Optimized Fuzzy C-Means Algorithm-Based Coronal Magnetic Resonance Imaging Scanning in Tracheal Foreign Bodies of Children

Abstract

In order to provide theoretical support for clinical diagnosis, the diagnostic value of the optimized fuzzy C-means (FCM) algorithm combined with coronal magnetic resonance imaging (MRI) scan was investigated in the diagnosis of tracheal foreign bodies in children. The anisotropic filtering was applied to optimize the traditional FCM algorithm, so as to construct a new MRI image segmentation algorithm, namely, AFFCM algorithm. Then, the traditional FCM algorithm, the FCM algorithm based on the kernel function (KFCM), and the FCM algorithm based on the spatial neighborhood information (RFCM) were introduced for comparison with the AFFCM. 28 children diagnosed with foreign bodies in the trachea were selected for MRI diagnosis, and AFFCM was used for segmentation. The partition coefficient, segmentation entropy, and the correlation degree between classes after fuzzy division of the four algorithms were recorded, and the location and distribution of foreign bodies in the trachea and the types of foreign bodies were also collected. Besides, the MRI scanning and chest X-rays of the children with foreign bodies in the trachea should also be recorded in terms of the positive rate, diagnosis rate, and indirect signs. The class division coefficient and interclass correlation degree after fuzzy division of AFFCM were markedly greater than those of FCM, KFCM, and RFCM (P < 0.05), while the segmentation entropy of AFFCM was less sharp than the entropies of FCM, KFCM, and RFCM (P < 0.05). Among the 28 children, there were 5 cases with foreign bodies in the trachea (17.86%), 10 cases in the left bronchus (35.71%), and 13 cases in the right bronchus (46.43%). Among the foreign body types, there were 10 cases of melon seeds (35.71%), 6 cases of peanuts (21.43%), and 5 cases of beans (17.86%). The positive rate (89.29%) and diagnosis rate (96.43%) of MRI for bronchial foreign bodies increased obviously in contrast to the rates of X-ray chest radiographs (57.14% and 67.86%) (P < 0.05). Therefore, it was indicated that AFFCM showed higher partition coefficient value, lower segmentation entropy, larger similarity among classes, and better image segmentation effect. Furthermore, AFFCM-based coronal MRI scan had a higher positive rate and diagnosis rate for children's tracheal foreign bodies, and the main signs were emphysema and atelectasis.

Figure 1

Algorithm's flow chart.

Figure 2

Comparison on partition coefficients, segmentation entropy, and the correlation degree between classes after fuzziness of different algorithms: (a) the partition coefficient; (b) the segmentation entropy; (c) the correlation degree between classes after fuzziness. ∗ indicates P < 0.05 compared with AFFCM; # indicates P < 0.05 compared with FCM.

Figure 3

Comparison on the segmentation performance of different algorithms: (a) the original image; (b) the image processed by FCM; (c) the image processed by KFCM; (d) the image processed by RFCM; and (e) the image processed by AFFCM constructed in this study.

Figure 4

X-ray and MRI image evaluation of a male case (2 years old): (a) a chest X-ray; (b) an MRI image.

Figure 5

The distribution of foreign bodies in the bronchus of 28 child patients: 1: the foreign bodies in the trachea; 2: the foreign bodies in the left bronchus; 3: the foreign bodies in the right bronchus.

Figure 6

Types of bronchial foreign bodies in 28 child patients.

Figure 7

Comparison on positive rates of MRI and X-ray diagnosis of bronchial foreign bodies: ∗ indicates P < 0.05 by comparison with MRI.

Figure 8

Comparison of the diagnosis rates of MRI and X-ray diagnosis of tracheal foreign bodies: ∗ indicates P < 0.05 compared to MRI.

Figure 9

Comparison of the tracheal foreign body signs of MRI and X-ray diagnosis: ∗ indicates P < 0.05 compared with X-ray.