける相関値の傾きを求めるようにしてもよいものである。但し、この方向の数は多いほどより位置合せに適した対応点が選択されるため、相関値の傾きを求める画素の数は、求められる位置合せの精度により決定すればよい。
【００７９】
なお、上述した条件（２）に関する評価を行うことにより、例えば照射野絞りを用いて放射線画像の撮影を行った場合のように、照射野エッジ上に対応点が求められても、この対応点は排除されるため、照射野エッジに沿って画像がずれて位置合せがなされることはなくなる。
【００８０】
また、上述した実施例においては、対応点が選択された後に、アフィン変換により位置合せを行うようにしているが、位置合せの方法はアフィン変換に限定されるものではなく、前述した特開昭58-163338 号に開示されたマーカーを用いて位置合せを行う方法において、マーカーの代わりに対応点を用いて位置ズレを補正するようにしてもよい。
【００８１】
さらに、上述した実施例において、評価を行った結果、対応点が２個以下となってしまった場合、すなわち、精度良く位置合せを行うには対応点が少なすぎる場合は、前述した評価における条件をゆるくして再度評価を行って対応点を求めるようにすればよい。
【００８２】
またこのような場合、上記（１）、（２）および（３）の条件のうちいずれの条件をも満たさない対応点から所定距離以上離れた位置、すなわち図14に示すように評価に適合しない対応点を含む領域の、条件を満たさない対応点から半径ｒ以上離れた位置において、対応点が位置合せを正確に行うことができる所定数以上となるまで、上述したテンプレート領域の設定、テンプレートマッチングおよび評価を繰り返し行うようにすればよい。
【００８３】
すなわち、図14に示すように領域14A で求められた基準対応点40A が評価に適合しない対応点であった場合、この対応点40A から半径ｒ離れた斜線の領域において、放射線画像4aに対して上述したフィルタリング処理を行ってテンプレート領域を設定し、この新たに設定されたテンプレート領域を放射線画像4b上において所定範囲内で移動させてテンプレートマッチングを行う。そしてこのテンプレートマッチングにより新たな対応点が求められた後、この対応点について上述した（１）、（２）および（３）の条件のうち少なくとも１つの条件を満たすか否かの評価を行う。そしてこれにより、位置合せを行う対応点の数が所定数以上（例えば３個以上）となった場合は、この対応点を用いて位置合せを行う。一方、位置合せを行う対応点の数が所定数以上とならない場合は、位置合せを行う対応点の数が所定数以上となるまで、上記テンプレート領域の設定、テンプレートマッチングおよび評価を繰り返し行う。このように、対応点の数が所定数以上となるまでテンプレート領域の設定等を繰り返し行うことにより、位置合せを行う対応点を確実に複数個求めることができ、このように複数個の対応点を用いて位置合せを行うことにより、より精度良く位置合せを行うことが可能となる。
【００８４】
ここで、評価に適合しなかった対応点から半径ｒ以上離れた放射線画像上において上記評価等を繰り返し行うのは、例えば評価に適合しなかった対応点において、相関値の分布が尾根状となるものであった場合、この対応点の近傍で上記評価等を再度行うと、求められた新たな対応点の近傍の領域において相関値の分布がまた尾根状となってしまうことがあり、上記（２）の条件に適合しないことがある。このため、評価に適合しなかった対応点から半径ｒ以上離れた放射線画像上において、再度上記評価等を繰り返し行うようにしたものである。なお、この半径ｒはこの尾根状となる部分を避けることができるように設定することが必要である。
【００８５】
なお、このように評価を繰り返し行う場合には、演算時間短縮のため繰り返し回数を制限することが好ましい。本出願人の胸部の放射線画像を対象とした実験によれば、約500 例のデータベースにおいて位置合せを行うための対応点の数を４個とした場合、平均的な対応点の数は繰り返しを行わない場合は２．８個、繰り返し回数を３回に制限した場合は３．８個であった。したがって、繰り返し回数を制限する場合は３回程度に制限することが好ましい。
【００８６】
また、上述した実施例においては、エネルギーサブトラクション処理を行うために２つの放射線画像の位置合せを行うようにしているが、重ね合せ処理を行うために２つの放射線画像の位置合せを行うようにしてもよい。すなわち、図15に示すように図２と同様の放射線画像撮影装置１においてフィルタ６を用いることなく２枚の蓄積性蛍光体シート５′，７′に被写体４の放射線画像を蓄積記録し、蓄積性蛍光体シート５′，７′から図３に示す放射線画像読取装置により２つの放射線画像を表す画像信号ＳＯ₁ ，ＳＯ₂ を得、２枚の蓄積性蛍光体シート５′，７′から得られた放射線画像のうち、シート７′から得られた放射線画像を上述した実施例と同様にアフィン変換を施して位置合せを行った後、重ね合せ処理、すなわち画像信号ＳＯ₂ の位置合せ後の画像信号ＳＯ₂ ′としたとき、
Ｓ２＝Ｗｃ・ＳＯ₁ ＋Ｗｄ・ＳＯ₂ ′ …(13)
但し、Ｗｃ，Ｗｄは重み付け係数
により、重み付け加算が行われ、これにより２つの放射線画像の和の画像に対応する画像信号Ｓ２を生成するものである。
【００８７】
また、上述した実施例においては、クロス型フィルタにより放射線画像をフィルタリング処理して、フィルタ出力の最大値を与える点を検出するようにしているが、この点としては例えば、分割された各領域の中の最大値を表す点、ある所定値以上の点等、いかなる点を用いてもよいものである。
【００８８】
さらに、上述した実施例においては、２つの放射線画像の位置合せについて説明しているが、位置合せを行う放射線画像の数は３枚以上であってもよいものである。この場合、３以上の放射線画像のうち、１つの放射線画像に上述した基準対応点を設定し、他の放射線画像についてテンプレートマッチングにより対応点を設定し、各放射線画像の対応点をアフィン変換により基準対応点に合わせるようにすればよい。
【００８９】
また、上述した実施例においては式(8) に示すアフィン変換により位置合せを行うようにしているが、一般式である
【００９０】
【数６】

【００９１】
（但し、ｕ，ｖは基準対応点の座標、ｘ，ｙは変換される対応点の座標、ａ，ｂ，ｃ，ｄは回転移動補正および拡大または縮小率補正を示す係数、ｅ，ｆは平行移動補正を示す係数）
により表されるアフィン変換を行うものであれば、いかなる係数により位置合せを行うようにしてもよいものである。
【００９２】
【発明の効果】
以上詳細に説明したように、本発明による放射線画像の位置合せ用対応点決定方法は、テンプレートマッチングにより求められた複数の対応点について前述した（１）、（２）および（３）のうち少なくとも１つの評価をするようにしたため、位置合せに適した対応点をより精度良く求めることができ、このように求められた対応点を用いて位置合せを行えば、より精度良く迅速に位置合せを行うことができる。
【００９３】
また、上記評価に適合する対応点が所定数未満であるときに、テンプレート領域の設定、テンプレートマッチングおよび評価を、評価に適合する対応点が所定数となるまで繰り返し行うことにより、位置合せのための対応点をより多く求めることができるため、この対応点を用いることにより位置合せをより正確に行うことができる。
【図面の簡単な説明】
【図１】本発明の実施例によるサブトラクションを行う放射線画像を得る放射線画像記録装置の概略図
【図２】各蓄積性蛍光体シートに蓄積記録された放射線画像を模式的に表す図
【図３】本発明による放射線画像位置合せ方法を実施するための装置を内包した画像読取装置の概略を表す図
【図４】放射線画像を分割して基準対応点を設定した状態を表す図
【図５】放射線画像にテンプレート領域を設定した状態を表す図
【図６】他の放射線画像に対応点を設定した状態を表す図
【図７】評価（２）を説明するための図
【図８】相関値の分布が尾根状となる状態を表す図
【図９】評価（３）を説明するための図
【図１０】選択された対応点を表す図
【図１１】本発明による放射線画像位置合せ方法により位置合せがなされた結果を表す図
【図１２】従来の放射線画像位置合せ方法により位置合せがなされた状態を表す図
【図１３】放射線としきい値との関係を表す図
【図１４】評価に適合しなかった対応点から半径ｒ以上離れた領域を斜線で表した図
【図１５】本発明の実施例による重ね合せを行う放射線画像を得る放射線画像記録装置の概略図
【符号の説明】
１ 放射線撮影装置
２ 放射線源
３ 放射線
４ 被写体
５ 第１の蓄積性蛍光体シート
６ フィルタ
７ 第２の蓄積性蛍光体シート
10 放射線画像読取装置
15 シート搬送手段
16 レーザ光源
17 光ビーム
18 モータ
19 回転多面鏡
20 集束レンズ
21 ミラー
22 輝尽発光光
23 光ガイド
24 フォトマルチプライヤ
25 ログアンプ
26 Ａ／Ｄ変換器
30 画像処理表示装置
31 キーボード
32 ＣＲＴディスプレイ
33 フロッピィディスク駆動装置
34 本体部
40Ａ〜40Ｄ 基準対応点
41Ａ〜41Ｄ テンプレート領域
44Ａ〜44Ｄ 対応点[0001]
[Industrial application fields]
The present invention relates to a method for obtaining corresponding points used when aligning images by correcting misalignment of a plurality of radiographic images, and in particular, the position of radiographic images for radiographic image superimposition processing or subtraction processing. The present invention relates to a method for obtaining corresponding points when performing alignment.
