SPECIFICATION [TITLE OF THE INVENTION]

Method and apparatus for cerebral haemorrhage segmentation [TECHNICAL FIELD]

The present invention relates to a method and apparatus for cerebral haemorrhage segmentation, more specifically to a method and apparatus for identifying the cerebral haemorrhage based on an

X-ray CT image of the head having the cerebral haemorrhage onset.

[BACKGROUND ART]

The cerebral haemorrhage is segmented on the head X-ray CT image for the diagnosis and treatment of the cerebral haemorrhage. The segmentation is manually conducted by the intervention of a specialist (for example, see patent reference 1) . [patent reference 1] JP-A-2005-118510 [DISCLOSURE OF THE INVENTION] [PROBLEM TO BE SOLVED BY THE INVENTION]

Manual segmentation by the intervention of a specialist takes time and labor. In addition the result of the segmentation is depending on the skill of the physician. The automation of the segmentation by a single threshold may not solve the problem because the CT value at the cerebral haemorrhage may vary in relation to the symptom, and may overlap to the CT values of healthy part.

An object of the present invention is to provide a method and apparatus for appropriate segmentation of cerebral haemorrhage lesion.

[MEANS FOR SOLVING THE PROBLEM]

First aspect of the present invention for solving the problem described above provides a method of segmenting a cerebral haemorrhage site in a medical image of a head, comprising the steps of: segmenting an internal region of a skull bone in the medical image of said head; segmenting a possible region in which a cerebral haemorrhage site is contained, out of a region segmented in said step of segmenting the internal region of the skull bone; and determining a cerebral haemorrhage site out of a region segmented in said step of segmenting the possible region in which the cerebral haemorrhage site is contained.

More preferably, said step of segmenting a possible region in which a cerebral haemorrhage site is contained includes: a step of segmenting a region with CT values gradually changing out of the region segmented in said step of segmenting the internal region of the skull bone ; a step of segmenting a region with the number of pixels larger than a predetermined number, out of the region segmented in said step of segmenting a region with CT values gradually changing; a step of segmenting a region with a comparatively large CT value out of the region segmented in said step of segmenting a region with the number of pixels larger than a predetermined number; and a step of segmenting a region with CT values gradually changing

out of the region segmented in said step of segmenting a region with a comparatively large CT value.

More preferably, said step of determining a cerebral haemorrhage site includes: for the region segmented in said step of segmenting the possible region in which the cerebral haemorrhage site is contained, a first step of determining whether or not said segmented region is a cerebral haemorrhage site based on a CT value and a size of said segmented region; and for a potential region of the cerebral hemorrhage site out of a region other than the region which has been determined as the cerebral hemorrhage site in said first step, a second step of determining whether or not said potential region is the cerebral haemorrhage site based on a CT value difference between said potential region and a vicinity thereof or a CT value of said region.

More preferably, said method further including a step of correcting the influence of a partial volume effect for the region which has been determined as the cerebral haemorrhage site in said step of determining a cerebral haemorrhage site.

More specifically, the present invention provides a method for cerebral haemorrhage segmentation comprising the steps of preprocessing, primary searching, analyzing and adjusting, filtering, secondary searching, analyzing and determining, and postprocessing; in which the preprocessing step, on the X-ray CT image of the head,

excludes pixels having a CT value larger than a first setting value and pixels having a CT value less than a second setting value, identifying the boundary of the skull bone based on a third setting value, and excludes the outer region of the skull bone based on the boundary identified; the primary searching step, on the image that has been processed by the preprocessing, searches the region where the CT value varies gradually, and labels each region found by the searching; the analyzing and adjusting step determines, on the image on which the first searching has been performed, the number of pixels for each of the regions and excludes the region having the number of pixels less than a fourth setting value; the filtering step determines, on the image on which the analyzing and adjusting step has been performed, the sum of the absolute value of the difference of the CT value pixel by pixel between adjacent two pixels in the direction that a two-dimensional coordinates i, j is increasing, for each of the regions, identifies the pixel position that the sum is more than a fifth setting value, determines the mean value of CT values at all of the pixel positions identified, and excludes the pixels having a CT value equal to or less than the mean value; the secondary searching step searches, on the image on which the filtering has been performed, a region where the CT value gradually changes, and newly labels each region found by the

searching; the analyzing and determining step sets, on the image that has been searched by the second searching, a first index IndeXc _T to

