METHOD FOR DIGITIZING A BIOLOGICAL PREPARATION OF BONE MARROW USING NEURAL NETWORKS Russian patent published in 2023 - IPC A61B10/02 G01N33/48 G06T7/00 G06N3/02 

Abstract RU 2793054 C1

FIELD: medicine.

SUBSTANCE: methods for digitizing biological preparations using automated microscopic systems and neural networks, used to detect and analyze formed elements in a bone marrow preparation. A method for digitizing a biological bone marrow preparation is proposed, which includes obtaining a smear image of the preparation at a low magnification, introducing the resulting image into the first neural network to obtain bone marrow smear masks, determining the location of areas of interest, transferring areas of interest to the second neural network, and determining the degree of suitability of areas of interest for preparation analysis. Digitization of the preparation is carried out in two stages, at the first stage, the bone marrow preparation, prepared by the slapping method, is placed on the stage of the microscopic system and digitized to obtain digital images at a low magnification of 10x, followed by cross-linking into one image, then the resulting image is entered into the first neural network, which is generative-adversarial with the ability to determine the coordinates of a smear on a slide, build a smear mask and a mask of the central smear spot, determine the coordinates of suitable areas of interest and store them in the computer's memory. Next, the saved areas of interest are introduced into the second neural network to determine the degree of suitability of the visual fields for analysis, while the pretrained second neural network is a deep convolutional neural network based on VGG16 with the ability to sort the found regions of interest according to the degree of suitability for further digitization at high magnification. If the first neural network does not determine the boundaries of the central spot, the mask of the central spot is formed from the smear mask by erosion, the coordinates of suitable areas of interest are determined and stored in the computer memory. Then, the saved areas of interest obtained at a magnification of 10x are divided into fields of view for obtaining images at a magnification of 100x, the selected fields of view are digitized and stored in the computer's memory, while the fields of view are digitized by moving the stage of microscope with the preparation in such a way that the field of view of the camera of the microscopic system was in the upper left corner of a certain area of interest, then an image of this field of view is obtained and transferred to an adjacent field of view, moving from left to right in an odd row and from right to left in an even row until all fields of view covering the area of interest are digitized. The degree of suitability is determined depending on the uniformity of the distribution of cells in the image and the presence or absence of artifacts, such as voids, paint bubbles, scratches.

EFFECT: ensuring high quality of the result of digitization of bone marrow preparations obtained by the slapping method.

2 cl, 8 dwg

Similar patents RU2793054C1

Title Year Author Number
METHOD FOR SELECTING IMAGES AT HIGH MAGNIFICATION USING NEURAL NETWORKS IN THE STUDY OF A BONE MARROW SMEAR AND A SYSTEM FOR IMPLEMENTING THE METHOD 2022
  • Berezovskij Stanislav Vladimirovich
RU2793060C1
METHOD FOR SCANNING BIOLOGICAL SUBSTANCES WITH THREE TYPES OF LOCALIZATION 2023
  • Berezovsky Stanislav Vladimirivich
RU2804992C1
METHOD FOR DETERMINING DEFECTS IN THE IMAGE OF COMPOSITE PRODUCTS 2022
  • Kremnev Ivan Aleksandrovich
  • Vorobev Ruslan Igorevich
RU2807288C1
METHOD FOR IDENTIFYING AND CALCULATING CELLS IN HUMAN AND ANIMAL BIOLOGICAL MEDIA AND DEVICE FOR ITS IMPLEMENTATION 2004
  • Kozlov Al'Bert Anatol'Evich
  • Golikov Ehduard Vjacheslavovich
  • Sidorov Mikhail Aleksandrovich
  • Ibragimova Irina Tagirovna
RU2303812C2
AUTOFOCUS METHOD FOR DIGITIZING A MICROSCOPIC PREPARATION 2022
  • Berezovskij Stanislav Vladimirovich
RU2794050C1
METHOD FOR AUTOMATIC CLASSIFICATION OF X-RAY IMAGES USING TRANSPARENCY MASKS 2019
  • Dabagov Anatolij Rudolfovich
  • Filist Sergej Alekseevich
  • Kondrashov Dmitrij Sergeevich
RU2716914C1
METHOD FOR SCANNING A CERVICAL PREPARATION PREPARED BY LIQUID CYTOLOGY METHOD 2022
  • Berezovskij Stanislav Vladimirovich
RU2785199C1
METHOD FOR DETERMINING FAULT POSITION BASED ON SEISMIC DATA 2022
  • Miftakhov Ruslan Fanisovich
  • Efremov Igor Ivanovich
  • Bazanov Andrej Konstantinovich
RU2783367C1
METHOD OF DIAGNOSING ONCOLOGICAL DISEASES OF THE BLOOD 2022
  • Nikitaev Valentin Grigorevich
  • Pronichev Aleksandr Nikolaevich
  • Nagornov Oleg Viktorovich
  • Tupitsyn Nikolaj Nikolaevich
  • Selchuk Vladimir Yurevich
  • Dmitrieva Valentina Viktorovna
  • Palladina Aleksandra Dmitrievna
  • Polyakov Evgenij Valerevich
RU2803281C1
METHOD FOR DIGITISING A BONE MARROW PUNCTATE SMEAR 2019
  • Li, Qiang
  • Lu, Ju
  • Li, Shun
  • Liu, Yongtao
  • Hu, Jiajia
RU2755247C1

RU 2 793 054 C1

Authors

Berezovsky Stanislav Vladimirivich Permsky Kray, G. Perm, Ul. Nikolaya Ostrovskogo,

Dates

2023-03-28Published

2022-11-29Filed