[0002]
[Prior art]
Using a stimulable phosphor, radiation image information of a subject such as a human body is temporarily recorded on a sheet of the stimulable phosphor (hereinafter referred to as a “storable phosphor sheet”), and this is scanned with excitation light. There is known a method of emitting light, photoelectrically reading the stimulated emission light to obtain an image signal, and processing the image signal to obtain a radiation image of a subject having good diagnostic suitability. This final image can be reproduced as a hard copy or reproduced on a CRT.
[0003]
On the other hand, a radiographic image overlay process has been known (for example, see JP-A-56-11399). In general, a radiographic image is used for diagnostic purposes and other purposes. However, the use of the radiographic image requires that a minute difference in radiation absorption of a subject be detected well. The degree of detection in a radiographic image is called contrast detection capability or simply detection capability. It can be said that the higher this detection capability, the higher the diagnostic performance and the higher the practical value. Therefore, in order to enhance the diagnostic performance, it is desired to increase this detection capability, but the biggest obstacle is various noises. The superimposition process is a method that greatly reduces the various noises and makes it possible to clearly observe a slight difference in radiation absorption of the subject in the final image, thereby greatly improving the detection capability. That is, radiographic images are taken (accumulated and recorded) on a plurality of storable phosphor sheets, and a plurality of image signals obtained by reading the plurality of sheets are subjected to addition processing. Is to decrease.
[0004]
Conventionally, in order to actually perform the superimposing process, for example, two sheets of stimulable phosphor sheets are put on a cassette and a subject is photographed, and normal reading is performed on the two sheets of stimulable phosphor sheets. A method is used in which processing is sequentially performed to obtain two sets of image signals, and the two sets of image signals are added.
[0005]
On the other hand, a subtraction process for a radiographic image has been conventionally known. This subtraction of radiographic images is obtained by photoelectrically reading out two radiographic images taken under different conditions to obtain digital image signals, and then subtracting these digital image signals in correspondence with the pixels of both images. This is a method for obtaining a difference signal for extracting a specific structure in a radiographic image. By using the difference signal thus obtained, a radiographic image in which only the specific structure is extracted can be reproduced.
[0006]
There are basically the following two methods for this subtraction process. That is, (1) Extracting a specific structure by subtracting the image signal of the radiographic image not injected with the contrast agent from the image signal of the radiographic image in which the specific structure is emphasized by contrast medium injection So-called temporal subtraction processing,
(2) The same subject is irradiated with radiation having different energy distributions, or the radiation after passing through the subject is irradiated to two radiation detection means while changing the energy distribution state, whereby a specific structure is formed. So-called energy subtraction in which different images exist between two radiographic images, and then an image of a specific structure is extracted by performing an appropriate weighting between the image signals of the two radiographic images and then subtracting them. It is processing.
[0007]
Since this subtraction process is a particularly effective method for medical diagnosis, it has attracted a great deal of attention in recent years, and its research and development have been actively promoted using electronic technology.
[0008]
However, the radiographic image overlay processing method and subtraction processing method using the stimulable phosphor sheet as described above have the following problems.