[equation 19]

( _^ l region Ci Mm

Index _cτ ^:

CT blood CT Mm

i f CT _1111n <— CT _Regl0n <— CT _BjOod , where CT _Reglon is the CT value of the pixel in the region, CT _max is sixth setting value, CT _1111n is seventh setting value, and CT _B i _∞d is eighth setting value; or sets to

[equation 20]

/IT /" ¹ T

¹ _^ J- max Li region

Index or ⁼

C 1 max ^~ Ci blood

if CT _Blood < CT^ _g10n <= CT _1112x or else sets to [equation 21 ]

Index _cτ = 0; when the surface area and the perimeter length of the region are indicated as Area and Perimeter, and the characteristics value of the region is indicated by

[equation 22]

Radian = Area / Perimeter ² then a second index Index _Radian is set to

Index^^ = 1 if Radian is more than a ninth setting value Radian^,

Index^ _an = 0 if Radian is less than a tenth setting value Radian^, and to

[equation 23]

Radian,egion ^~ RadiariMm

Index r _a d _o n =

Radian _Max - Radiatio _n

if Radian is less than or equal to the ninth setting value Radian^ and more than or equal to the tenth setting value Radian^, sets a third index index ₀ to

[equation 24]

X-idex ₀ = Index _cτ * Index _radlan , then determines that the regions has the cerebral haemorrhage if index ₀ >= 20%, or that the region has not the cerebral haemorrhage if Index,, <= 3%, or, if 3 % < Indexo < 20%,

is the CT value of the pixel of the surrounding region of the region, and C T _Average _ _A1 i_ _Reg i _o ii is the mean CT value of all regions , then sets a fourth index to

[equation 25]

IndeXSub — CT Region ^~ CT Around ^' Region r and a fifth index Indexo^ to [ equation 26]

lnCieX _OrdθJ: — Ll _Reglon — CT _Average _ _M1 - _Re gion I and determines that the region has a cerebral haemorrhage if Index _^b >= 8 or Index _Older >= 10, or determines that the region has not a cerebral haemorrhage if Index _sub <= 0 or Index _Ordβr <= -5, or, if 0 < IndeX _sub < 8 and -5 < Index _Order < 10 , then set a sixth index Index _βnal to

[equation 27]

Index ^■ Final = Index * Indexes - (S)

10 - (-5)

and determines that the region has a cerebral haemorrhage if Index _Fma i >= 50%, or determines that the region has not a cerebral haemorrhage if Index _Fmal < 50%; and the postprocessing step compensates for the influence of partial volume effect on the region determined as having a cerebral haemorrhage in the analyzing and determining step.

A second aspect of the present invention for solving the problem described above provides an apparatus for segmenting a cerebral haemorrhage site in amedical image of a head, comprising: means for segmenting an internal region of a skull bone in the X-ray CT image of said head; means for segmenting a possible region in which a cerebral haemorrhage site is contained, out of a region segmented by said means for segmenting the internal region of the skull bone; and

means for determining a cerebral haemorrhage site out of a region segmented by said means for segmenting a possible region in which a cerebral haemorrhage site is contained.

More preferably, saidmeans for segmenting a possible region in which a cerebral haemorrhage site is contained includes: means for segmenting a region with CT values gradually changing out of the region segmented by said means for segmenting the internal region of the skull bone; means for segmenting a region with the number of pixels larger than a predetermined number, out of the region segmented by said means for segmenting a region with CT values gradually changing; means for segmenting extracting a region with comparatively large CT value out of the region segmented by said means for segmenting a region with the number of pixels larger than a predetermined number; and means for segmenting a region with CT values gradually changing out of the region segmented in said step of segmenting a region with comparatively large CT value.

More preferably, said means for determining a cerebral haemorrhage site includes: for the region segmented by said means for segmenting the possible region in which the cerebral haemorrhage site is contained, first means for determining whether or not said segmented region is a cerebral haemorrhage site based on a CT value and a size of said segmented region; and

for a potential region of the cerebral hemorrhage site out of a region other than the region which has been determined as the cerebral hemorrhage site by said first means, second means for determining whether or not said potential region is the cerebral haemorrhage site based on a CT value difference between said potential region and a vicinity thereof or a CT value of said region.

