Automatic verification method for clinical test source data

Abstract

The invention discloses an automatic verification method for clinical test source data. The method comprises steps that an obtained source data image of a clinical test is identified through utilizinga CTPN network model, the text area is determined, and the text area is cut to obtain lines of text; vertical projection column cutting for each line of the text is performed to obtain the valid textarea of each line of the text; a set of valid text areas is sequentially inputted into the trained CRNN network to obtain the variable length sequence identification result, and a regular expressionis utilized to extract the text identification result; the text identification result is corrected to obtain the error correction result; feature values are extracted from the error correction resultone by one according to a feature value set, the feature values are compared with standard feature values recorded in a database, the extracted feature values inconsistent with the standard feature values are marked in an alarm state, and an error reminder is formed. The method is advantaged in that CPTN and CRNN are used as the core to carry out clinical test source data image text identification, and then automatic data verification is achieved.

Description

technical field [0001] The invention relates to the technical field of data verification, in particular to an automatic verification method for clinical trial source data. Background technique [0002] Clinical trial electronic data capture system (EDC, Electric Data Capture system) is suitable for the core information system of drug clinical trials, medical randomized controlled trials and medical cohort studies. Its core purpose is to record the information of subjects and form electronic follow-up form. For clinical trials, the most important thing is data accuracy, and human error in the input process is one of the main reasons for inaccurate data. To this end, the clinical trial implementation team needs to assign a dedicated data auditor to verify the source data: manually verify the original paper materials (test sheets, case records, ICU monitoring records / reports, etc.) The consistency between records and source data is called Source Data Validation (SDV, Source D...

AI Technical Summary

Problems solved by technology

[0003] In the current EDC system, since the source data verification is usually performed manually, there are two problems: first, it is time-consuming and labor-intensive, and auditors need to maintain a high degree of concentration to verify key data one by one. Enter the data once; secondly, the quality cannot be guaranteed. Considering factors such as labor cost and time period, the test execution team cannot conduct a comprehensive source data verification, and usually uses spot checks, which cannot fully guarantee the data quality

Images