[0009]
That is, in each of the processing methods using the stimulable phosphor sheet, two (or more than three) stimulable phosphor sheets should be inserted into the imaging table sequentially or simultaneously to overlap or subtract. Radiation is obtained by taking a radiation image, and then inserting stimulable phosphor sheets into the reader individually, and detecting the stimulating light emitted by irradiating the stimulable phosphor sheet with excitation light each time. Although the image is read out, in this process, even if the mechanical accuracy of all the apparatuses involved in photographing and reading is increased, a positional shift and a rotational shift occur between the images to be superimposed or subtracted. As a result, in the superimposition process, various noises are averaged and reduced by this process, but the entire image including the edge of the structure in the image is blurred, and the image to be observed is not suitable for observation. Further, in the subtraction process, the image to be erased is not erased, or conversely, the image to be extracted is erased and a false image is generated, so that an accurate subtraction image cannot be obtained. As described above, it has been found that the above-described positional shift and rotational shift cause serious troubles in diagnosis.
[0010]
When such a deviation occurs between the radiation image information accumulated and recorded on the stimulable phosphor sheet, the radiation image is accumulated and recorded in the stimulable phosphor as a latent image, so that the X-ray image is regarded as a visible image. Unlike an X-ray photographic film that can be used, it is impossible to visually match two X-ray photographs, which makes correction of displacement extremely difficult.
[0011]
Further, even if a positional deviation and a rotational deviation occurring between two radiographic images can be detected by any means, a conventionally known calculation process is performed to correct the read radiographic image data. A great deal of time is consumed, which is a very big problem in practical use.
[0012]
Therefore, the present applicant has proposed a radiation image subtraction processing method using a marker having a shape that provides a reference point or a reference line in Japanese Patent Laid-Open No. 58-163338. In this method, a marker is recorded on two stimulable phosphor sheets at a fixed position with respect to a radiographic image, the marker is detected when the radiographic image is read, a positional shift and a rotational shift are calculated, and subtraction is performed. One of the deviations of the radiographic image is rotated and / or moved on the digital data, and the image data is subtracted between corresponding pixels of the radiographic image. The alignment process in the radiographic subtraction processing method using the marker can also be applied to the above-described registration processing method. In that case, the image data may be added between the corresponding pixels of the radiation image after alignment.
[0013]
However, in this method, each time a radiographic image is taken, the marker as described above must be accumulated and recorded on the stimulable phosphor sheet together with the subject. Then, there is a problem that image information of the subject cannot be obtained from a portion overlapping the marker position of the accumulated and recorded radiographic image.
[0014]
Therefore, a method for performing alignment of radiation images without using a marker or the like for alignment has been proposed by the applicant of the present application (Japanese Patent Application No. 4-318533). In this method, a template area is set in one radiographic image among a plurality of radiographic images to be aligned, and template matching is performed on another radiographic image using the template area, and at least two markers are provided in each radiographic image. In this method, the corresponding points are obtained by substituting the corresponding points so that the corresponding points of the radiographic images coincide with each other, and the rotational movement correction, the enlargement / reduction ratio correction, and the parallel movement correction are performed on each radiographic image. is there.
[0015]
According to this method, since a corresponding point instead of a marker is found and used from an image, quick and highly accurate alignment can be performed without recording a marker or the like together with the subject for alignment. can do.
[0016]
[Problems to be solved by the invention]
However, in the above alignment method, the corresponding point of each radiographic image is obtained by template matching. For example, noise in the radiographic image may be obtained as the corresponding point. The positions of the radiographic images are not common to each other, and the alignment cannot be performed with high accuracy. In addition, when the corresponding point is obtained at a position where the distribution of correlation values such as on the edge of the radiation image has a ridge shape, the image is shifted along the ridge at the time of alignment, and as a result, the alignment is accurately performed. There was a case that could not be done.
[0017]
In view of the above circumstances, an object of the present invention is to provide a method for determining corresponding points for alignment of radiographic images, which can obtain corresponding points suitable for alignment in order to perform alignment with high accuracy. Is.
[0018]
[Means for Solving the Problems]
The corresponding point determination method for alignment of radiographic images according to the present invention is a method for determining corresponding points for performing alignment of a plurality of radiographic images.
Setting a template region on one of the radiographic images;
By performing template matching to obtain a correlation value between the template region and a part of the other radiation image by moving the template region on another radiation image other than the one radiation image, Find multiple corresponding points that correspond to each other,
The corresponding points of one radiographic image in the radiographic images are set as reference corresponding points,
For the plurality of corresponding points,
(1) The correlation value between the template region and a part of the other radiation image obtained by the template matching is not less than a predetermined value.
(2) the distribution of the correlation values in the vicinity of the plurality of corresponding points is not ridged;
(3) The ratio between the distance between any two reference corresponding points and the distance between the two corresponding points on the other radiographic image corresponding to the two reference corresponding points is within a certain range.
Evaluate whether or not at least one of the conditions is satisfied,
Corresponding points that match the evaluation are used as corresponding points for the alignment.
[0019]
If the number of corresponding points that match the evaluation is less than a predetermined number, the setting of the template region, the template matching, and the template matching on the radiation image that is more than a predetermined distance away from the corresponding points that did not match the evaluation It is preferable to repeat the evaluation until the number of corresponding points that match the evaluation reaches the predetermined number.
[0020]
Here, when the distribution of correlation values is ridge-shaped, for example, when a template region including an edge portion of a radiographic image is set, the correlation value obtained by template matching in the vicinity of the edge of another radiographic image is When this correlation value distribution is viewed three-dimensionally, it becomes a ridge shape as shown in FIG. Such a state is referred to as a state where the distribution of correlation values is ridge-like.
[0021]
Here, the template matching is a process of searching for the most matching place by moving the template area on another radiographic image when a template area is set on one radiographic image as described above. The point representing the location gives the coordinates of the corresponding point.
[0022]
In such template matching, examples of the evaluation scale representing the degree of matching include a correlation method and SSDA (Sequential Similarity Detection Algorithm).