More preferably, said apparatus further including means for correcting the influence of a partial volume effect for the region which has been determined as the cerebral haemorrhage site by said means for determining a cerebral haemorrhage site.

More specifically, the present invention provides an apparatus for segmentation of a cerebral haemorrhage region on an X-ray CT image of a head, comprising a preprocessing means, a primary searchingmeans, an analyzing and adjustingmeans, a filteringmeans, a secondary searching means, an analyzing and determining means, and a postprocessing means; in which the preprocessing means, on the X-ray CT image of the head, excludes pixels having a CT value larger than a first setting value and pixels having a CT value less than a second setting value, identifies the boundary of the skull bone based on a third setting value, and excludes the outer region of the skull bone based on the boundary identified; the primary searching means, on the image that has been preprocessed by the preprocessing, searches the region where the CT value gradually changes, and labels each region found by the

searching; the analyzing and adjusting means, on the image that has been searched by the primary searching, determines the number of pixels for each of the regions, and excludes the region having the number of pixels less than a fourth setting value; the filtering means, on the image that has been analyzed and adjusted, for each of the regions, determines the sum of the absolute value of the difference of the CT value between adjacent pixels for each pixel in the direction wherein a two-dimensional coordinate i, j is increasing, identifies the pixel position that the sum is more than a fifth setting value, determines the mean value of CT values at all of the pixel positions identified, and excludes the pixels having a CT value less than the mean value; the secondary searching means, on the image that has been filtered by the filtering, searches a region where the CT value gradually changes, and newly labels each region found by the searching; the analyzing and determining means, on the image that has been searched by the secondary searching, sets a first index InCJeX ₀₁ to

[equation 28]

C-- J region Oi Min

IndexcT ^'

CT blood CT Min

¹ ^ CT _nUn <" CT _Regioll <— CT _Blood , or sets to

[equation 29]

C- 1 max C/ ; region IndexcT ^~

Ci max C- 1 blood

i f U l _B i _00C i < C I _j ^g _10n <= L, l _maχ , or else sets to [equation 30]

Index _cτ = 0, where CT _Reeon is the CT value of the pixel in the region,

CT _∞ax is a sixth setting value,

CT _1111n is a seventh setting value, and

CT _Blood is an eighth setting value ; when the surface area and the perimeter length of the region are indicated as Area and Perimeter, and the characteristics value of the region is indicated by

[equation 31 ]

Radian = Area / Perimeter ² then sets a second index Index _Radian to

1 if Radian is more than a ninth setting value Radian _max ,

Index _Rildim = 0 if Radian is less than a tenth setting value Radian _^ , and to

[equation 32 ]

Rαdiαriregwn ^~ RαdiαuMm

Index _n o _w , =

RαdimiMαx ^~ Rαdicm _Mm

if Radian is less than or equal to the ninth setting value Radian^ and more than or equal to the tenth setting value Radian^, sets a third index index ₀ to

[equation 33] Index,, = Index _cτ * Index _radian , then determines that the regions has the cerebral haemorrhage if index ₀ >= 20%, or determines that the region has not the cerebral haemorrhage if index ₀ <= 3%, or if 3 % < index ₀ < 20%, and when CT^^^ is the CT value of the pixel of the surrounding region of the region, and ^CT _ATCrage - _A ii- _Reg i _on is the mean CT value of all regions, then sets a fourth index Index^ to

[equation 34]

Lϊlύ&XSub *~- L Re gion (-> J- Around Re gion and a fifth index Index _o ^ to

[equation 35] Ind.ex _0rder =CT _Resion -CT _Avexage _ _Mj _ _{Region r} and determines that the region has a cerebral haemorrhage if Index^ >= 8 or Index _Oκter >= 10, or determines that the region has not a cerebral haemorrhage if Index _sub <= 0 or Index _Oκler <= -5, or, if 0 < IndeX _sub < 8 and -5 < Index _Order < 10, then set a sixth index Index _Flnal to

[equation 36]

Index _mml = Index, ^^ex _o> , _er - (-5) ^Fmal ° 10 - (-5)

and determines that the region has a cerebral haemorrhage if Index _Final >= 50%, or determines that the region has not a cerebral haemorrhage if Index _Final < 50%; the postprocessing means compensates for the influence of partial volume effect on the region determined as having a cerebral haemorrhage in the analyzing and determining.

