MRC determined that to increase customer satisfaction, retention and acquisition (as well as affirm their position as a market leader), the workflow of collating medical records had to be significantly increased. As it’s a key business need for solicitors and barristers to have access to the collated medical records as soon as possible.

They also faced a challenge in terms of their ability to scale the business and meet customer demand, when processing a much larger volume of documents. They were able to process around 2,000 pages per business day but were being faced with challenges of staff hire and turnover, due to the repetitive nature of manual document sorting, as well as limitations to the speed and accuracy of record collation for their customers.

To meet the challenges of increasing the speed and capacity of the manual processes of pagination, MRC looked towards automation and refinement within their workflow, for record collation and towards cloud adaption, especially using AI/ML services.

Looking to use AI/ML, meant MRC were looking for a partner that could guide them through technological advancements and the cloud. MRC needed to be provided full understanding, confidence and training, across all stakeholders, to embrace the new innovation.

Working with Firemind, the path to automate and optimise MRC’s processes were defined, then validation of challenges in handwritten medical documents, such as doctor’s notes, were provided.

Furthermore, by working backwards from MRC’s desired outcomes, MRC and Firemind explored together the opportunity to create an Intellectual Property for collating records that would have wide appeal. Not only for their own customers but also as an individual service with licensing ability.

For this specific project, we had to gain a thorough understanding of the many challenges that this build faced:

• Document volume – MRC focus on clinical negligence and personal injury cases. Due to the complexity and variety in these cases, some case documents could span thousands of pages in total, whilst others may consist of 50 or less.
• Classification variety – To detect handwriting from doctor’s notes as well as specialist notes for personal injury consultations. It would have to do this as well as quickly adjusting to typed copy. The quality of each document also becomes a factor when classifying blurry and low-quality scans.
• Value differentiation – To build accurate models, there is a need to understand and differentiate between numbers and characters that shared similar visuals. For example, ‘1’ and ‘7’, or ‘6’ and ‘G’. Failure to differentiate numerical and character differences would result in unstructured page formatting and classification for indexing purposes.
• Data modelling – A complex variety of label types to identify such as hospitals, patient details, medical procedures, and processes.

Due to the complexity of the build and the requirement for validation to the business and its stakeholders, a Proof-of-Concept (PoC) focused on the above-mentioned challenges was to be delivered.

To combat differentiation, the PoC we built had to assign a category/subcategory for each document as well as understand the classification variety. This would help replicate the current system and requirement used by MRC through their manual process.

To begin streamlining MRC’s operation and collation processes, we worked on building a PoC that could fully automate the document processing, sorting, and collation system. Using Amazon Textract as the core service, we could intelligently extract texts from handwritten, scanned, or electronic medical records. Then, by applying Amazon Comprehend capabilities, we could extract insights from the documents such as key information or document category.

We used Comprehend Custom Classifier to manage the data model training and predictive capabilities. A TFIDF (term frequency–inverse document frequency) text classification model was built as a precursor to modelling performance and understanding the data at hand.

These ML (Machine Learning) services were built on a serverless infrastructure using AWS Lambda and Amazon DynamoDB. This provided greater scalability, flexibility, and reduced cost during the data training and concept building. As a result, at the end of this AI/ML driven process, MRC can create a neatly ordered, paginated, and indexed document.

We saw a 1,920% speed increase when comparing the automated solution to the current human led workflow. The use of server-less infrastructure that can scale allows for faster, more secure data movement, whilst minimising MRC staffing costs when performing the same tasks manually. During our PoC, AWS costs dropped by over 80%, showing that through refinement of data modelling, less resources are required to complete the pagination service.

The PoC provided validation for a production ready automated process, which in time will significantly reduce the laborious efforts of manual collation, ensuring MRC staff can work on more engaging and problematic medical cases. This will lead to improved scalability in terms of the document volume MRC can handle, providing a provisional capacity increase of over 500% when data trained models are trained further.

The PoC showed a quick advancement in accuracy and correct pagination during the data training process. After further data training and modelling, the AI/ML process will be able to perform at a higher consistency of accuracy than a human operator.

This PoC addressed both the ability to replace a manual and labour-intensive process, whilst driving the creation of a new IP for MRC. As owners of all the code and models, MRC will raise the benchmark within their industry.