[0023]
In this correlation method, a product is calculated for each corresponding pixel, and a value obtained by standardizing the sum of the products (hereinafter referred to as a standardized value) is used as a superimposition scale. This standardization calculates the sum of the products (squares) of the pixels themselves in each region, further calculates the product of each sum, and uses the square root of this product as the denominator of the sum of the products for each corresponding pixel. This is done. When the superposition is complete, the products of the numerators are not all sums of squares due to noise or the like, and therefore the standardized value is considered to be the maximum value closest to 1 even if it is not 1. Therefore, it is considered that the overlay is achieved by moving the template region in various ways on the radiographic image and moving the standardized value as described above. However, the movement with the maximum standardized value cannot be determined unless all movements are completed. Details of this method are described in, for example, Smith et al. “Automated cloud tracking using precisely aligned digital ATS pictures” ibid. July 1972, volume c-21, pages 715-729.
[0024]
In addition, SSDA is a superposition scale that is the sum of absolute values of differences (residuals) for each pixel. When the superposition is complete, it is considered that the residual is minimized even if it does not become zero due to noise. Therefore, it is considered that the overlay is achieved by moving the template region in various ways on the image and by moving the template region to minimize the residual. At this time, if the registration is shifted, the residual increases rapidly when the pixels are sequentially added. Therefore, this SSDA is a method of aborting the addition as soon as the residual exceeds a threshold value during the addition and moving to the next movement. Since the calculation to be used is only addition, and in many cases, the calculation is terminated in the middle, the calculation time is greatly shortened. Details of this method are described in, for example, Barnea et al., “A class of algorithms for fast digital image registration” IEEE. Trans. February 1972, c-21, 179-186.
[0025]
[Action]
According to the method for determining the corresponding point for alignment of the radiographic image according to the present invention, the corresponding point that is not suitable for alignment, such as noise in the radiographic image, is evaluated by evaluating whether or not the condition (1) is satisfied. The corresponding points that are likely to shift along the edges, such as the corresponding points obtained in the edge portion of the radiographic image, are excluded by evaluating whether or not the condition (2) is satisfied. Furthermore, when the radiographic images completely match each other, the distance between the two corresponding points corresponding to each other in each radiographic image is a constant value. Therefore, each radiographic image is evaluated by evaluating whether or not the condition (3) is satisfied. Corresponding points that are not common to
Will be removed. In this way, by evaluating whether or not any one of the three conditions is satisfied, corresponding points unsuitable for alignment are eliminated, and only corresponding points more suitable for alignment are selected. The Rukoto.
[0026]
When imaging is performed by superimposing a plurality of recording media, the distance between any two corresponding points on the radiation image on the side close to the radiation source, and the radiation source corresponding to the two corresponding points are far from each other. The distance between the two corresponding points on the side is larger because the latter is irradiated with radiation. Therefore, in the evaluation of (3), by setting the ratio of the former and the latter within a certain range, a plurality of radiographic images are obtained by performing imaging while superposing a plurality of recording media. However, it is possible to accurately obtain corresponding points suitable for alignment.
[0027]
In addition, since none of the conditions (1), (2), and (3) is satisfied, the number of corresponding points may be too small, and alignment may not be performed accurately. In such a case, that is, when the number of corresponding points that match the evaluation is less than a predetermined number, the setting of the template region, the template matching, and the By repeatedly performing the evaluation until the number of corresponding points that match the evaluation reaches a predetermined number or more, it is possible to obtain more corresponding points for alignment. Therefore, by using these corresponding points, the alignment can be performed more accurately. It can be carried out.
[0028]
Here, the above evaluation or the like is repeatedly performed on a radiographic image that is a predetermined distance or more away from a corresponding point that does not match the evaluation. In the case of the corresponding point, if the above evaluation or the like is performed again in the vicinity of the corresponding point, the distribution of the correlation value becomes a ridge shape at the obtained corresponding point, which may not satisfy the condition (2). . For this reason, the above-described evaluation and the like are repeatedly performed again on a radiographic image that is a predetermined distance or more away from a corresponding point that does not match the evaluation.
[0029]
【Example】
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0030]
FIG. 1 is a schematic view of a radiation imaging apparatus which is an embodiment of an apparatus for recording a radiation image used in the present invention. The radiographic image obtained by this imaging is used for energy subtraction processing.
[0031]
Storable phosphor sheets 5 and 7 are stacked with the filter 6 in between, with the sheet 7 facing down. Above this, a radiation source 2 that emits radiation 3 via a subject 4 is arranged. As described above, the radiation imaging apparatus 1 is configured.
[0032]
The subject 4 is irradiated with radiation 3 emitted from the radiation source 2. The radiation 3a that has passed through the subject 4 is irradiated onto the first stimulable phosphor sheet 5, and a part of the energy of the radiation 3a is recorded on the first stimulable phosphor sheet 5, thereby causing the sheet 5 to have the subject 4 Radiographic images are accumulated and recorded. The radiation 3b that has passed through the sheet 5 further passes through the filter 6, and the radiation 3c that has passed through the filter 6 is applied to the second stimulable phosphor sheet 7. As a result, the radiation image of the subject 4 is also accumulated and recorded on the sheet 7.
[0033]
FIG. 2 is a diagram schematically showing radiographic images accumulated and recorded on the stimulable phosphor sheets 5 and 7. The radiation images 4a and 4b of the subject 4 are accumulated and recorded on substantially the entire surfaces of the stimulable phosphor sheets 5 and 7. That is, the radiation image 4a is a radiation image obtained from the upper storage phosphor sheet 5, and the radiation image 4b is a radiation image obtained from the lower storage phosphor sheet 7.
[0034]
FIG. 3 shows an image processing display which is an embodiment of an arithmetic unit which performs a subtraction process by implementing a radiological image reading apparatus which is an embodiment of a radiographic image reading unit used in the present invention and an alignment method of the present invention. It is a perspective view of an apparatus.
[0035]
After imaging is performed by the radiation imaging apparatus 1 shown in FIG. 1, the first and second stimulable phosphor sheets 5 and 7 are set one by one at a predetermined position of the radiation image reading apparatus 10. Here, the case of reading the first radiation image accumulated and recorded on the first stimulable phosphor sheet 5 will be described.