The identification of the skull boundary in the preprocessing is preferably performed by detecting the CT value changing point from a value smaller than the third setting value to a larger value, or the CT value changing point from a value larger than the third setting value to a smaller value, in order to appropriately identify the boundary.

The region search in the primary searching and the second searching is preferably performed by searching a region that has the difference of CT value between adjacent two pixels of 5 or less, in order to appropriately search a region.

The compensation in the postprocessing is preferably performed by a dilation calculation with respect to the region in order to appropriately compensate for a region.

Preferably the first setting value is 245, the second setting value is 30, the third setting value is 190, the fourth setting value is 300, the fifth setting value is 4, the sixth setting value is

100, the seventh setting value is 40, the eighth setting value is 70, the ninth setting value is 0.015, and the tenth setting value is 0.003 in order to perform a segmentation in a high precision. [EFFECT OF THE INVENTION]

In accordance with the above aspects of the present invention, the index Index _cτ with respect to the CT value of the pixels in a candidate region and the index Index _Radian with respect to the characteristics of the candidate region are used to generate the index Index ₀ , the region of interest is determined to have a cerebral haemorrhage if index ₀ >= 20%, or the region is determined not to have a cerebral haemorrhage if index ₀ <= 3%. If 3% < index ₀ < 20% then the index Index^ and index Indexo _rde . _r are generated for the region, then the region is determined to have a cerebral haemorrhage if Index, _ub >= 8 or Index _Order >= 10, or the region is determined not to have a cerebral haemorrhage if Index^ <= 0 or Index^^ <= -5. If 0 < Index^ < 8 and -5 < Index _Order < 10, an index Index _Fma iis generated to determine that the region has a cerebral haemorrhage if Index _Fmal >= 50%, or th°at the region has not a cerebral haemorrhage if Index _Flnal < 50%. In this manner a method and apparatus for appropriately performing the cerebral haemorrhage segmentation is achieved. [BEST MODE FOR CARRYING OUT THE INVENTION]

Some best modes for carrying out the present invention will be described in greater details with reference to the accompanying drawings. It should be noted here that the present invention is not limited to the best modes for carrying out the invention. Now

referring to the drawings, Fig. 1 shows a schematic block diagram of an image processing apparatus.

This apparatus is an exemplary best mode for carrying out the invention. The arrangement of the apparatus illustrates an exemplary best mode for carrying out the invention for achieving an apparatus for the cerebral haemorrhage segmentation. The operation of the apparatus illustrates an exemplary best mode for carrying out the invention for achieving a method for the cerebral haemorrhage segmentation.

As shown in Fig. 1, the apparatus includes a data processing unit 10, a display unit 20, an operation console 30, a storage unit 40, and an input and output unit 50. The data processing unit 10 performs a predetermined data processing on the data stored in the storage unit 40 based on the interactive operation by a user through the display unit 20 and the operation console 30.

The data processing unit 10 also performs data input and output through the input and output unit 50 to an external device. The X-ray CT images to be subject of the cerebral haemorrhage segmentation will be input through the input and output unit 50.

Some typical examples of the external devices include an X-ray CT apparatus and a medical image server. The apparatus may also be part of an X-ray CT apparatus or a medical image server. In the latter case the input and output unit 50 is not necessarily required. Fig .2 shows an example of X-ray CT image to be subj ect of the cerebral haemorrhage segmentation.

The cerebral haemorrhage segmentation to be performed on the apparatus will be described in greater details herein below. Fig. 3 shows steps of cerebral haemorrhage segmentation. As shown in Fig. 3, the cerebral haemorrhage segmentation is performed through seven steps Pl to P7.

The process step Pl is a preprocessing step. The process step P2 is a first searching step. The process step P3 is an analyzing and adjusting step. The process step P4 is a filtering step. The process step P5 is a secondary searching step. The process step Pβ is an analyzing and determining step. Finally the process step P7 is a postprocessing step.

These steps are executed by the data processing unit 10. The data processing unit 10 is an example of the preprocessing means, an example of the primary searching means, an example of the analyzing and adjusting means, an example of the filtering means, an example of the secondary searching means, an example of the analyzing and determining means, and an example of the postprocessing means. These steps will be described in greater details herein below.

