Analysis method and analyzer
Disclosed is an analysis method for a specimen using an analyzer connected to a host computer, the analysis method including: obtaining, with respect to each of a plurality of cells contained in the specimen, feature data of the cell; generating classification information in which each of the cells is classified into a plurality of cell types, by analyzing the feature data with use of an artificial intelligence algorithm and performing classifying; generating a measurement result of the specimen on the basis of the classification information; displaying, on a display part of the analyzer, the measurement result and at least a part of the classification information; and transmitting, to the host computer, output data that includes the measurement result and in which at least a part of the classification information has been removed.
1. An analysis method for a specimen using an analyzer connected to a host computer, the analysis method comprising:
preparing a measurement sample by mixing the specimen and a reagent in a chamber of the analyzer;
optically interrogating the measurement sample with an optical detector of the analyzer to obtain feature data of individual cells contained in the measurement sample;
analyzing the feature data with a deep learning algorithm trained to output a set of values of probabilities in response to an input of the feature data of the individual cell, wherein each of the values corresponds to one cell type of multiple cell types and the value represents a probability that the individual cell belongs to the corresponding cell type;
generating classification information for the individual cells, wherein the classification information comprises at least following information:
(i) a primary cell type to which the individual cell belongs with the highest probability,
(ii) a secondary cell type to which the individual cell belongs with the second highest probability, and
(iii) the values of probabilities for the primary and secondary cell types;
generating a measurement result of the specimen on the basis of the primary cell type in the classification information;
displaying, on a display part of the analyzer, the measurement result and at least a part of the classification information;
validating the measurement result; and
transmitting, upon the validation, output data to the host computer, wherein the output data includes the measurement result and the classification information from which the information (ii) or the information (iii) is removed.
2. The analysis method of claim 1 , wherein
the multiple cell types include abnormal cell, and
the classification information in the output data includes a probability that the individual cell belongs to the abnormal cell.
3. The analysis method of claim 2 , wherein
the abnormal cell is a cell other than a monocyte, a lymphocyte, a neutrophil, an eosinophil, and a basophil.
4. The analysis method of claim 3 , wherein
the abnormal cell includes an abnormal lymphocyte, a blast, and an immature granulocyte.
5. The analysis method of claim 1 , wherein
the measurement result includes a number of cells for each of the multiple cell types, wherein the number of cells is obtained by counting cells on the basis of the primary cell type.
6. The analysis method of claim 1 , wherein
the deep learning algorithm has been trained using, as teaching data, feature data regarding morphology of cells and information regarding types of the cells.
7. The analysis method of claim 1 , wherein
the analyzing of the feature data is performed by using a processor and a parallel-processing processor that operates under an order of the processor.
8. The analysis method of claim 1 , wherein
the feature data is a matrix data digitally representing the morphological feature of the optically interrogated cells.
9. The analysis method of claim 1 , wherein
the optical detector includes:
(i) a flow cell through which the cells in the measurement sample flows;
(ii) at least one light source configured to irradiate the cells flowing through the flow cell for optical interrogation; and
(iii) at least one light detector each configured to sense light from a respective one of optically interrogated cells, the sensed light carrying an analog waveform signal indicative of a morphological feature of the respective one of the optically interrogated cells.
10. The analysis method of claim 9 , wherein
the feature data is a matrix data digitally representing the morphological feature of the optically interrogated cells, the matrix data being obtainable by digitally converting the analog waveform signal at a plurality of time points.
11. The analysis method of claim 9 , wherein the analyzer is further configured to sample the analog waveform signal at a predetermined sampling rate to convert the analog waveform signal to obtain the matrix data.
12. An analyzer communicably connected to a host computer, the analyzer comprising:
a preparator configured to prepare a measurement sample by mixing a specimen and a reagent in a chamber;
an optical detector configured to optically interrogate the measurement sample to obtain feature data of individual cells contained in the measurement sample;
at least one processor programmed to analyze the feature data to generate a measurement result; and
a display configured to display the measurement result ,
wherein the processor is programmed to:
analyze the feature data with a deep learning algorithm trained to output a set of values of probabilities in response to an input of the feature data of the individual cell, wherein each of the values corresponds to one cell type of multiple cell types and the value represents a probability that the individual cell belongs to the corresponding cell type;
generate classification information for the individual cell, wherein the classification information includes at least the following information:
(i) a primary cell type to which the individual cell belongs with the highest probability;
(ii) a secondary cell type to which the individual cell belongs with the second highest probability; and
(iii) the values of probabilities for the primary and secondary cell types;
generate a measurement result of the specimen on the basis of the primary cell type in the classification information;
cause the display to display the measurement result and at least a part of the classification information,
validate the measurement result; and
transmit, upon receiving the validation, output data to the host computer, wherein the output data includes the measurement result and the classification information from which the information (ii) or the information (iii) is removed.
13. The analyzer of claim 12 , wherein
the multiple cell types include abnormal cell, and
the classification information in the output data includes a probability that the individual cell belongs to the abnormal cell.
14. The analyzer of claim 13 , wherein
the abnormal cell is a cell other than a monocyte, a lymphocyte, a neutrophil, an eosinophil, and a basophil.
15. The analyzer of claim 14 , wherein
the abnormal cell includes an abnormal lymphocyte, a blast, and an immature granulocyte.
16. The analyzer of claim 12 , wherein the measurement result includes a number of cells for each of the multiple cell types. wherein the number of cells is obtained by counting cells on the basis of the primary cell type.
17. The analyzer of claim 12 , wherein
the deep learning algorithm has been trained using, as teaching data, data regarding morphology of cells and information regarding types of the cells.
18. The analyzer of claim 12 , wherein
the at least one processor includes a processor and a parallel-processing processor that operates under an order of the processor.
19. The analyzer of claim 12 , wherein
the feature data is a matrix data digitally representing the morphological feature of the optically interrogated cells.
20. The analyzer of claim 12 , wherein
the optical detector includes:
(i) a flow cell through which the cells in the measurement sample flows;
(ii) at least one light source configured to irradiate the cells flowing through the flow cell for optical interrogation; and
(iii) at least one light detector each configured to sense light from a respective one of optically interrogated cells, the sensed light carrying an analog waveform signal indicative of a morphological feature of the respective one of the optically interrogated cells.
21. The analyzer of claim 20 , wherein
the feature data is a matrix data digitally representing the morphological feature of the optically interrogated cells, the matrix data being obtainable by digitally converting the analog waveform signal at a plurality of time points.
22. The analyzer of claim 20 , wherein the analyzer is further configured to sample the analog waveform signal at a predetermined sampling rate to convert the analog waveform signal to obtain the matrix data.