[0036]
The stimulable phosphor sheet 5 on which the first radiation image is accumulated and recorded at a predetermined position is conveyed in the direction of arrow Y by a sheet conveying means 15 such as an endless belt driven by a driving means (not shown). Scanned). On the other hand, a light beam 17 emitted from a laser light source 16 is reflected and deflected by a rotating polygon mirror 19 driven by a motor 18 and rotated at high speed in the direction of arrow Z, passes through a focusing lens 20 such as an fθ lens, and then is reflected by an optical path by a mirror 21. Is changed to enter the stimulable phosphor sheet 5 and main scanning is performed in an arrow X direction substantially perpendicular to the sub-scanning direction (arrow Y direction). From the location where the light beam 17 of the stimulable phosphor sheet 5 is irradiated, the amount of stimulated emission light 22 corresponding to the stored and recorded radiographic image information is emitted. And is detected photoelectrically by a photomultiplier (photomultiplier tube) 24. The light guide 23 is made by molding a light guide material such as an acrylic plate, and is arranged such that a linear incident end face 23a extends along the main scanning line on the stimulable phosphor sheet 5. The light receiving surface of the photomultiplier 24 is coupled to the emission end surface 23b formed in an annular shape. The stimulated emission light 22 that has entered the light guide 23 from the incident end surface 23a repeatedly undergoes total internal reflection in the light guide 23, proceeds from the emission end surface 23b, is received by the photomultiplier 24, and receives a radiographic image. The stimulated emission light 22 that is represented is converted into an electrical signal by the photomultiplier 24. The analog signal S output from the photomultiplier 24 is logarithmically amplified by the log amplifier 25, then input to the A / D converter 26, and sampled to obtain a digital image signal SO. This image signal SO represents the first radiation image accumulated and recorded on the first stimulable phosphor sheet 5, and the first image signal SO₁Call it. This first image signal SO₁Is once recorded in an internal memory in the image processing display device 30.
[0037]
The image processing display device 30 is loaded with a keyboard 31 for inputting various instructions, a CRT display 32 for displaying a visible image based on auxiliary information and image signals for instructions, and a floppy disk as an auxiliary storage medium. A floppy disk drive 33 and a main body 34 incorporating a CPU and an internal memory are provided.
[0038]
Next, in the same manner as described above, the second image signal SO representing the second radiation image accumulated and recorded on the second stimulable phosphor sheet 7 is used.₂To obtain the second image signal SO.₂Is also temporarily stored in an internal memory in the image processing display device 30.
[0039]
In this way, two image signals SO to be subtracted are calculated.₁, SO₂Are stored in the internal memory, these two image signals SO₁, SO₂Are read out and these two image signals SO are read out.₁, SO₂The images are aligned so that the corresponding subtraction calculation is performed between the pixels of the radiation images carried by the.
[0040]
Here, in this embodiment, the image signal SO₁, SO₂A method for aligning two radiographic images represented by will be described.
[0041]
In the radiographic image alignment method according to the present invention, the radiographic image 4a obtained from the stimulable phosphor sheet 5 located near the radiation source 2 in FIG. 1 is divided into four equal parts as shown in FIG. Regions 14A, 14B, 14C and 14D are defined. For each region, the image is filtered using a cross-type filter to detect a point that gives the maximum output value of the filter.
[0042]
Here, when the filter matrix is Aij (i = 1, 2,... N, j = 1, 2,... N), the filter expression is
[0043]
[Expression 1]

[0044]
However, a cross-type filter means that when an element a = Ai, j (i = j or i = nj-1) and b = Ai, j (i ≠ j and i ≠ nj−1), a ≠ b (a> b), and in this embodiment, for example,
[0045]
[Expression 2]

[0046]
This matrix is used.
[0047]
By using such a cross-type filter, as shown in FIG. 4, the density changes extremely when viewed from a complicated portion of a structure such as a cross edge where the ribs and ribs intersect, that is, from the surrounding area. A point can be detected, and in this embodiment, four points 40A, 40B, 40C, and 40D can be detected. A plurality of such cross edges exist in the four areas 14A to 14D. In this embodiment, the points 40A, 40B, 40C, and 40D at which the output of the filter reaches the maximum value in each area are detected. And
[0048]
When the four points 40A to 40D are detected in this way, these four points are set as reference corresponding points, and a template region centered on each reference corresponding point is set on the radiation image 4a. That is, as shown in FIG. 5, template regions 41A to 41D centering on the reference corresponding points 40A to 40D are set in the respective regions 14A to 14D divided into four equal parts.
[0049]
Next, template matching is performed by moving the template regions 41A to 41D within a predetermined range on the radiation image 4b. Here, the template matching is performed using the above-described correlation method or SSDA. In the correlation method, as described above, the point having the maximum standardized value gives the coordinates of the corresponding point described below. Also in SSDA, as described above, the point with the smallest residual sum gives the coordinates of the corresponding point.
[0050]
By performing template matching in this way, a plurality of corresponding points are obtained in the radiation image 4b as shown in FIG.
[0051]
Next, it is evaluated whether or not the following conditions are satisfied for the plurality of corresponding points obtained in this way. That is,
(1) The correlation value between the template regions 41A to 41D and a part of the radiation image 4b obtained by template matching is not less than a predetermined value;
(2) The distribution of correlation values in the vicinity of the corresponding points should not be ridge-like,
(3) The ratio between the distance between any two reference corresponding points and the distance between the two corresponding points on the radiation image 4b corresponding to the two reference corresponding points is within a certain range.
Perform an evaluation. This evaluation will be described in detail below.
[0052]
First, the evaluation (1) will be described.
[0053]
For a plurality of corresponding points obtained on the radiographic image 4b shown in FIG. 6, a correlation value is obtained when template matching is performed at each corresponding point, and a corresponding point having a correlation value less than 0.84 at each corresponding point is obtained. Exclude. That is, when the correlation value at each corresponding point is CORR (i, j) (i, j are the coordinates of the corresponding point),
CORR (i, j) <0.84 (3)
The corresponding points that have become are excluded from the plurality of corresponding points shown in FIG. As a result, out of a plurality of corresponding points obtained in the radiation image 4b, corresponding points unsuitable for alignment such as noise are excluded.
[0054]
Next, the above-described evaluation (2) will be described.
[0055]
First, with respect to a plurality of corresponding points obtained on the radiographic image 4b shown in FIG. For example, the slope of the correlation value at the time of template matching in the eight pixels 51A to 51H around the pixel 50 that is the corresponding point shown in FIG.