Fig. 4 shows detailed substeps of the preprocessing step Pl. As shown in Fig. 4, the preprocessing step Pl, in step 101, selects pixels of an X-ray CT image of a head. The pixel selection is by excluding the pixels having a CT value larger than a first setting value and the pixels having a CT value less than a second setting value from the entire image. The first setting value may be 245,

and the second setting value may be for example 30. Therefore the pixels having a CT value larger than 245 and the pixels having a CT value less than 30 will be excluded.

In step 102 the skull boundary is identified. The identification of the skull boundary is performed based on a third setting value. The identification of the skull boundary is by detecting for the entire image any CT value changing points from a value less than the third setting value to a larger value, or any CT value changing point from a value more than the third setting value to a lesser value. The third setting value maybe for example 190. Based on the skull boundary thus identified, the external region of the skull bone is excluded in step 103, thus the internal region of the skull bone is segmented.

In steps 101 to 103, any pixels having a CT value larger than 245, and pixels having a CT value less than 30, and the region external to the skull bone are excluded. By the preprocessing as such, an image such as shown in Fig. 5 may be obtained. In Fig. 5, region blotted by black is indicative of a region not excluded. This black region is the target of next processing step.

Fig. 6 shows detailed substeps of the primary searching step P2. As shown in Fig. 6, in primary searching step P2, on the preprocessed image, a region in which the CT value changes gradually is searched in step 201. The search for the region having a gradually changing CT value is by searching any region that a difference of the CT value between adjacent two pixels is 5 or less for example.

In step 202 the regions detected by the search are labeled, thus the region having a gradually changing CT value is segmented.

Fig. 7 shows detailed substeps of the analyzing and adjusting step P3. As shown in Fig. 7 , in the analyzing and adjusting step P3, the image having the primary search performed is counted region by region the number of pixels in step 301. In step 302 any region having the number of pixels less than a fourth setting value is excluded, thus the region having the number of pixels no less than a fourth setting value is segmented.

The fourth setting value may be for example 300, when defined field of view (defined FOV) is 25 cm. The setting value may be adjusted to an appropriate value other than 300 if the defined FOV is not 25 cm.

From the primary searching and the analyzing and adjusting as described above, an image such as shown in Fig. 8 can be obtained. In Fig, 8, the region blotted by black is indicative of the region having the number of pixels of 300 or over and a gradual CT value change. This region becomes the target of the next processing step.

Fig. 9 shows detailed substeps of the filtering step P4. As shown in Fig. 9, the filtering step P4 determines, on the image that has been analyzed and adjusted, pixel by pixel the sum of the absolute value of the difference of CT value between adjacent two pixels in the direction that two dimensional coordinate i, j increases, for each pixel and for each region, in step 401. In other words a calculation as shown below is performed.

[equation 37]

G [i, j ] =abs (F [i, j ] -F [i+l , j ] ) +abs (F [i, j ] -F [i, j+l ] ) where F[i, j] is the CT value of the pixel at the two- dimensional coordinate i, j; F[i+1, j] is the CT value of an adjacent pixel in the direction that the coordinate i is incrementing; F[i, j+1] is the CT value of an adjacent pixel in the direction that the coordinate j is incrementing. The relationship among CT values F[i, j], F[i+1, j], F[i, j+1], and F[i+1, j+1] is as shown in Fig. 10.

In step 402, a pixel position is identified where G[i, j] becomes more than the fifth setting value. The fifth setting value may be for example 4. In step 403, a mean value of the CT values at all of the pixel positions identified is determined, and in step 404 any pixels having a CT value equal to or less than the mean value is excluded, thus the region with remained pixels is segmented.

The mean of the CT value is determined from the original image for each region. The exclusion of pixels having a CT value equal to or less than the mean value is performed on the original image region by region. This allows the removal of background for each region.

By the filtering as described above, an image such as shown in Fig. 11 can be obtained. In Fig. 11, the region blotted by black is indicative of the region not excluded. This black region is the target of the next process step.

Fig. 12 shows detailed substeps of the secondary searching step P5. As shown in Fig. 12, the secondary searching step P5

searches regions having CT value gradually changing, in step 501, for the filtered image. The search for regions having CT value gradually changing is by searching the region that the difference of CT value between adjacent two pixels is equal to or less than 5. In step 502, the regions found by searching are newly labeled, thus the region having CT value gradually changing is segmented. By the secondary searching as described above, an image such as shown in Fig. 13 can be obtained.

