Use of multi-frequency impedance cytometry in conjunction with machine learning for classification of biological particles
This disclosure provides methods and systems for classifying biological particles, e.g., blood cells, microbes, circulating tumor cells (CTCs). Using impedance flow cytometry, such as multi-frequency impedance cytometry, in conjunction with supervised machine learning, the disclosed methods and systems demonstrated improved accuracy in classifying biological particles.
1. A system for classifying biological particles, comprising:
a non-transitory, computer-readable memory;
one or more processors; and
a computer-readable medium containing programming instructions that, when executed by the one or more processors, cause the system to:
measure an impedance response of one or more biological particles in a sample at multiple frequencies simultaneously and using a multi-frequency lock-in amplifier to generate impedance response data associated with the one or more biological particles in the sample, wherein the multiple frequencies are from 100 kHz to 10 MHz;
determine physical properties of the generated impedance response data at the multiple frequencies, wherein the determined physical properties of the generated impedance response data comprise electrical properties that comprise amplitude of the impedance response, phase of the impedance response, or both; and
classify the one or more biological particles in the sample into categories based on the determined physical properties of the generated impedance response data at the multiple frequencies by applying a machine learning model trained with measurements of the multi-frequency lock-in amplifier and amplitudes of the impedance response at the multiple frequencies to the generated impedance response data.
2. The system of claim 1 , wherein the step of measuring the impedance response is performed using multi-frequency impedance cytometry.
3. The system of claim 1 , wherein the machine learning model comprises Support Vector Machine, K Nearest Neighbors, Logistic Regression, Random Forests, or Deep Learning.
4. The system of claim 1 , wherein the biological particles comprise cells.
5. The system of claim 4 , wherein the cells comprise white blood cells, red blood cells, or cancer cells.
6. The system of claim 5 , wherein the cancer cells comprise circulating cancer cells (CTCs).
7. The system of claim 1 , wherein the step of determining the physical properties of the generated impedance response data comprises detrending and denoising the generated impedance response data.
8. A system for determining a type of a biological particle, comprising:
a non-transitory, computer-readable memory;
one or more processors; and
a computer-readable medium containing programming instructions that, when executed by the one or more processors, cause the system to:
measure an impedance response of a biological particle in a sample at multiple frequencies simultaneously using multi-frequency impedance cytometry and a multi-frequency amplifier to generate impedance response data associated with the biological particle in the sample, wherein the multiple frequencies are from 100 kHz to 10 MHz;
determine physical properties of the generated impedance response data at the multiple frequencies, wherein the determined physical properties of the generated impedance response data comprise electrical properties that comprise amplitude of the impedance response, phase of the impedance response, or both; and
determine a type of the biological particle in the sample based on the determined physical properties of the generated impedance response data at the multiple frequencies by applying a machine learning model trained with measurements of the multifrequency lock-in amplifier and amplitudes of the impedance response at the multiple frequencies to the generated impedance response data.
9. A method of classifying biological particles, comprising:
measuring an impedance response of one or more biological particles in a sample at multiple frequencies simultaneously and using a multi-frequency lock-in amplifier to generate impedance response data associated with the one or more biological particles in the sample, wherein the multiple frequencies are from 100 kHz to 10 MHz;
determining physical properties of the generated impedance response data at the multiple frequencies, wherein the determined physical properties of the generated impedance response data comprise electrical properties that comprise amplitude of the impedance response, phase of the impedance response, or both; and
classifying the one or more biological particles in the sample into categories based on the determined physical properties of the generated impedance response data at the multiple frequencies by applying a machine learning model trained with measurements of the multi-frequency lock-in amplifier and amplitudes of the impedance response at the multiple frequencies to the generated impedance response data.
10. The method of claim 9 , wherein the step of measuring further comprises measuring the impedance response using multi-frequency impedance cytometry.
11. The method of claim 9 , wherein the machine learning model comprises Support Vector Machine, K Nearest Neighbors, Logistic Regression, Random Forests, or Deep Learning.
12. The method of claim 9 , wherein the biological particles comprise cells.
13. The method of claim 12 , wherein the cells comprise white blood cells, red blood cells, or cancer cells.
14. The method of claim 13 , wherein the cancer cells comprise circulating cancer cells (CTCs).
15. The method of claim 9 , wherein the step of determining the physical properties of the generated impedance response data comprises detrending and denoising the generated impedance response data.