[0056]
That is,
Ak = (CORR (i, j) −CORRk (i, j)) / 10 (4)
Where CORR (i, j): correlation value at corresponding point
CORRk (i, j): Correlation value 10 pixels ahead in the k direction
Ak: inclination in k direction
The angle is obtained by
Ak <Th (Th = 0.00075) (5)
If so, the pixel is excluded from a plurality of corresponding points shown in FIG.
[0057]
Thus, when the slope of the correlation value is small in a certain direction, it is determined that the pixel exists in a portion where the correlation value distribution as shown in FIG. Thus, if a corresponding point exists in a portion where the distribution of correlation values is ridge-like, the image will be shifted along this ridge when alignment is performed. Therefore, as shown in FIG. 8, when the distribution of correlation values in the vicinity of a certain pixel 52 has a ridge shape and the position of the corresponding point is likely to shift in the direction of arrow A, whether or not the condition (2) is satisfied is determined. The corresponding points are excluded from a plurality of corresponding points as being unsuitable for alignment by evaluation.
[0058]
Next, the above-described evaluation (3) will be described.
[0059]
First, the distance between any two of the four reference corresponding points 40A to 40D on the radiation image 4a and the distance between the two corresponding points on the radiation image 4b corresponding to these two points are obtained. For example, as shown in FIG. 9, the distance AB between the reference corresponding points 40A and 40B on the radiation image 4a and the distance A'B 'between the corresponding points 40A' and 40B 'on the radiation image 4b are obtained. That is, when the coordinates of the reference corresponding points 40A, 40B are (Ui, Vi), (Uj, Vj) and the coordinates of the corresponding points 40A ', 40B' are (Xi, Yi), (Xj, Yj).
[0060]
[Equation 3]

[0061]
To obtain the distances AB and A'B '.
[0062]
Next, a ratio between the distances AB and A'B 'is obtained, and it is checked whether or not this ratio is within a predetermined threshold range. Ie
Th1 <A'B '/ AB <Th2 (7)
Only the corresponding points satisfying the condition of Equation (7) are used as corresponding points suitable for alignment.
[0063]
Here, the ratio A′B ′ / AB in the equation (7) should be 1 when the position of each point truly corresponds, but at the time of radiographic image capturing shown in FIG. The radiation image accumulated and recorded on the stimulable phosphor sheet 7 is a slightly larger image than the radiation image accumulated and recorded on the stimulable phosphor sheet 5. This is a phenomenon that occurs when radiation is spread and irradiated because the stimulable phosphor sheet 7 is located farther from the sheet 5 when viewed from the radiation source 2. Therefore, the condition (3) can be evaluated in such a way that the condition A′B ′ / AB in the equation (7) is not set to 1, but is in the range of the threshold values Th1 to Th2 slightly larger than 1. desirable.
[0064]
In this way, the plurality of corresponding points shown in FIG. 6 are evaluated, and finally four corresponding points 44A to 44D are included in the four regions 43A to 43D as corresponding points suitable for alignment as shown in FIG. Is selected.
[0065]
Subsequently, the coordinates of the reference corresponding points 40A to 40D on the radiation image 4a are set as (ui, vi) (i = 1 to 4), and the affine transformation is performed.
[0066]
[Expression 4]

[0067]
Where m is an enlargement / reduction ratio, θ is a rotation amount, A and B are coefficients indicating rotational movement correction and enlargement / reduction ratio correction, and C and D are coefficients indicating parallel movement correction.
Accordingly, the first radiation image 4a and the second radiation image 4b are overlapped by converting the coordinates of the corresponding points (xi, yi). Here, in the coordinate transformation based on Expression (3), the entire second radiation image is enlarged or reduced independently in the X direction and the Y direction, the entire second radiation image is rotated, and the second All of the radiographic images are translated in the X and Y directions at the same time.
[0068]
Here, how to obtain the coefficients A, B, C, and D included in the equation (8) will be described.
[0069]
In the present invention, the coefficients A, B, C, and D are determined by the least square method. First, the positional error E between the reference corresponding point (ui, vi) and the corresponding point (xi, yi) is calculated. ,

And solving equation (9) to minimize the square error of error E,
[0070]
[Equation 5]

[0071]
It becomes. Therefore, when the linear equation is solved from the equation (10), the coefficients A, B, C, and D are obtained as follows.
[0072]
A = [d · (e + f) −b · i −c · j] / Δ
B = [d · (g + h) + c · i−b · j] / Δ
C = [− b · (e + f) + c · (−g + h) + a · i] / Δ
D = [− c · (e + f) −b · (−g + h) + a · j] / Δ
However, Δ = a · d−b²-C²
a = Σ (xi²+ Yi²), B = Σxi, c = Σyi, d = Σ
e = .SIGMA.xi.ui, f = .SIGMA.yi.vi,
g = Σyi · ui, h = Σxi · vi, i = Σui, j = Σvi (11)
By performing the coordinate transformation by the equation (8) using the coefficients A, B, C, and D thus obtained, the error between each corresponding point 44A to 44D and the reference corresponding point 40A to 40D is minimized. Thus, the second radiation image 4b can be matched with the first radiation image 4a substantially equally. That is, as shown in FIG. 11, when the data obtained by aligning 55 sets of radiation images by the alignment method of the present invention is used, the reference level is set to | R | = 4.5 and max | Ri | = 1.00. In this case, all the data are within the reference level, and | R | = 3.15, max | Ri | = 0.7, and compared with the alignment result by the conventional method shown in FIG. It can be seen that the alignment is performed.
[0073]
After the affine transformation is performed in this way, subtraction processing, that is, the image signal SO₂The image signal after the alignment of₂When ′
S1 = Wa / SO₁-Wb ・ SO₂'+ C (12)
Where Wa and Wb are weighting factors,
C represents a bias component.
[0074]
Thus, the weighted subtraction is performed, and thereby the image signal S1 corresponding to the difference image between the two radiation images is generated. The image signal S1 is sent to the CRT display 32 of the image processing display device 30, and a visible image (energy subtraction image) based on the image signal S1 is reproduced and displayed on the CRT display 32. Note that the function (combination of hardware and software) for performing the above-described subtraction processing executed in the main body 34 is considered as an example of the arithmetic unit of the present invention.
