Device for performing multi-detection analysis on sample using deep learning-based decoding of encoded magnetic particles, and method therefor

1 . A multi-detection analysis method of a biological sample, comprising:

a step of preparing a plurality of rod-shaped magnetic encoded microparticles encoded in different types, each encoding type corresponding to a different target material and each magnetic encoded microparticle being formed by aligning superparamagnetic nanoparticles in a photocurable polymer matrix in a predetermined direction under an external magnetic field and curing the polymer, wherein each magnetic encoded microparticle has a capture antibody specific to the corresponding target material attached thereto, and wherein the encoding types include a structure in which a plurality of nanoparticle blocks, each having particles aligned in a different direction, are arranged sequentially along the length of the microparticle;

a step of introducing the plurality of magnetic encoded microparticles into a sample of an analysis target and, subsequent to introducing the microparticles, introducing into the sample a detection antibody that binds to the target material and a fluorescent material that attaches to the detection antibody, thereby allowing the detection antibody to bind to the target material captured on the magnetic encoded microparticles and label the microparticles with the fluorescent material;

a step of acquiring a video of the sample in a state where the sample is exposed to a rotating magnetic field;

a step of inputting time-series images from the video into a pre-trained deep learning algorithm, the deep learning algorithm including a convolutional long-short term memory network for analyzing image and time variation information, and analyzing rotation pattern characteristics of each magnetic encoded microparticle in the video over time through the deep learning algorithm to classify each magnetic encoded microparticle; and

a step of, for each classified magnetic encoded microparticle, identifying an encoding type of the microparticle based on the classification result, determining an intensity of the fluorescent material on the microparticle, and quantifying a concentration of the target material corresponding to the encoding type, thereby detecting a plurality of different target materials in the sample.

2 . The multi-detection analysis method of claim 1 , wherein

the rotation pattern characteristics of the magnetic encoded microparticles include one or more of, a rotation speed of the magnetic encoded microparticle, a rotation cycle of the magnetic encoded microparticle; whether the magnetic encoded microparticle undergoes continuous rotation or an intermittent stop during rotation; a frequency of occurrence of an intermittent stop during rotation; and a duration of an intermittent stop.

3 . A multi-detection analysis device for a biological sample, the multi-detection analysis device comprising:

one or more units being configured and executed by a processor using algorithm, the algorithm which when executed, causing the processor to perform the one or more units, the one or more units comprising:

a video acquisition unit configured to acquire an image of a sample of an analysis target while a rotating magnetic field is applied to the sample, the sample having been prepared by introducing a plurality of rod-shaped magnetic encoded microparticles encoded in different types into the sample each encoding type corresponding to a different target material and each magnetic encoded microparticle having a capture antibody specific to the corresponding target material attached thereto, and by introducing, prior to application of the rotating magnetic field, a detection antibody that binds to the target material and a fluorescent material attached to the detection antibody into the sample to react with the target material captured on the magnetic encoded microparticles such that the target material is labeled with the fluorescent material, wherein the encoding types of the magnetic encoded microparticles include a structure in which a plurality of nanoparticle blocks, each having particles aligned in different directions, are arranged sequentially along the length of the microparticle;

a decoding unit configured to input time-series images from the video into a pre-trained deep learning algorithm including a convolutional long-short term memory network, analyze rotation pattern characteristics of each magnetic encoded microparticle in the video over time via the deep learning algorithm, and classify each magnetic encoded microparticle; and

a detection unit configured to, based on the classification result for each magnetic encoded microparticle, identify an encoding type of each magnetic encoded microparticle, determine an intensity of the fluorescent material attached to the detection antibody on the magnetic encoded microparticle, and quantify a concentration of the target material corresponding to the encoding type, thereby multi-detecting a plurality of types of target materials in the sample.

4 . The multi-detection analysis device of claim 3 , wherein

the rotation pattern characteristics of the magnetic encoded microparticles include one or more of: a rotation speed of the magnetic encoded microparticle; a rotation cycle of the magnetic encoded microparticle; whether there is continuous rotation of the magnetic encoded microparticle or an intermittent stop during rotation, a frequency of occurrence of an intermittent stop during rotation; and a time length of an intermittent stop during rotation.