How Myriad Genetics achieved quick, correct, and cost-efficient doc processing utilizing the AWS open-source Generative AI Clever Doc Processing Accelerator
This publish was written with Martyna Shallenberg and Brode Mccrady from Myriad Genetics.
Healthcare organizations face challenges in processing and managing excessive volumes of advanced medical documentation whereas sustaining high quality in affected person care. These organizations want options to course of paperwork successfully to satisfy rising calls for. Myriad Genetics, a supplier of genetic testing and precision drugs options serving healthcare suppliers and sufferers worldwide, addresses this problem.
Myriad’s Income Engineering Division processes 1000’s of healthcare paperwork day by day throughout Ladies’s Well being, Oncology, and Psychological Well being divisions. The corporate classifies incoming paperwork into lessons comparable to Take a look at Request Types, Lab Outcomes, Medical Notes, and Insurance coverage to automate Prior Authorization workflows. The system routes these paperwork to acceptable exterior distributors for processing based mostly on their recognized doc class. They manually carry out Key Data Extraction (KIE) together with insurance coverage particulars, affected person info, and take a look at outcomes to find out Medicare eligibility and assist downstream processes.
As doc volumes elevated, Myriad confronted challenges with its current system. The automated doc classification resolution labored however was expensive and time-consuming. Data extraction remained guide resulting from complexity. To handle excessive prices and gradual processing, Myriad wanted a greater resolution.
This publish explores how Myriad Genetics partnered with the AWS Generative AI Innovation Heart (GenAIIC) to rework their healthcare doc processing pipeline utilizing Amazon Bedrock and Amazon Nova basis fashions. We element the challenges with their current resolution, and the way generative AI lowered prices and improved processing pace.
We study the technical implementation utilizing AWS’s open source GenAI Intelligent Document Processing (GenAI IDP) Accelerator solution, the optimization methods used for doc classification and key info extraction, and the measurable enterprise influence on Myriad’s prior authorization workflows. We cowl how we used immediate engineering strategies, mannequin choice methods, and architectural selections to construct a scalable resolution that processes advanced medical paperwork with excessive accuracy whereas decreasing operational prices.
Doc processing bottlenecks limiting healthcare operations
Myriad Genetics’ day by day operations depend upon effectively processing advanced medical paperwork containing essential info for affected person care workflows and regulatory compliance. Their current resolution mixed Amazon Textract for Optical Character Recognition (OCR) with Amazon Comprehend for doc classification.
Regardless of 94% classification accuracy, this resolution had operational challenges:
- Operational prices: 3 cents per web page leading to $15,000 month-to-month bills per enterprise unit
- Classification latency: 8.5 minutes per doc, delaying downstream prior authorization workflows
Data extraction was fully guide, requiring contextual understanding to distinguish essential scientific distinctions (like “is metastatic” versus “is just not metastatic”) and to find info like insurance coverage numbers and affected person info throughout various doc codecs. This processing burden was substantial, with Ladies’s Well being customer support requiring as much as 10 full-time workers contributing 78 hours day by day within the Ladies’s Well being enterprise unit alone.
Myriad wanted an answer to:
- Cut back doc classification prices whereas sustaining or bettering accuracy
- Speed up doc processing to get rid of workflow bottlenecks
- Automate info extraction for medical paperwork
- Scale throughout a number of enterprise models and doc sorts
Amazon Bedrock and generative AI
Fashionable massive language fashions (LLMs) course of advanced healthcare paperwork with excessive accuracy resulting from pre-training on large textual content corpora. This pre-training allows LLMs to know language patterns and doc constructions with out characteristic engineering or massive labeled datasets. Amazon Bedrock is a totally managed service that provides a broad vary of high-performing LLMs from main AI corporations. It supplies the safety, privateness, and accountable AI capabilities that healthcare organizations require when processing delicate medical info. For this resolution, we used Amazon’s latest basis fashions:
- Amazon Nova Professional: A cheap, low-latency mannequin supreme for doc classification
- Amazon Nova Premier: A complicated mannequin with reasoning capabilities for info extraction
Answer overview
We carried out an answer with Myriad utilizing AWS’s open supply GenAI IDP Accelerator. The accelerator supplies a scalable, serverless structure that converts unstructured paperwork into structured information. The accelerator processes a number of paperwork in parallel via configurable concurrency limits with out overwhelming downstream companies. Its built-in evaluation framework lets customers present anticipated output via the person interface (UI) and consider generated outcomes to iteratively customise configuration and enhance accuracy.
The accelerator affords 1-click deployment with a selection of pre-built patterns optimized for various workloads with completely different configurability, price, and accuracy necessities:
- Pattern 1 – Makes use of Amazon Bedrock Information Automation, a totally managed service that provides wealthy out-of-the-box options, ease of use, and easy per-page pricing. This sample is advisable for many use instances.
- Pattern 2 – Makes use of Amazon Textract and Amazon Bedrock with Amazon Nova, Anthropic’s Claude, or customized fine-tuned Amazon Nova fashions. This sample is good for advanced paperwork requiring customized logic.
- Pattern 3 – Makes use of Amazon Textract, Amazon SageMaker with a fine-tuned mannequin for classification, and Amazon Bedrock for extraction. This sample is good for paperwork requiring specialised classification.
