COMPUTER ASSISTED CANCER DIAGNOSIS SYSTEM

USING PET/CT DELAYED SCAN IMAGE

Rui Zhang, Takako Sato and Hiroshi Arisawa

Graduate School of Yokohama Nat. Univ., 79-7, Tokiwadai, Hodogayaku, Yokohama, 240-8501,Japan

Keywords: PET/CT, Whole-body delayed scan, CAD, CADS, Tumor detection, SUV, DTA.

Abstract: In this paper, a new method for CADS (computer assisted diagnosis system) is proposed: whole-body

PET/CT delayed scan. Whole-body PET/CT imaging is quite useful for detecting cancerous regions,

however sometimes too many “suspicious areas” due to ambiguous signs within 3D images. We propose a

method in which two images i.e. original images and whole-body delayed scan images are compared and

the true positive regions are identified. For the purpose of detection, a matching algorithm for comparing

each region on both two images has been developed. It contributes not only to improve accuracy of a

diagnosis but to reduce “false positive” regions. We compare this new method with the routine one and

show its supporting technologies and advantages. In the end we indicate our research emphasis.

1 BACKGROUND

Since computer-aided diagnosis (CAD) was first

researched by the laboratory of Kurt Rossmann

during the 1980’s, the research on CAD systems has

been increasing, which constituted a sturdy

foundation for the basis of CAD systems. One

application for these systems is aiding radiologist

with growing burden of making effective and

efficient diagnoses from increasing amounts of

imaging data. For that end, a large number of

research on computer assisted diagnosis system

(CADS) has been proposed (Jiang et al, 2001)

(Toriwaki,

2000) (Tsai and Lee, 2001) (Cheng,

Akiyama, Wang, Itoh, 1998) (Ukai, 2000).

The diagnostic method for cancer detection using

the PET/CT images is a core technology in CADS; it

has shown its superiority in the clinical management

of cancer. During cancer inspection by PET/CT scan,

drugs called FDG (fluorodeoxyglucose) are injected

into patient. After 1 hour, gamma rays emitted from

the patient are photographed by the nuclear imaging

system. Because much more FDG is taken into a

cancer cell than a normal cell, as a result, we

observed the concentration of FDG absorbed by the

tissues and organs to distinguish between normal and

abnormal areas.

To further improve the accuracy of the diagnosis,

a second scan, known as PET delayed scan is

performed. A half hour after the PET/CT scan a PET

delayed scan is performed at specific areas where

potential tumor was shown in the original

whole-body scan. By comparing the original image

with the delayed scan image, a more accurate

diagnosis is achieved. If the value of SUV (standard

uptake value) of a certain part in the delayed scan

image is higher than the corresponding part’s SUV in

the whole-body scan, this part is likely to be a

malignant pattern which indicates a true positive.

Otherwise this part is likely to be a benign pattern

which means it is a true negative. PET delayed scan

is not very prevalent in the diagnosis for tumor

detection. Only a few hospitals use it for about 10%

of their patients.

Still PET delayed scan is one of the promising

technologies for the medical image processing

system. But this technology is only applied to

specific parts of the body but not to the entire body.

So we propose a new method of cancer diagnosis：

Whole-body PET/CT delayed scan with CADS.

2 PROCEDURE OF NEW

METHOD OF WHOLE-BODY

PET/CT DELAYED SCAN

A large number of data and experiments testify the

usefulness of dual time imaging in tumor detection

(Matthies et al, 2002) (Zhuang et al, 2001).

399

Zhang R., Sato T. and Arisawa H..

COMPUTER ASSISTED CANCER DIAGNOSIS SYSTEM USING PET/CT DELAYED SCAN IMAGE .

DOI: 10.5220/0003287203990404

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 399-404

ISBN: 978-989-8425-34-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Part-body delayed scan could distinguish between

malignant and inflammatory lesions by comparing

SUVs of corresponding sites in two images thereby

reducing the false positive results. FDG is absorbed

adequately by the tumor during the half hour after

the first scan so that cancerous areas are diagnosed

more easily, which means delayed scan could

decrease false negative results. However, the

disadvantage of dual time imaging for part of the

body is also obvious. The cancerous cells have the

characteristic of normal metastasis. The site where a

tumor is may be not found if this site shows a false

positive result in the first scan and is excluded from

the delayed scan. So we propose the new method of

PET/CT whole-body delayed scan.

Figure 1 is the description of radiologist’s routine

method of PET/CT scan. Figure 2 describes the

diagnostic process of a delayed scan which is

utilized but not so prevalently nowadays, and Figure

3 shows the procedures of tumor detection using a

whole-body PET/CT delayed scan in the future.

