Implementing Handwritten Text Recognition to Automate Clinical Trials

Clinical trials are the evaluation of medical interventions, including medicinal; surgical, and behavioral interventions. Such studies help researchers to determine if the medical devices and medicines being observed are efficient and whether their exploitation has any side effects.

Clinical trials together with drug descriptions are complex, delicate processes, which require high accuracy. Partially automating these might not only save resources but improve overall productivity and facilitate scientific research.

The data directly relevant to conducted clinical trials must be:

• Up-to-date

• Reliable

• Verifiable

• Traceable

  
  

Manual processing of data (overviews, summaries, study reports) is significantly slowed down and inefficient. To ensure the data is up-to-date, reliable, and easily traceable, research institutions should introduce smart tools for simplified data processing, for example, solutions utilizing HTR (handwritten text recognition).

AI-driven handwritten text recognition for optimized data management

Although more and more healthcare facilities are digitizing patient records and other important documents, most historical medical records remain handwritten, which complicates various processes such as clinical trials. Processing information in form of notes and questionnaires requires significant resource allocation and carries great risks associated with information inaccuracy.

Using optical character recognition might provide valuable benefits such as business automation.

ML-based handwritten text recognition might optimize:

• Information gathering and structuring
• Statistical processing
• Up-to-date monitoring
• Data exchange between sponsor and investigator

But how does this work exactly?

Implementation and business benefits

Data management and operability

Clinical trials consist of four phases:

• Phase 1 – Human pharmacology (safety and right dosage)
• Phase 2 – Therapeutic exploratory (efficacy and side effects)
• Phase 3 – Therapeutic confirmatory (efficacy and adverse reactions)
• Phase 4 – FDA approval (evaluation of long-term risks and benefits)

At each research stage, there might be challenges associated with data management:

• Large amounts of paper-based source protocols
• Information inaccuracy
• Poor accessibility
• Complicated monitoring

Implementing optical character recognition (OCR) might solve these challenges and optimize document operability. The gathered information can be easily digitalized and accessed at any desired moment, as well as exchanged through an internal portal.

The recognized raw data might include:

• Protocols
• Calendars
• Questionnaires
• Photographs
• Patient diaries
• Participant interviews
• Phone logs
• Hospital records
• Letters from referring physicians
• Medication logs
• Original radiological films
• Pharmacy dispensing records
• Data accompanying source documents
• Clinic and office charts:

  • Nurses notes

  • Lab reports

  • Pathology reports

  • Surgical reports

  • Radiology reports

  • Referring physician’s progress notes

    
    

A solution utilizing OCR might help extract and process data from standard, manually filled medical forms.

An example of a manually filled medical form can be seen below:

Manually filled screening form

To optimize document management, a case report form can be previously designed, printed, and then filled in. This process might provide for more accurate reports and significant time savings.

Some examples of pre-designed medical forms can be seen below:

Pre-designed case report form

Pre-designed case report form. Example 2

Pre-designed case report form. Example 3

Insight-driven decision-making at each trial stage

Besides improved data management and operability as well as flexibility, there are other benefits to mention. Implementing optical character recognition might provide an accurate statistical overview and enhance thought-out decision-making.

The solution might facilitate each stage of the clinical trial:

• The design of sheets and figures
• The incorporation of appendixes, which include:
  • Protocols
  • Investigator related information
  • Patient listings, including demographics and baseline patient data
  • Sample Case Report Forms
  • Additional information
  • Related publications
  • Technical statistical documentation
  • Technical statistical details

The reports containing tests and calculations from more than one case study can be done way more accurately. To achieve high accuracy, gathered information has to be placed in a homogeneous database that contains historical data.

Properly designed sheets and figures with included additional appendixes might provide insightful oversight. This way, research institutions might increase overall productivity and save valuable resources.

Electronic clinical study report

Summary report (“Summary of exposure to study drug”)

Rounding up

Although most research institutions are going more digital, clinical trials are usually documented manually. Paper-based routines have many significant downsides, which include erroneous data being stored in the general database and affecting the study.

Implementing OCR might be game-changing in terms of improved time and cost efficiency, as well as flexibility. A solution with optical character recognition might improve several processes, from accurate data extraction and analysis to reporting.

Other benefits may include:

• Data interpretability and operability
• Risk management
• Simplified monitoring
• Compliance with good practices

Introducing handwritten text recognition to digitize paper-based documents used in clinical trials can save valuable resources, eliminate risks, improve security and improve the overall productivity of the conducted study. Another thing, using HTR can optimize data exchange between sponsor and investigator.