A Multi-Objective Flow Cytometry Profiling for B-Cell Lymphoma Diagnosis
| Autor: | Shou-Hsuan Stephen Huang, Ming-Chih Shih, Youli Zu |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
0301 basic medicine
Engineering drawing medicine.diagnostic_test Computer science business.industry Low Confidence computer.software_genre medicine.disease Automation Minimal residual disease Confidence interval Flow cytometry 03 medical and health sciences 030104 developmental biology Test case medicine Profiling (information science) Data mining B-cell lymphoma business computer |
| Zdroj: | BCB |
| DOI: | 10.1145/2975167.2975169 |
| Popis: | Flow cytometry, a powerful tool for the diagnosis of hematolymphoid malignancies including B-cell lymphomas, is an innovative technique that measures the fluorescence of suspended cells. Although several computerized methods are available for flow cytometry data processing, most current automatic techniques have not been fully developed. In our previous publication, we proposed a multi-profile detection algorithm for several types of B-cell lymphoma that is capable of replacing the current conventional manual gating technique to provide a more efficient and accurate technique through automation and improved data interpretation. The automated diagnosis also comes with a confidence level which provides more information for pathologists; a low confidence will alert pathologists they need to investigate the case more. In this paper, we strengthen the previous algorithm to give a much better performance of the system. We also tested the improved algorithm by analyzing test cases of minimal residual disease (MRD), obtained from a group of patients with fewer B-cell lymphoma cells, which is a condition, when exhibited, that greatly increases the difficulty of automated diagnosis. The validity of our automated system is supported by comparing the results from the automated system diagnosis with those of the conventional manual gating process. The average accuracy rate of the automated diagnosis system is 95.2% for the dataset and lowers to 91.8% when tested with MRD data. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|