Fig. 14 shows detailed substeps of the analyzing and determining step P6. As shown in Fig. 14, in the analyzing and determining step P6, a first index Index^. is determined for the labeled image on which the second search has been performed in step 601.

The index Index _cτ sets

[equation 38] l _^ -l region t- _' -/ Mm

Index _cτ ^~

CT blood CT, MIn

^χ f CT _1111n <— CT _Re&on <— CT _Blood , or [equation 39]

Ci max Cl legion

IndexcT ^{~ '}

(— ' J- JTiaϊc (-✓ 1 blood

if CT _Blood < CT _Regl0I1 <= CT _mEK , or else [equation 40 ] Index _aτ = 0

where CT _Reglon is the CT value of the region pixels, CT _101x is the sixth setting value, CT ₁₁₁₁₁₁ is the seventh setting value, and CT _Blood is the eighth setting value.

Here, the sixth setting value CT _max may be for example 100, the seventh setting value CT ₁₁₁₁₁₁ may be for example 40, and the eighth setting value CT _Blood may be for example 70.

In step 602, a second index Index _Radlan is determined. In the second index Index^^, when the surface area and the perimeter length of the region is identified as Area and Perimeter, and the characteristics value of the region is identified as [equation 41]

Radian = Area / Perimeter ² ,

Index _Radian = 1 if Radian is larger than the ninth setting value Radian _^ , Index _Radl!m = 0 if Radian is less than the tenth setting value Radia-w and

[equation 42]

Radian, egion ^~ Radianum lnύ&X radian

Radian _Max - Radianu _m

if Radian is less than or equal to the ninth setting value Radian _max and more than or equal to the tenth setting value Radian,^.

Here the ninth setting value Radian _∞ax may be for example 0.015, the tenth setting value Radian,^ may be for example 0.003.

In step 603, a third index index ₀ is set to

[equation 43]

Index ₀ ~ Index _cτ * Index _radiaa .

In step 604, it is determined whether the region of interest has a cerebral haemorrhage or not in accordance with the value of third index index ₀ . More specifically, if index ₀ >= 20% then the region has a cerebral haemorrhage (CH) , if index ₀ <= 3% then the region has not a cerebral haemorrhage (not CH) , else if 3% < index ₀ < 20% then the process proceeds to the next step 605.

In step 605, the fourth index Index^ and the fifth index Index^,. are determined. More specifically, the fourth index Index^ is set to

[equation 44]

Index Sub - CT Region ~ CT Around^ gwn and the fifth index Index^,. is set to

[equation 45]

In αex _Order =C T _Regi0R - CT _AverSge _ _M _z _ _{Regα on}

, when the CT value of the pixels of the surrounding region around the region 3% < index ₀ < 20% is CT^,^^, and the mean CT value of the entire region which is 3% < index ₀ < 20% is CT _Averate _ _A11 _ _Regum .

In step 606, it is determined whether the region of interest has a cerebral haemorrhage or not in accordance with the value of the fourth index Index^ and with the value of the fifth index Index^^ . More specifically, a region is determined to have a cerebral haemorrhage (CH) if Indeχ. _ub >= 8 or IndeXo _rder >= 10, or a region is determined not to have a cerebral haemorrhage (not CH) if Index^, <= 0 or Index _Ordcr <= -5, else the process goes to next step 607 if 0 < IndeXg _ub < 8 and -5 < Index _Order < 10 .

In step 607, the sixth index Index _Final is set to [equation 46]

Index- , = Index * ^IndeXa * ^{r ~ (r5'} ) .

F ⁱ n ^a l ⁰ I ₀ - (-5)

In step 608, a region is determined whether or not to have a cerebral haemorrhage in correspondence with the value of the sixth index Index _Froal . More specifically, a region is determined to have a cerebral haemorrhage (CH) if Index _Fmal >= 50%, otherwise a region is determined not to have a cerebral haemorrhage (not CH) if Index _Fma[ < 50%.

An accurate segmentation of a cerebral haemorrhage region is then performed in accordance with the three-step-analyze and determination as have been described above, and an image such as shown in Fig. 15 may be obtained. In Fig. 15, the region blotted by black is indicative of a cerebral haemorrhage. The region blotted byblack indicates the actual position and extent of cerebral haemorrhage with a high precision.

