Deep Learning in Medical Image Analysis by Gobert Lee & Hiroshi Fujita

Author:Gobert Lee & Hiroshi Fujita
Language: eng
Format: epub
ISBN: 9783030331283
Publisher: Springer International Publishing

Fig. 11ROC curves of classification

Discussion

As can be seen in Table 2, there was a significant difference between the characteristic features of the benign and malignant nodules with regard to SUV. The SUV obtained by PET examination was confirmed to be useful for distinguishing between benign and malignant. However, inflammatory diseases can also produce high SUVs; hence, SUVs should be combined with other features for correct classification. In addition, some texture features were effective for distinguishing between benign and malignant nodules. The difference in features with regard to spicules was of low significance. This is because many spicules are observed even in inflammatory diseases. Furthermore, the average CTmax in benign nodules was higher than that in malignant nodules. This is due to calcification inside the benign nodules. The distributions of all characteristic features overlapped between the two classes, which makes it difficult to distinguish between the two classes using a single feature. Therefore, integration of multiple characteristic features using a classifier would be more effective.

From the ROC curves (Fig. 11), the proposed method using the CT image and two-phase PET images shows that the incorrect rate of detection for benign nodules was 0.278 (accuracy rate of 0.722 for benign nodules), whereas the accuracy rate of malignant nodule detection was 0.944. The target nodules in this study were “difficult nodules” for differentiation by image diagnosis using CT and PET/CT. Most of the benign cases were not confirmed in follow-up examinations but were confirmed after biopsy. Our results indicate that biopsy examination with its accompanying physical hardship to the patient, especially in benign cases, could be reduced by 72.2%.

In addition, compared to classification using only CT images, the result of using PET images together with CT improved the AUC of the ROC curves, as shown in Fig. 11. This proves the effectiveness of using both anatomical and functional information together. Based on the results of the p-values of the ROC curves, we found a significant difference between analysis using CT alone and analysis using both CT and PET images.

These results indicate that the proposed method may be useful for improving the accuracy of malignancy analysis.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.