In figure 1, patients accept the whole-body

PET/CT scan only once, and then radiologist makes

the final diagnosis by the PET/CT image. For the

figure 2 and 3, there are three differences between

these two figures: firstly, it is obvious that the

delayed scan of figure 3 is taken over the whole

body while the other is only the suspicious area;

secondly, recently, only 10% of patients could be

subject to a second PET scan on the suspicious parts;

we hope in the future this new cancer diagnosis

method of whole-body PET/CT delayed scan could

be used for all the patients after the first PET scan.

Thirdly, the computer assisted diagnosis system

(CADS) in figure 3 is utilized.

In figure 3, most diagnoses and comparisons are

made by the CADS, not by the doctor. The

whole-body delayed scan process produces the

double workload of reading images for radiologists

because the area of the second image increases to the

whole body. That is the main reason why

whole-body delayed imaging has not been used in

clinical medicine. The CADS will solve this problem

by reducing false positive results significantly and

having the task of comparing the original images

with the delayed images done by a computer. The

doctor only does the final diagnose which is

simplified by the CADS. So the CADS constitutes

the principal advantage and precondition for the

deploying of whole-body delayed scan. Now

Arisawa laboratory in Japan is devoted to the study

of the CADS.

The CADS is a system which imitates diagnosis

of radiologist. The most important part of the CADS

is a diagnosis algorithm which could interpret a

doctor’s methodology during PET/CT scan diagnosis

and uses the proper PET/CT terminology to describe

the program’s action so that it would simulate the

performance of the radiologist. This system is

designed so that a doctor can monitor the process of

diagnosis from the local to the whole body level, to

evaluate diagnosis’s validity and to recommend

improvement.

Figure 1: Routine method of PET/CT scan.

Figure 2: Diagnostic process of delayed scan which is

utilized but not so prevalently nowadays.

3 SUPPORTING

TECHNOLOGIES

The realization of this new method needs many

technologies. Now we introduce supporting

technologies which we are devoted to research.

3.1 Multi-organ Identification and

Cancer Diagnosis

Obtaining an organ’s shape and position is crucial

Doctor’s

diagnosis

PET/CT scan

Final result

images

PET Delayed

scan

for the

suspicious

areas

Delayed

images for

10% patients

Doctor’s

diagnosis

PET/CT

First scan

Doctor’s diagnosis by

comparing two kinds of

images

30mins later, 10%

patients

Final

result

images

10% original

images

90%patients

HEALTHINF 2011 - International Conference on Health Informatics

400

Figure 3: Procedure of tumor detection with the new

method of PET/CT scan for the whole body in the future.

step before tumor detection. Some researchers think

that the first processing step of a CAD system in the

future should understand the normal structure of the

human body by the processing of input images

(Kobatake, 2009). The outline image of an organ

from a PET/CT scan is very useful to confirm the

location of the accumulation of cancer. In general,

PET images readily allow identification of certain

organs like the brain due to high metabolism, while

CT images are most suitable for extracting the areas

of the lungs and the skeleton due to their distinctly

differences in density compared to soft tissue (Wen,

Leung, Eberl, Feng, Bai, 2008). But this issue is the

difficult point in the research of cancer diagnosis all

along.

Multi-organ cancer diagnosis is the future of

computer assisted diagnosis system. There are

different criterions for tumor detection in different

organs. Existing works are mostly focusing on

specific organs such as lung and breast (Takeo,

Shimura, Imamura, Shimizu, Kobatake, 2005).

Research about other organs such as liver,

gallbladder and so on are in the primary stage.

3.2 Dynamic Threshold Adjustment

Setting appropriate threshold values such as SUVs in

PET/CT images contributes to not only identify the

organs and acquire the figures of organs exactly but

also finding tumors and extract the shape of cancer.

Even in one organ, dynamic threshold values are

needed to obtain the organ’s boundary and ensure the

location of cancerous cells. DTA (dynamic threshold

adjustment) method (Arisawa, 2009) is proposed for

this situation. DTA is useful for extracting the

outline of an organ and position of a tumor, it could

elevate the precision of diagnosis and it is an

important focus for further research

In cancer detection, DTA method can notice fine

differences of threshold values of abnormality. That

is, in the preliminary stage, detecting whole

“critical” areas (c-areas) which have sharp increasing

of SUV locally, then calculating the average value

(c-value) of the c-area. This c-area should be

extended if some spots have a higher SUV than

c-value around the original c-area. Finally, for all the

extended c-area refiltering will be done depending

on the organ-specific value. The remaining c-areas

are considered “abnormal” accumulations (Arisawa,

2009).

Figure 4: Target areas for tumour detection.

Figure 5: Extracting the areas which have sharp increasing

of SUV locally, and then calculate the average value of

c-area (c-value of area No1 is 3.0; c-value of No 2 is 2.8;

c-value of No 3 is 4.0).

Figure 6: Obtaining the suspicious areas based on the

c-value of No.1 area.

