A multi-disease exhaled breath screening method and system based on heterogeneous sensor fusion and interpretable risk output

Abstract

A multi-disease exhaled breath screening method and system based on heterogeneous sensor fusion and interpretable risk output belong to the technical field of exhaled breath detection data processing and disease screening. The method receives multi-dimensional feature vectors, sample quality information, environment and sensor state information and historical detection records, obtains original risk output in each disease direction through common base and special disease branch, and performs quality correction and result degradation to form interpretable screening results containing disease direction, risk level and retest suggestion; when the historical detection record meets the comparison condition, the trend comparison result is output. It is suitable for multi-disease exhaled breath screening and risk prompt.

Description

Technical Field

[0001] This invention relates to the field of exhaled breath detection data processing and disease screening technology, and in particular to a multi-disease exhaled breath screening method and system based on heterogeneous sensor fusion and interpretable risk output. Background Technology

[0002] Multi-sensor exhaled breath detection devices typically acquire responses from multiple main sensors, environmental compensation data, and process auxiliary data in a single test. Outputting only a single probability value or threshold judgment can easily lead to poor interpretability, lack of sample quality involvement in decision-making, failure to compensate for environmental and sensor conditions, and insufficient scalability across multiple diseases. Existing technologies often only perform binary classification for a single disease or directly provide model results, failing to simultaneously generate disease direction, risk level, retest recommendations, and trend comparison results. Furthermore, they cannot perform result degradation when dealing with marginal or invalid samples, significant environmental biases, or aging critical sensors. Therefore, they struggle to meet the application requirements for multi-disease joint screening, risk assessment, and auxiliary judgment. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-disease exhaled breath screening method and system based on heterogeneous sensor fusion and interpretable risk output, so as to solve the problems of insufficient interpretability of multi-disease exhaled breath screening results, ineffective participation of sample quality and environmental conditions in result correction, and imperfect trend analysis and retest prompting mechanisms in the prior art.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a multi-disease exhaled breath screening method, comprising receiving multi-dimensional feature vectors, sample quality information, environmental and sensor status information, and historical detection records obtained from a multi-sensor exhaled breath detection process; performing general fusion processing on the multi-dimensional feature vectors based on a shared base, and inputting them into the corresponding disease-specific branches to obtain the original risk output for each disease direction; performing quality correction and result downgrading on the original risk output based on the sample quality information, environmental and sensor status information; mapping the quality-corrected output to an interpretable screening result containing disease direction, risk level, and retesting recommendations; and outputting trend comparison results when the historical detection records meet the comparison conditions.

[0005] In one implementation, the sample quality information includes one or more of high-quality samples, marginally usable samples, invalid samples, and abnormal samples; when the sample is a high-quality sample, a standard screening result is output; when the sample is a marginally usable sample, the reliability of the result is reduced and a retest suggestion is output; when the sample is an invalid or abnormal sample, no formal disease risk conclusion is output, only a sample abnormality prompt is output.

[0006] In one embodiment, the environmental and sensor status information includes one or more of the following: dynamic baseline status, temperature, humidity, drift correction status, and sensor lifetime status; the quality correction and result degradation include weighting, downweighting, or pausing the output of the original risk output for each disease direction based on environmental deviations and sensor health status.

[0007] In one implementation, the risk level includes one or more of the following: low risk, medium risk, high risk, recommended retesting, and insufficient current sample for judgment; the risk level is generated based on the original risk output, sample quality level, environmental and sensor state correction results, and preset stratification rules.

[0008] In one implementation, the trend comparison result is obtained by comparing one or more of the following: risk score, risk level, feature vector change, and sample quality consistency between the current detection result and the historical detection result. The trend comparison result includes one or more of the following: increased risk, basically stable, decreased risk, and large fluctuations with a recommendation for retesting.

[0009] This invention also provides a multi-disease exhaled breath screening system, including an input receiving module, a disease fusion module, a quality correction module, an interpretation mapping module, a trend analysis module, and a result display module. The disease fusion module includes a common base and disease-specific branches corresponding to different disease areas. The common base is used to perform one or more of the following: sample quality evaluation, environmental compensation state integration, general feature fusion, and reliability management. The disease-specific branches are used to call feature subsets, weight parameters, or judgment logic corresponding to the disease area.

[0010] Compared with the prior art, the present invention has at least the following beneficial effects: by using a unified architecture of shared base and disease-specific branches, it enables parallel screening output for multiple diseases, which facilitates subsequent expansion; by incorporating sample quality, environmental bias, and sensor health into the results layer, the results are made more robust; and by combining the output of risk level, retest recommendations, and trend comparison results, the interpretability and consistency of the screening results are improved. Attached Figure Description

[0011] Figure 1 This is a flowchart of the exhaled breath screening method for multiple diseases according to the present invention.

[0012] Figure 2 This is a diagram of the common base and specialized branch architecture of the present invention.

[0013] Figure 3 This is a schematic diagram of the risk stratification mapping of the present invention.

[0014] Figure 4 This is a schematic diagram illustrating the role of quality status in the output decision-making process of this invention.

[0015] Figure 5 This is a schematic diagram comparing trends in this invention.

[0016] Figure 6 This is a schematic diagram of the structured output of the present invention. Detailed Implementation

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] In this embodiment, Figure 1 The system shown includes an input receiving module 1, a disease fusion module 2, a quality correction module 4, an interpretation mapping module 5, a trend analysis module 6, and a result display module 7. The input receiving module 1 receives multidimensional feature vectors 8, sample quality information 9, environmental and sensor status information 10, and historical detection records 11. The disease fusion module 2 performs general fusion on the input information based on a shared base 12, and sends the fused results to a metabolic-related branch 13, a respiratory-related branch 14, a digestive or liver disease-related branch 15, and an inflammation or infection-related branch 16, respectively, to obtain the original risk output 17 for the corresponding disease.

