Bimodal radiomics image analysis method for pulmonary nodule classification

Abstract

The invention discloses a bimodal radiomics image analysis method for pulmonary nodule classification, which relates to the technical field of pulmonary nodule image analysis, and comprises the following steps: collecting pulmonary nodule patients subjected to PET/CT examination and capable of preprocessing PET/CT images and making clear pathological diagnosis on pulmonary nodules by a center, using a PET/CT machine, and fasting each patient for more than 6 hours before examination, the blood glucose is measured to be smaller than 11.1 mmol/L through fingertip blood sampling, after 18F-FDG is intravenously injected according to the body weight of a patient, the patient has a quiet rest for about 1 hour, the body is scanned from the skull to the upper portion of the thigh, image collection is conducted from the tail of a bed, scanning lasts for 2 minutes for each bed, 9-10 beds are scanned for each patient, 3D-Slicer software is used for segmenting the images. A semi-automatic method, an NVIDIA AI-Assisted Annotation method and a boundary-based CT segmentation model are respectively used for segmenting PET and CT images of the pulmonary nodules. According to the method, information can be provided for treatment decision and prognosis judgment of the pulmonary nodule patient, and personalized treatment is promoted in a non-invasive mode.

Description

technical field [0001] The invention relates to the technical field of pulmonary nodule image analysis, in particular to a dual-mode radiomics image analysis method for pulmonary nodule classification. Background technique [0002] With the wide application of multi-slice spiral CT and low-dose CT in lung cancer screening, the detection rate of pulmonary nodules has been continuously improved, and its diagnosis and differential diagnosis, treatment decision-making and prognosis judgment have become a huge challenge for clinicians. , pathological examination is still the gold standard for lesion diagnosis and prognosis, but it requires high puncture techniques and is difficult, and the heterogeneity of tumors is likely to cause sampling deviations. There are certain limitations in the application, and an ideal non-invasive examination method is urgently needed to classify pulmonary nodules, so as to provide reference for clinical treatment decisions and prognosis judgment of ...

AI Technical Summary

Problems solved by technology

[0002] With the wide application of multi-slice spiral CT and low-dose CT in lung cancer screening, the detection rate of pulmonary nodules has been continuously improved, and its diagnosis and differential diagnosis, treatment decision-making and prognosis judgment have become a huge challenge for clinicians. , pathological examination is still the gold standard for lesion diagnosis and prognosis, but it requires high puncture techniques and is difficult, and the heterogeneity of tumors is likely to cause sampling deviations. There are certain limitations in the application, and an ideal non-invasive examination method is urgently needed to classify pulmonary nodules, so as to provide reference for clinical treatment decisions and prognosis judgment of pulmonary nodules

[0003] At present, radiomics is a promising and very popular diagnostic method. With the help of mathematical and statistical methods, the spatial relationship between voxels can be quantitatively described, and high-throughput imaging data from regions of interest Extract characteristic spatial data and capture potential lesion information beyond the ability of naked eyes to improve the accuracy of disease diagnosis. It has the advantages of being objective, non-invasive, fast, low-cost, and easy to operate. Previous studies have shown that CT texture features in The identification of lung cancer, prediction of tumor growth, gene expression and efficacy evaluation are of great significance, but it can only provide morphological information of pulmonary nodules, and the morphological characteristics of pulmonary nodules overlap to some extent and are easily affected by subjective factors , the identification efficiency needs to be improved, therefore, we urgently need a non-invasive, objective and accurate image analysis method to classify pulmonary nodules

Images