Figure 7: Preserving the areas which overlap with No.1

area.

First PET/CT

scan

Comparing the two

kinds of images by

CADS

Whole-body PET/CT

delayed scan

Delayed

images for

all patients

Simply diagnosis by doctor

images

Final result

30mins later, all

patients

2 3

COMPUTER ASSISTED CANCER DIAGNOSIS SYSTEM USING PET/CT DELAYED SCAN IMAGE

401

Figure 8: Applying the same operation on the No.2 and

No.3 areas.

Figure 9: Result of tumour candidates by the DTA method.

3.3 Position Adjustment of Images

Position adjustment between previous PET/CT

image and delayed scan PET/CT image is an

important step to ensure whether the areas have

cancer or not. There is a wait of about 30 minutes

between first PET/CT scan and the delayed scan, the

two images subject to many differences caused by

changes in a patient’s laying positions and persistent

absorption of FDG by organs during the 30 minute

interval. For example, the position of the body in

both images is changed; the areas of FDG

accumulation in both pictures are altered. Research

involving position adjustment of images includes

two aspects: (1) image adjustment, and (2)

accumulation area adjustment (HONG, Sato,

Arisawa, 2008).

(1)Image adjustment includes rescaling of size

and barycenter between two images and so on. We

rescale the size of delayed image by the method of

Linear Interpolation.

In normal condition, the patient’s barycenters of

original image and second image are different.

Figure 10: The difference between barycenters of two

images.

The computing formula of patient’s barycenter of

original image:

（Xc, Yc）

(1)

The computing formula of patient’s barycenter of

delayed image:

（X’c, Y’c）

(2)

The differentials between patient’s barycenters of

two images:

Δx = X’c – Xc

(3)

Δy = Y’c – Yc

(4)

So we can rescale the delayed image by the

formulas of relationship between the original

image’s point (x, y) and delayed image’s point（x’,

y’）:

x’ = x + Δx

(5)

y’ = y + Δy

(6)

(2)

Accumulation area adjustment is to make

accumulation mapping for two images. There are

four possibilities for accumulations’ changes by

comparing the original image and delayed image: (a)

the SUV of delayed image becomes higher than the

SUV of original image(in figure, SUV = 6.7→7.9);

(b)

Two accumulations combine together; (c) One

accumulation separates to two parts; (d) The position

of accumulation in image is changed.

For the accumulation of (a), this part indicates

true positive and accumulation of (d), we could

conclude it is false positive result because the

cancer’s accumulation could not

transfer.

Accumulation (b) and (c), we should do further

analysis to make sure whether there is cancer or not.

Figure 11: The changes between original image’s

accumulations and delayed image’s accumulations.

2 3

HEALTHINF 2011 - International Conference on Health Informatics

402

3.4 Comparing the SUVs of Two

Images

During the interval of 30 minutes between first scan

and delayed scan, normal cells will release the FDG

so that SUV of normal region decreases; meanwhile

cancer cell continue to absorb the FDG in that 30

minutes so that its SUV increases. Therefore,

comparing the SUVs of the first scan image with the

ones of the delayed scan image is a research subject

for this new method. Generally speaking, lesions

such as in the lung area with decreased or stable

SUVs over time are likely to have a benign etiology.

In contrast, lesions with increased SUVs over time

are likely to be caused by malignancy (Matthies et al,

2006). At present there are no identified standards

for whole-body tumor detection by comparing SUVs

of two images, which is one of our research

objectives.

4 EXPERIMENTAL RESULT

Using the above procedure, we have already

implemented a preliminary experiment. The result

shows that a delayed scan with CADS reduces the

false positive results, which mean a more precise

diagnosis is achieved.

Figure 12: Experimental result by 3 different methods.

Figure 13: PET image of case 1.

5 ADVANTAGES

5.1 Accomplishment of More

Comprehensive Diagnosis

More comprehensive diagnosis could be achieved by

the whole-body PET /CT delayed scan. Now PET

delayed scan of specific parts of the body allows a

doctor to make omissions of cancerous area during

the diagnosis process. Unlike other diseases, cancer

is not necessarily only limited to one primary organ,

but also may involve local and distant metastases to

the lymphatic system and other organs (Knoepp and

Ravenel, 2006). So PET /CT delayed scan for the

whole body diagnosed is a potentially promising

method for the tumor detection.

5.2 Achieving More Accurate Diagnosis

The CADS makes the result of tumor detection more

reliable and precise. Mistakes maybe made by the

radiologist’s diagnoses because of doctor’s

negligence for the tiny accumulation of high SUV.

But in CADS, there are standard SUVs which are

determined by computer to distinguish tumor areas

and normal areas. The CADS has a higher sensitivity

for the SUVs so as to make more accurate diagnosis.

