federalnoe gosudarstvennoe avtonomnoe obrazovatelnoe uchrezhdenie vysshego obrazovaniya "Sankt-Peterburgskij politekhnicheskij universitet Petra Velikogo" (FGAOU VO "SPbPU")
发明人:
Utkin Lev Vladimirovich (RU),Уткин Лев Владимирович (RU),Ryabinin Mikhail Andreevich (RU),Рябинин Михаил Андреевич (RU),Meldo Anna Aleksandrovna (RU),Мелдо Анна Александровна (RU),Lukashin Aleksej And
申请号：
RU2018118626
公开号：
RU0002668699C1
申请日:
2018.05.21
申请国别(地区):
RU
年份:
2018
代理人:
摘要：
FIELD: medicine.SUBSTANCE: invention relates to medicine, namely to the diagnosis of lung cancer. Method comprises the processing of images of the patient's lungs obtained by computed tomography, as a result of which voxels with density values according to the Hounsfield scale that do not correspond to the density values of the lung tissue are masked in the graphic image； subsequent segmentation of voxels located on the surface of the "candidates" of the neoplasms； construction of a set of chords formed by combinations of pairs of points located in the allocated voxels on the surface of the "candidates" of the neoplasms； the construction of a histogram of the distribution of the length of the chords with the reduction to the maximum length of the chord constructed within the boundaries of each "candidate" of the neoplasm； the formation of a feature vector including the data of the constructed histogram of the distribution of chord lengths, average value of density according to the Hounsfield scale of each "candidate" of the neoplasm, the total number of voxels in each "candidate" of the neoplasm. After that, according to the formed vector of characteristics, each candidate "candidate" of neoplasms is classified as a true malignant neoplasm using the machine learning algorithm that implements the functions of the classifier.EFFECT: invention provides a reduction in the number of detected false positive neoplasms in the lungs.1 cl, 5 dwgИзобретение относится к медицине, а именно к диагностике рака легких. Способ содержит обработку изображений легких пациента, полученных методом компьютерной томографии, в результате которой в графическом изображении маскируют воксели со значениями плотности по шкале Хаунсфилда, не соответствующими значениям плотности тканям легких； последующую сегментацию вокселей, расположенных на поверхности «кандидатов» новообразований； построение множества хорд, образованных комбинациями пар точек, находящихся в выделенных вокселях на поверхности «