When the cerebral haemorrhage is in proximity of the skull bone, the partial volume effect may cause the CT value of the cerebral haemorrhage region to be changed, resulting in an inaccurate segmentation of the cerebral haemorrhage.

Fig. 16 shows an example. In Fig. 16 the region blotted by black is the region segmented. The contour of this region at the right hand side does not reach the skull bone boundary. This is because the CT value which has been changed due to the partial volume

effect was excluded. The segmentation result as described above may be compensated for the influence of the partial volume effect in the postprocessing step P7.

Fig. 17 shows detailed substeps of the postprocessing step P7. As shown in Fig. 17, in the postprocessing step P7, the region that has been determined to have a cerebral haemorrhage by the analyzing and determining, namely the CH region, is determined whether or not to have the CT value of the surrounding pixels larger than the mean CT value of the CH region in step 701. If the CT value of the surrounding pixels is larger, then a dilation calculation is performed until the perimeter of the CH region reaches the skull bone boundary. With this postprocessing, as shown in Fig. 18, an image with the influence of the partial volume effect being compensated for on the CH region may be obtained. [BRIEF DESCRIPTION OF THE DRAWINGS]

Fig.1 is a schematic diagram of an exemplary image processing apparatus in accordance with the best mode for carrying out the invention/

Fig.2 is a schematic diagram illustrating an exemplary X-ray CT image as gray-scale photograph;

Fig. 3 is a schematic diagram illustrating the steps for the cerebral haemorrhage segmentation;

Fig. 4 is a schematic diagram illustrating the details of the preprocessing step;

Fig. 5 is a schematic diagram illustrating an exemplary X-ray

CT image after the preprocessing step as gray-scale photograph;

Fig. 6 is a schematic diagram illustrating the details of the primary searching step;

Fig. 7 is a schematic diagram illustrating the details of the analyzing and adjusting step;

Fig. 8 is a schematic diagram illustrating an exemplary X-ray CT image after the primary searching and the analyzing and adjusting step as gray-scale photograph;

Fig. 9 is a schematic diagram illustrating the details of the filtering step;

Fig. 10 is a schematic diagram illustrating the spatial positional relationships between pixel values;

Fig.11 is a schematic diagram illustrating an exemplary X-ray CT image after the filtering step as gray-scale photograph;

Fig. 12 is a schematic diagram illustrating the details of the secondary searching step;

Fig.13 is a schematic diagram illustrating an exemplary X-ray CT image after the secondary searching as gray-scale photograph;

Fig. 14 is a schematic diagram illustrating the details of the analyzing and determining step;

Fig.15 is a schematic diagram illustrating an exemplary X-ray CT image after analyzing and determining step as gray-scale photograph;

Fig.16 is a schematic diagram illustrating an exemplary X-ray CT image after analyzing and determining step as gray-scale

photograph/

Fig. 17 is a schematic diagram illustrating the details of the postprocessing step; and

Fig.18 is a schematic diagram illustrating an exemplary X-ray CT image after postprocessing step as gray-scale photograph. [REFERENCE NUMERALS]

10 data processing unit

20 display unit

30 operation console

40 storage unit

50 input and output unit

Fig. 1 input and output unit 50 data processing unit 10 display unit 20 operation console 30 storage unit 40 Fig. 3 preprocessing step Pl first searching step P2 analyzing and adjusting step P3 filtering step P4 second searching step P5 analyzing and determining step P6 postprocessing step P7 Fig. 4 pixel selection 101 identification of skull bone boundary 102 exclusion of outer region of the skull bone 103 Fig. 6 searching region with CT value gradually changing 201 labeling the region 202 Fig. 7 determining the number of pixels for each region 301 excluding regions having fewer pixels 302 Fig. 9

G[i, j] = abs(F[i, j] - F[i+1, j]) + abs(F[i, j] - F[i, j+1] ) 401 determining pixel position of G[i, j] = 4 402 finding the mean of CT value at the determined pixel position 403 excluding pixels less than mean value 404 Fig. 12 searching region in which CT value gradually changes 501 labeling the region 502 Fig. 14 determining Indexc _T 601 determining Index _Radιan 602 determining index ₀ 603 index ₀ 604 determining Index^ and Index _Order 605

Index^, Index^.^ 606 determining Index _Fmal 607

Index _Fmal 608 Fig . 17 determining CT value of surrounding pixels 701 performing dilation computation 702