5.3 Decreasing the Burden of

Radiologist

Nowadays, a great number of diagnoses by doctors

are needed in tumor detection. Adding CADS

processing to doctors’ diagnoses by the whole-body

PET/CT delayed scans will simplify the diagnosis

for radiologists, especially for the doctors in the

countries where PET/CT imaging for tumor

detection is a regular health test for the patients.

6 CONCLUSIONS

The new method of PET/CT delayed scan for the

whole body has been briefly introduced. We hope

delayed scan in the future could apply to all the

patients, not only 10% patient, and achieve

whole-body PET/CT delayed scan instead of

part-body PET delayed scan. For the realization of

this new method, computer algorithms are extremely

important to make the diagnosis active and

automated. Whole-body PET/CT delayed scan

should be one of the trends in computer assisted

diagnosis system. In the future, our research will

COMPUTER ASSISTED CANCER DIAGNOSIS SYSTEM USING PET/CT DELAYED SCAN IMAGE

403

focus on position adjustment of images in many

situations such as patient’s arms putting down beside

body during the first scan in order to capture the

arms but holding them up over the head during the

second scan in order to reduce noises, which has a

big influence in PET/CT imaging for the breast area.

ACKNOWLEDGEMENTS

We express our appreciation to Dr. Ryogo

Minamimoto of National Center for Global Health

and Medicine, Professor Tomio Inoue of the

Yokohama City University Graduate School of

Medical Studies and Dr. Seishi Jinnouchi of Atsuchi

Neurosurgical Hospital PET Center.

REFERENCES

Jiang, H., Masutoh, N., Nishimura, O., Okumura, T.,

Yamamoto, S., Iisaku, S., Matsumoto, T., Tateno, Y.,

Iinuma T., Matsumoto, M., 2000. Computer-Aided

Diagnosis System of Lung Cancer Screening by

X-Ray CT, IEICE TRANS.(D-II)

，

J83-D-II (1),

333-341.

Toriwaki, J., 2000. Trends and Future Problems in

Computer Aided Diagnosis of X-ray Images, IEICE

TRANS.(D-II), J83-D-II (1), 3-26.

Tsai, D., Lee, Y., 2001. Fuzzy Reasoning Based

Computer-Aided Diagnosis for Automated

Discrimination of Myocardial Heart Disease from

Ultrasonic Images, IEICE TRANS.(A)

，

j84-A(12)

，

1431-1438.

Cheng, X., Akiyama, I., Wang, Y., Itoh, K., 1998. Breast

Tumor Diagnosis System Using Three Dimensional

Ultrasonic Echography，IEICE TRANS.(D-II)

，

J81-D-II

(1), 1402-1410.

Ukai, Y., Niki, N., Satoh, H., Watanabe, S., 2000.

Computer Assisted Diagnosis System for Coronary

Calcifications Based on Helical CT Images，IEICE

TRANS.(D-II)

，

J83-D-II (1), 342-350.

Matthies, A., Hickeson, M., Cuchiara, A., et al.2002,Dual

time point 18F-FDG PET for the evaluation of

pulmonary noudles. J. Nucl. Med., 43;871-875.

Zhuang, H., Pourdehnad, M., Lambright, E. S., et al.2001.

Dual time point 18F-FDG PET imaging for

differentiating malignant from inflammatory processes.

J. Nucl. Med., 42(9);1412-1417.

Takeo H., Shimura K., Imamura T., Shimizu A., Kobatake

H., 2005, Detection System of Clustered

Microcalcifications on CR Mammogram (Biological

Engineering), IEICE transactions on information and

systems, E88-D(11), 2591-2602

Hidefumi Kobatake, 2009.Trends in Research on CAD in

Japan, The Institute of Electronics,Information and

Communication Engineers—IECE Technical Report

217~218.

Lingfeng Wen, Wilson Leung, Stefan Eberl, Dagan Feng,

Jing Bai, 2008, A preliminary study on the

knowledge-based delineation of anatomical structures

for whole body PET-CT Studies, Proceedings of the

5th International Conference on Information

Technology and Application in Biomedicine

Hiroshi Arisawa, 2009 . Improvement of Automated

Cancer Detection System Using PET-CT Images,

RNSA pp.921

Hoang Thuy Hong, Takako Sato, Hiroshi Arisawa, 2008

Improvement of Cancer Detecting Ability in PET

Automated Diagnosis System by Using Delayed Scan

Images, Paper Volume of the 7th Information

Technology Forum of Information

ProcessingAacademy,pp.453-454, Sep.

U. W. Knoepp and J. G. Ravenel, "CT and PET imaging in

non-small cell lung cancer," Critical Reviews in

Oncology Hematology, vol. 58, pp. 15-30, 2006.

HEALTHINF 2011 - International Conference on Health Informatics

404