[0075]
In the above embodiment, the corresponding points detected by template matching from the radiation image obtained from the stimulable phosphor sheet 7 located far from the radiation source 2 in FIG. 1 are affine transformed. This is due to the following reason. That is, when radiographic images are accumulated and recorded on two stimulable phosphor sheets by so-called one-shot as shown in FIG. 1, the two sheets 5 and 7 are spatially different from each other with the filter 6 interposed therebetween. Accordingly, the distance between the two sheets from the radiation source 2 and the subject 4 is different, and therefore the size of the radiation image recorded on each of the sheets 5 and 7 is different and far from the radiation source 2. The radiation image 4b accumulated and recorded on the sheet 7 at the position is an image that is blurred or has more scattered rays than the radiation image 4a accumulated and recorded on the sheet 5. Further, when the above-described affine transformation is performed, the image quality of the transformed image is somewhat deteriorated because it is necessary to perform some interpolation between the pixels. Therefore, the image quality of the resulting subtraction image is guaranteed when the radiation image 4b, which is inferior in image quality compared to the radiation image 4a, is converted rather than affine transformation of the radiation image 4a with good image quality. It will be. Therefore, the radiation image 4b obtained from the stimulable phosphor sheet 7 located far from the radiation source 2 is subjected to affine transformation.
[0076]
Further, in the above-described embodiment, it is used for alignment of corresponding points satisfying all of the above three conditions (1), (2) and (3). Corresponding points satisfying at least one condition may be used for alignment. However, as the number of conditions increases, corresponding points that are more suitable for alignment are selected, so the number of conditions may be determined based on the required alignment accuracy.
[0077]
Further, in the above-described embodiment, 0.84 is used as the threshold value of the correlation value in the condition (1). However, this threshold value may be any value, and further, as shown in FIG. You may make it enlarge a threshold value, so that a part with large. The reason why the threshold value is changed in accordance with the dose in this way is that when the radiation dose is low, the noise increases and the output value of the filter decreases, making it difficult to obtain the feature point.
[0078]
In the above-described embodiment, when the condition (2) is evaluated, the slopes of the correlation values in the eight pixels 51A to 51H in the vertical and horizontal directions and the 45 degrees oblique direction of the corresponding pixel 50 are obtained. However, the inclination of the correlation value in the four pixels 51A, 51C, 51E, 51G in the vertical and horizontal directions of the pixel 50 may be obtained. However, since the corresponding points more suitable for alignment are selected as the number of directions increases, the number of pixels for which the slope of the correlation value is obtained may be determined by the accuracy of the obtained alignment.
[0079]
Even if a corresponding point is obtained on the irradiation field edge, for example, when a radiographic image is captured using the irradiation field stop by performing the evaluation on the condition (2), the corresponding point is obtained. Therefore, the image is not shifted and aligned along the irradiation field edge.
[0080]
In the above-described embodiment, the alignment is performed by the affine transformation after the corresponding point is selected. However, the alignment method is not limited to the affine transformation. In the method of performing alignment using the markers disclosed in 58-163338, the positional deviation may be corrected using corresponding points instead of the markers.
[0081]
Furthermore, in the above-described embodiment, when the number of corresponding points is less than two as a result of the evaluation, that is, when the number of corresponding points is too small for accurate alignment, the conditions in the evaluation described above It is sufficient to loosen and evaluate again to find the corresponding points.
[0082]
Further, in such a case, a position that is a predetermined distance or more away from a corresponding point that does not satisfy any of the above conditions (1), (2), and (3), that is, does not conform to the evaluation as shown in FIG. In the region including the corresponding points, at the position away from the corresponding points that do not satisfy the condition by a radius r or more, the template region setting and template matching described above are performed until the number of corresponding points is equal to or greater than a predetermined number that enables accurate alignment. The evaluation may be repeated.
[0083]
That is, as shown in FIG. 14, when the reference corresponding point 40A obtained in the region 14A is a corresponding point that does not match the evaluation, the region corresponding to the radiation image 4a in the hatched region away from the corresponding point 40A by the radius r. A template region is set by performing the filtering process described above, and the template setting is performed by moving the newly set template region within a predetermined range on the radiation image 4b. Then, after a new corresponding point is obtained by this template matching, it is evaluated whether or not at least one of the above conditions (1), (2) and (3) is satisfied for the corresponding point. As a result, when the number of corresponding points to be aligned becomes a predetermined number or more (for example, 3 or more), alignment is performed using the corresponding points. On the other hand, if the number of corresponding points to be aligned does not exceed a predetermined number, the template region setting, template matching, and evaluation are repeated until the number of corresponding points to be aligned exceeds a predetermined number. In this way, by repeatedly setting the template area and the like until the number of corresponding points reaches a predetermined number or more, a plurality of corresponding points to be aligned can be reliably obtained. It is possible to perform alignment with higher accuracy by performing alignment using.
[0084]
Here, the above evaluation or the like is repeatedly performed on the radiographic image separated from the corresponding point that does not conform to the evaluation by a radius r or more, for example, the correlation value distribution becomes a ridge shape at the corresponding point that does not conform to the evaluation. In the case of the corresponding point, if the above evaluation or the like is performed again in the vicinity of the corresponding point, the distribution of the correlation value may be ridged again in the region in the vicinity of the new corresponding point. The condition of 2) may not be met. For this reason, the above-described evaluation and the like are repeated again on a radiographic image that is separated from the corresponding point that does not match the evaluation by a radius r or more. The radius r needs to be set so as to avoid the ridge-like portion.
[0085]
When the evaluation is repeated as described above, it is preferable to limit the number of repetitions in order to shorten the calculation time. According to the experiment on the radiographic image of the applicant's chest, when the number of corresponding points for alignment in the database of about 500 cases is four, the average number of corresponding points is repeated. The number was 2.8 when not performed and 3.8 when the number of repetitions was limited to 3. Therefore, when limiting the number of repetitions, it is preferable to limit to about 3 times.
