電腦輔助型態測定法在肝臟細針抽取抹片之研究

Computerized morphometry study in liver fine needle aspiration

溫千慧(Chien-Hui Wen)

指導教授 ： 蔡志仁 ； 黃明賢

高雄醫學大學/醫學院/醫學研究所/碩士(2006年)

摘要

肝臟細胞癌(Hepatocellular carcinoma)是肝臟原發性腫瘤中最常見的惡性腫瘤。在肝臟細針抽取抹片中，有時候在肝臟細胞癌的診斷是很困難的，特別是在鑑別診斷良性再生性非典型性變化的肝細胞與分化良好的肝臟細胞癌細胞。有時候在細胞學抹片診斷過程，因為看到型態變化不明顯的疑似惡性腫瘤細胞而導致病理醫師無法很有信心的診斷為一個惡性腫瘤，這種情形也會因為不同的病理醫師而有不同的診斷差異性存在，因此一個診斷惡性腫瘤的標準界定值(Cutoff value)是必須的。在本研究中，我們嘗試利用電腦輔助型態測定這種客觀、科學電腦輔助的影像分析方法，分析所選定的四種細胞核型態參數值在良性再生性非典型性變化的肝細胞、分化良好的肝臟細胞癌與懷疑惡性肝腫瘤細胞上的表現，進一步的再由分析的結果去找出一個診斷為分化良好肝臟細胞癌的標準界定值，藉由選定的標準值在往後肝臟細針抽取抹片上，希望可以用來輔助診斷細胞型態變化不明顯的一些懷疑惡性的個案，降低因不同的判讀病理醫師所產生的人為誤差以及提升在細胞學抹片診斷的準確率。
材料與方法：
所選定的78個樣本分別為30個細胞學抹片診斷為惡性(Positive)之後切片診斷為分化良好的肝臟細胞癌；30個細胞學抹片診斷為良性(Negative) 之後切片診斷為良性再生性非典型性變化的肝細胞；18個細胞學抹片診斷為懷疑惡性(Suspicious) 之後切片診斷為分化良好的肝臟細胞癌，所選取的樣本，利用電腦輔助型態測定研究，分析所選定的四項細胞核型態參數(Parameter)，包括細胞核面積(Area)、細胞核周長(Perimeter)、細胞核最大徑與最小徑的比例(L/S ratio)、細胞核面積的變異係數(Coefficient of variation of the nuclear area, NACV)，在不同樣本群上的表現情形。利用Student’s t檢定方法分析所選取的細胞核參數在不同樣本群的表現，分析在統計上有顯著差異的型態參數，再以公式求得診斷為分化良好肝臟細胞癌的標準界定值，進一步分析其敏感度(Sensitivity)、準確度(Accuracy)與特異度(Specificity)。最後，是將選定的標準界定值，拿來與懷疑惡性肝腫瘤細胞樣本群之細胞核型態參數之表現值作一比較，分析利用此標準界定值在此樣本群中對於界定分化良好的肝臟細胞癌的準確度。
結果：
所選取的四個細胞核型態參數，除了細胞核面積的變異係數之外，細胞核面積、細胞核周長、細胞核最大徑與最小徑的比例在良性再生性非典型性變化與分化良好的肝臟細胞癌細胞上的表現在統計結果中有顯著性的差異性。而在之後所選取的三個細胞核型態參數對於界定分化良好肝臟細胞癌的標準界定值中，細胞核面積與細胞核周長也都有很高的特異度(分別為100.00%與100.00%)、準確度(分別為93.33%與93.33%)，與敏感度 (分別為86.66%與86.66%)，細胞核最大徑與最小徑的比例有比較高的敏感度(93.33%)，特異度比較低(30.00%)。將選定的診斷為分化良好肝臟細胞癌的標準界定值與懷疑為惡性肝腫瘤細胞上的細胞核參數數值比較，大部分的數值都是在標準界定值以上的表現，少數是落在低值的表現，在細胞核面積與細胞核最大徑與最小徑的比例皆有有達到94.4%的準確度，而在細胞核周長也有達到88.89%的準確度。
結論：
在細胞學抹片中，細胞的型態變化是很多樣化的，有時候確實會造成診斷上的困難，且細胞學診斷是很主觀的，當細胞型態變化不明顯時，往往會因為不同的判讀者而有不同的診斷結果，利用電腦輔助型態測定法之研究在肝臟細針抽取抹片中，可以比較客觀的去分析所選取細胞核參數的在不同細胞上的表現數據，減低人為判讀之間的差異性。在本次研究結果，所選取的細胞核參數除了細胞核面積的變異係數之外，其餘三個選定的細胞核參數在良性與惡性肝腫瘤細胞上的表現有顯著性的差異，且使用所選取的診斷為分化良好肝臟細胞癌的標準界定值與在診斷懷疑為惡性肝腫瘤細胞的樣本都可以得到很高的準確度，因此電腦輔助型態測定法在肝臟細針抽取抹片中，是可以用來作為一個很好的輔助診斷的工具。