[0019] like Figure 2 As shown, the common base 12 performs at least sample quality assessment, environmental compensation status integration, general feature fusion, and credibility management. Disease-specific branches 13 to 16 respectively call the feature subsets, weight parameters, or judgment logic corresponding to different disease directions to generate the original risk output 17 for each disease direction.

[0020] like Figure 3 As shown, after the original risk output 17 enters the interpretation mapping module 5, it generates results such as low risk 181, medium risk 182, high risk 183, and insufficient current sample for judgment 184 according to the hierarchical mapping rule 18. For cases near the hierarchical threshold or where the quality confidence decreases, a suggested retest result can also be output.

[0021] like Figure 4 As shown, sample quality status participates in the final output decision. High-quality sample path 21 corresponds to standard result output 24, marginal sample path 22 corresponds to downgraded result output 25, and invalid sample path 23 corresponds to outputting only sample anomaly prompts 26. By incorporating sample quality information 9 into the result layer, overly strong conclusions can be avoided when there are insufficient samples, abnormal air blowing, or significant environmental interference.

[0022] like Figure 5 As shown, the trend analysis module 6 receives the current detection result 24 and historical detection results 25, and outputs the following: risk increased 261, basically stable 262, risk decreased 263, or significant fluctuations suggesting retesting 264. Trend analysis can be based on one or more of the following: changes in risk score, changes in risk level, changes in feature vector magnitude, and consistency of sample quality.

[0023] like Figure 6 As shown, the results display module 7 provides a structured presentation of disease type, risk level, sample quality description, retesting recommendations, and trend comparison results, so that users can understand the current screening conclusions and their changing trends.

[0024] This invention is not limited to the embodiments described above. Any equivalent substitutions or modifications made by those skilled in the art to the number of disease directions, feature fusion methods, quality correction rules, risk level classification methods, and trend comparison logic without departing from the spirit and essence of this invention should fall within the protection scope of this invention.

Claims

1. A multi-disease exhaled breath screening method based on heterogeneous sensor fusion and interpretable risk output, characterized in that, The process includes the following steps: S1, receiving multidimensional feature vectors, sample quality information, environmental and sensor status information, and historical detection records obtained from the exhaled breath detection process; S2, performing general fusion processing on the multidimensional feature vectors based on a shared base, and inputting them into the corresponding disease-specific branches to obtain the original risk output for each disease-specific direction; S3. Based on the sample quality information and the environmental and sensor status information, perform quality correction and result downgrading on the original risk output; S4. Map the quality-corrected output to an interpretable screening result, wherein the interpretable screening result includes at least the disease type, risk level, and retesting recommendation; S5. When the historical detection record meets the comparison conditions, compare the current detection result with the historical detection result to determine the trend, and output the trend comparison result.

2. The method according to claim 1, characterized in that, The sample quality information includes one or more of the following: high-quality samples, marginally usable samples, invalid samples, and abnormal samples. When a sample is a high-quality sample, a standard screening result is output. When a sample is a marginally usable sample, the reliability of the result is reduced and a retest suggestion is output. When a sample is an invalid or abnormal sample, no formal disease risk conclusion is output, only a sample abnormality prompt is output.

3. The method according to claim 1, characterized in that, The environmental and sensor status information includes one or more of the following: dynamic baseline status, temperature, humidity, drift correction status, and sensor lifespan status; step S3 includes weighting, downweighting, or pausing the output of the original risk output for each disease direction based on environmental deviations and sensor health status.

4. The method according to claim 1, characterized in that, The risk level includes one or more of the following: low risk, medium risk, high risk, recommended retesting, and insufficient current sample for judgment; the risk level is generated based on the original risk output, sample quality level, environmental and sensor state correction results, and preset stratification rules.

5. The method according to claim 1, characterized in that, The trend comparison results are obtained by comparing the current test results with historical test results in one or more of the following: risk score, risk level, feature vector change, and sample quality consistency. The trend comparison results include one or more of the following: risk increase, basically stable, risk decrease, and large fluctuations with recommendation for retesting.

6. A multi-disease exhaled breath screening system based on heterogeneous sensor fusion and interpretable risk output, characterized in that, include: The input receiving module is used to receive multidimensional feature vectors, sample quality information, environmental and sensor status information, and historical detection records; the disease fusion module is used to perform general fusion processing on the multidimensional feature vectors based on the common base, and input the disease-specific branches of the corresponding disease direction to obtain the original risk output of each disease direction. The quality correction module is used to perform quality correction and result downgrading on the original risk output based on the sample quality information and the environmental and sensor status information; The interpretation mapping module is used to map the quality-corrected output to interpretable screening results; The trend analysis module is used to output trend comparison results when historical detection records meet the comparison conditions; The results display module is used to display the interpretable screening results and trend comparison results.

7. The system according to claim 6, characterized in that, The disease fusion module includes a common base and disease-specific branches corresponding to different disease directions. The common base is used to perform one or more of the following: sample quality evaluation, environmental compensation state integration, general feature fusion, and credibility management. The disease-specific branches are used to call feature subsets, weight parameters, or judgment logic corresponding to the disease direction.

8. The system according to claim 6, characterized in that, The results display module outputs one or more of the following: disease type, risk level, sample quality description, retesting recommendations, and trend comparison results, presented in a structured manner.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the method according to any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 5.

Images