[0086]
In the above-described embodiment, the two radiographic images are aligned in order to perform the energy subtraction process, but the two radiographic images are aligned in order to perform the superimposing process. Also good. That is, as shown in FIG. 15, the radiographic image of the subject 4 is accumulated and recorded on the two stimulable phosphor sheets 5 ′ and 7 ′ without using the filter 6 in the radiographic imaging apparatus 1 similar to FIG. Signal SO representing two radiographic images from the fluorescent phosphor sheets 5 'and 7' by the radiographic image reader shown in FIG.₁, SO₂Among the radiographic images obtained from the two stimulable phosphor sheets 5 'and 7', the radiographic image obtained from the sheet 7 'is subjected to affine transformation in the same manner as in the above-described embodiment, and is aligned. After performing the overlay processing, that is, the image signal SO₂Image signal SO after alignment of₂When ′
S2 = Wc · SO₁+ Wd · SO₂' …(13)
Where Wc and Wd are weighting factors
Thus, weighted addition is performed, thereby generating an image signal S2 corresponding to the sum of the two radiation images.
[0087]
In the above-described embodiment, the radiation image is filtered by the cross-type filter to detect the point that gives the maximum value of the filter output. Any point may be used, such as a point representing the maximum value among them, or a point exceeding a certain predetermined value.
[0088]
Furthermore, in the above-described embodiment, the alignment of two radiographic images has been described. However, the number of radiographic images to be aligned may be three or more. In this case, among the three or more radiographic images, the reference corresponding point described above is set for one radiographic image, the corresponding points are set by template matching for the other radiographic images, and the corresponding points of each radiographic image are determined by affine transformation. It is only necessary to match the corresponding points.
[0089]
In the above-described embodiment, the alignment is performed by the affine transformation shown in the equation (8).
[0090]
[Formula 6]

[0091]
(Where u and v are coordinates of reference corresponding points, x and y are coordinates of corresponding points to be converted, a, b, c and d are coefficients indicating rotational movement correction and enlargement or reduction ratio correction, and e and f are Coefficient indicating translation correction)
As long as the affine transformation represented by the above is performed, the alignment may be performed by any coefficient.
[0092]
【The invention's effect】
As described in detail above, the radiographic image alignment corresponding point determination method according to the present invention is the method of determining at least one of the above-described (1), (2), and (3) for a plurality of corresponding points obtained by template matching. Since one evaluation is performed, it is possible to obtain corresponding points suitable for alignment with higher accuracy. If alignment is performed using the corresponding points thus obtained, alignment can be performed more accurately and quickly. It can be carried out.
[0093]
In addition, when the number of corresponding points that match the above evaluation is less than a predetermined number, template region setting, template matching, and evaluation are repeatedly performed until the predetermined number of corresponding points that match the evaluation is reached. Since more corresponding points can be obtained, alignment can be performed more accurately by using the corresponding points.
[Brief description of the drawings]
FIG. 1 is a schematic diagram of a radiographic image recording apparatus for obtaining a radiographic image for performing subtraction according to an embodiment of the present invention.
FIG. 2 is a diagram schematically showing radiation images accumulated and recorded on each stimulable phosphor sheet.
FIG. 3 is a diagram showing an outline of an image reading apparatus including an apparatus for carrying out a radiographic image alignment method according to the present invention.
FIG. 4 is a diagram illustrating a state in which a reference image is set by dividing a radiation image
FIG. 5 is a diagram showing a state where a template region is set in a radiation image
FIG. 6 is a diagram illustrating a state in which corresponding points are set in another radiographic image.
FIG. 7 is a diagram for explaining evaluation (2);
FIG. 8 is a diagram showing a state in which the distribution of correlation values is ridged
FIG. 9 is a diagram for explaining evaluation (3);
FIG. 10 is a diagram showing selected corresponding points
FIG. 11 is a diagram showing a result of alignment performed by the radiographic image alignment method according to the present invention.
FIG. 12 is a view showing a state in which alignment is performed by a conventional radiographic image alignment method;
FIG. 13 is a diagram showing the relationship between radiation and a threshold value
FIG. 14 is a diagram in which a region separated by a radius r or more from a corresponding point that does not match the evaluation is represented by hatching
FIG. 15 is a schematic view of a radiographic image recording apparatus for obtaining radiographic images to be superposed according to an embodiment of the present invention.
[Explanation of symbols]
1 Radiography system
2 Radiation sources
3 radiation
4 Subject
5 First stimulable phosphor sheet
6 Filter
7 Second stimulable phosphor sheet
10 Radiation image reader
15 Sheet conveying means
16 Laser light source
17 Light beam
18 Motor
19 Rotating polygon mirror
20 Focusing lens
21 mirror
22 photostimulated light
23 Light guide
24 photo multiplier
25 Log amplifier
26 A / D converter
30 Image processing display device
31 keyboard
32 CRT display
33 floppy disk drive
34 Main unit
40A to 40D reference points
41A ~ 41D Template area
44A ~ 44D Corresponding point

Claims (3)

In a method for determining corresponding points for aligning a plurality of radiographic images,
Setting a template region on one of the radiographic images;
By performing template matching to obtain a correlation value between the template region and a part of the other radiation image by moving the template region on another radiation image other than the one radiation image, Find multiple corresponding points that correspond to each other,
Each corresponding point of one radiographic image in each radiographic image is set as a reference corresponding point,
For the plurality of corresponding points, a condition that a ratio between a distance between any two reference corresponding points and a distance between two corresponding points on another radiation image corresponding to the two reference corresponding points is within a constant value range. A corresponding point determination method for alignment of radiographic images, characterized in that evaluation is performed as to whether or not the condition is satisfied, and a corresponding point that matches the evaluation is used as a corresponding point for the alignment. The correlation value between the template region and a part of the other radiation image obtained by the template matching is equal to or greater than a predetermined value, and the distribution of the correlation values in the vicinity of the plurality of corresponding points is ridge-shaped. 2. The radiation according to claim 1 , further evaluating whether or not at least one of the conditions for not to be satisfied is satisfied, and setting a corresponding point that matches the evaluation as a corresponding point for the alignment. Corresponding point determination method for image alignment. If the number of corresponding points that match the evaluation is less than a predetermined number, the setting of the template region, the template matching, and the evaluation are performed on a radiation image that is a predetermined distance or more away from the corresponding points that did not match the evaluation. The method of determining corresponding points for alignment of radiographic images according to claim 1, wherein the matching is repeated until the predetermined number of corresponding points matching the evaluation reaches the predetermined number.

Images