Μελέτη με MRI μετακτινικών αλλοιώσεων στα οστά ασθενών με μεταστατικούς ή πρωτοπαθείς όγκους που υποβάλλονται σε ακτινοθεραπεία

Bone marrow can be affected by lymphoproliferative disorders and metastatic disease but also by several therapeutic approaches. MRI is the most suitable method for the detection of metastases and post-treatment follow-up. Image analysis techniques are now used to extract additional diagnostic information. This study focuses on the early radiation-induced changes that can be detected by MRI and compares the established methods for the identification and characterization of these lesions with an automated classification system. METHODS: 36 patients with histologically confirmed primary malignancy and associated bone metastases were included in the study. All patients underwent radiation therapy (RT) to treat bone metastases to the spinal column or the pelvis. Magnetic resonance imaging (MRI) was performed just before the start of RT, 12 to 18 days and up to 3 months after the start of RT. Images were obtained within, adjacent and outside the radiation field. Qualitative assessment was performed independently by two experienced radiologists. For quantitative assessment, specific measurements were selected and evaluated by the method of the region of interest (ROI). In addition, textural features of 1st and 2nd class were exported and inserted into a probabilistic neural network classifier, in order to create an automatic classification system for these lesions. RESULTS: Following qualitative and quantitative assessment, within the radiation field, 22.22% and 33.33% of patients respectively showed fatty conversion of the bone marrow, 19.44% and 16.67% of patients showed haemorrhage, while 11.11% and 16.67% of the patients demonstrated bone marrow oedema. Adjacent to the radiation field, 11.11% and 19.44% of patients showed fatty conversion, 8.33% showed haemorrhage, while 2.78% and 8.33% demonstrated bone marrow oedema. Outside of the radiation field, 5.56% of patients showed changes compatible with fatty conversion, while the remaining 94.44% showed no significant change. There was no statistically significant change of the enhancement index after gadolinium administration. In multivariate analysis, none of the studied parameters did not appear to affect significantly the appearance of any of the radiation-induced lesions. The largest overall classification accuracy of the system designed to distinguish between the pre- radiation and radiation-induced images was 93.02% using the LSFT-PNN classification system of multiple sequences and the ECV method. Discrimination accuracy of the classification system designed to distinguish between the three main types of post-radiation lesions was 86.67%. CONCLUSIONS: This study shows that a significant proportion of patients undergoing RT will experience at least one of the common radiation-induced bone marrow changes. Fatty marrow conversion is the most often featured change in the examined period. Qualitative analysis of the MRI images lacks sensitivity comparing to quantitative measurements. The proposed classification system, based on the neural network, can be used as a very helpful tool for the characterization of these lesions.