關鍵字

分化良好的肝臟細胞癌細胞 ； 良性再生性非典型性變化細胞 ； 懷疑惡性肝腫瘤的細胞 ； 電腦輔助型態測定研究 ； 標準界定值 ； 肝臟細針抽取細胞學抹片

並列摘要

Hepatocellular carcinoma (HCC) is the most common primary hepatic malignancy. Especially, in differential diagnosis of well-differentiated hepatocellular carcinoma(WD-HCC) from benign hepatocytes with reactive or regenerative atypia is always difficult to make in liver fine needle aspiration (FNA ) cytology because of the greater variation in cellular morphology may be found in cytological materials than in histological sections. Based on the equivocal of cell morphology features presented in FNA cytology, the diagnosis may vary between the different cytopathologists, so a standard cutoff value for diagnosis in malignant tumor is essential. Herein, we used this objective computer-assisted image analysis system to evaluate the expression of chose four nuclear parameters in benign hepatocytes with regenerative atypia, WD-HCC and suspicious for malignant cells in liver FNA cytology smears. Further more, we want to define a standard cutoff value of the diagnosis in WD-HCC and hope to reduce the variation between the different cytopathologists and to elevate the diagnosis accuracy in suspicious cells in liver FNA cytology.
Materials and Methods
All 78 cases, including 30 diagnosed as negative were combine with histologically confirmed regenerative atypia; 30 diagnosed as positive and 18 diagnosed as suspicious were combine with histologically confirmed Grade I (well differentiated) HCC. Liver FNA cytology smears were examined regard to nuclear area, nuclear perimeter, largest to smallest diameter ratio of the nuclei (L/S ratio), and coefficient of variation of the nuclear area (NACV) using the computerized morphometric study. Correlations between the nuclear parameters and the different hepatocytes of normal and malignant were investigated. Further we chose the standard cutoff value of the diagnosis in WD-HCC and analyzed the sensitivity, specificity and accuracy. Finally, the cutoff values of the diagnosis in WD-HCC were compared with the parameters in the suspicious group and the diagnostic accuracy was analyzed.
Results
Except of the NACV, all other three parameters were significantly correlated with the benign with regenerative atypia and WD-HCC studied. Our result indicates the high sensitivity and specificity when using the cutoff value of the nuclear area and perimeter, and the high sensitivity and low specificity when using the cutoff value of the L/S ratio. On the other hand, our results show that most of the suspicious group were in the high value and with the highly accuracy when using the chosen cutoff values. It is possible to increase the accuracy in the diagnosis of suspicious cells to 94.44% of nuclear area and LS ratio and 88.89% of nuclear perimeter.
Conclusion
The variety of cytomorphological features in liver FNA cytology smear may lead to difficulties in interpretation. Base on the equivocal of cell morphology features presented in FNA cytology, the diagnosis may vary between the different cytopathologists, Using the objective image analysis technique may reduce the diagnostic variation between different cytopathologists, especially for the suspicious cases. In the presented study, all parameters were significantly correlated with the benign with regenerative atypia and WD-HCC groups except that of the NACV. On the other hand, our results show the highly accuracy in diagnosis for the suspicious cells. In conclusion, computerized nuclear morphometry is a useful diagnostic tool for liver FNA cytology.

並列關鍵字

well differentiated hepatocellular carcinoma ； regenerative atypia ； suspicious cell ； computerized morphometry ； cutoff value ； liver FNA cytology.

參考文獻

1. Hamilton S.R., Aaltonen L.A.(Eds): World Health Organization Classification of Tumors. Pathology and Genetics of Tumours of the Digestive System. IARC Press: Lyon 2000:157-199.

2. Orell SR, Sterrett GF, Whitaker D. Fine Needle Aspiration Cytology.

4th ed.Churchill-Livingstone, 2005:293-316.

3. Hertz G, Reddy VB, Green L, Spitz D, Massarani-Wafai R, Selvaggi SM, Kluskens L, Gattuso P. Fine-needle aspiration biopsy of the liver: a multicenter study of 602 radiologically guided FNA. Diagn Cytopathol. 2000 Nov;23(5):326-8.

4. Franca AV, Valerio HM, Trevisan M, Escanhoela C, Seva-Pereira T,