Sample 2 proved best suited for this venture, assembly the essential requirement of low price whereas providing flexibility to optimize accuracy via immediate engineering and LLM choice. This sample affords a no-code configuration – customise doc sorts, extraction fields, and processing logic via configuration, editable within the internet UI.
We custom-made the definitions of doc lessons, key attributes and their definitions per doc class, LLM selection, LLM hyperparameters, and classification and extraction LLM prompts through Sample 2’s config file. In manufacturing, Myriad built-in this resolution into their current event-driven structure. The next diagram illustrates the manufacturing pipeline:
- Doc Ingestion: Incoming order occasions set off doc retrieval from supply doc administration programs, with cache optimization for beforehand processed paperwork.
- Concurrency Administration: DynamoDB tracked concurrent AWS Step Function jobs whereas Amazon Simple Queue Service (SQS) queues recordsdata exceeding concurrency limits for doc processing.
- Textual content Extraction: Amazon Textract extracted textual content, format info, tables and kinds from the normalized paperwork.
- Classification: The configured LLM analyzed the extracted content material based mostly on the custom-made doc classification immediate supplied within the config file and classifies paperwork into acceptable classes.
- Key Data Extraction: The configured LLM extracted medical info utilizing extraction immediate supplied within the config file.
- Structured Output: The pipeline formatted the leads to a structured method and delivered to Myriad’s Authorization System through RESTful operations.
Doc classification with generative AI
Whereas Myriad’s current resolution achieved 94% accuracy, misclassifications occurred resulting from structural similarities, overlapping content material, and shared formatting patterns throughout doc sorts. This semantic ambiguity made it tough to tell apart between comparable paperwork. We guided Myriad on immediate optimization strategies that used LLM’s contextual understanding capabilities. This strategy moved past sample matching to allow semantic evaluation of doc context and objective, figuring out distinguishing options that human specialists acknowledge however earlier automated programs missed.
AI-driven immediate engineering for doc classification
We developed class definitions with distinguishing traits between comparable doc sorts. To determine these differentiators, we supplied doc samples from every class to Anthropic Claude Sonnet 3.7 on Amazon Bedrock with mannequin reasoning enabled (a characteristic that permits the mannequin to reveal its step-by-step evaluation course of). The mannequin recognized distinguishing options between comparable doc lessons, which Myriad’s material specialists refined and included into the GenAI IDP Accelerator’s Pattern 2 config file for doc classification prompts.
Format-based classification methods
We used doc construction and formatting as key differentiators to tell apart between comparable doc sorts that shared comparable content material however differed in construction. We enabled the classification fashions to acknowledge format-specific traits comparable to format constructions, subject preparations, and visible parts, permitting the system to distinguish between paperwork that textual content material alone can’t distinguish. For instance, lab stories and take a look at outcomes each comprise affected person info and medical information, however lab stories show numerical values in tabular format whereas take a look at outcomes observe a story format. We instructed the LLM: “Lab stories comprise numerical outcomes organized in tables with reference ranges and models. Take a look at outcomes current findings in paragraph format with scientific interpretations.”
Implementing detrimental prompting for enhanced accuracy
We carried out detrimental prompting strategies to resolve confusion between comparable paperwork by explicitly instructing the mannequin what classifications to keep away from. This strategy added exclusionary language to classification prompts, specifying traits that shouldn’t be related to every doc sort. Initially, the system steadily misclassified Take a look at Request Types (TRFs) as Take a look at Outcomes resulting from confusion between affected person medical historical past and lab measurements. Including a detrimental immediate like “These kinds comprise affected person medical historical past. DO NOT confuse them with take a look at outcomes which comprise present/current lab measurements” to the TRF definition improved the classification accuracy by 4%. By offering express steering on frequent misclassification patterns, the system averted typical errors and confusion between comparable doc sorts.
Mannequin choice for price and efficiency optimization
Mannequin choice drives optimum cost-performance at scale, so we performed complete benchmarking utilizing the GenAI IDP Accelerator’s evaluation framework. We examined 4 basis fashions—Amazon Nova Lite, Amazon Nova Professional, Amazon Nova Premier, and Anthropic Claude Sonnet 3.7—utilizing 1,200 healthcare paperwork throughout three doc lessons: Take a look at Request Types, Lab Outcomes, and Insurance coverage. We assessed every mannequin utilizing three essential metrics: classification accuracy, processing latency, and price per doc. The accelerator’s price monitoring enabled direct comparability of operational bills throughout completely different mannequin configurations, guaranteeing efficiency enhancements translate into measurable enterprise worth at scale.
The analysis outcomes demonstrated that Amazon Nova Professional achieved optimum steadiness for Myriad’s use case. We transitioned from Myriad’s Amazon Comprehend implementation to Amazon Nova Professional with optimized prompts for doc classification, attaining important enhancements: classification accuracy elevated from 94% to 98%, processing prices decreased by 77%, and processing pace improved by 80%—decreasing classification time from 8.5 minutes to 1.5 minutes per doc.
Automating Key Data Extraction with generative AI
Myriad’s info extraction was guide, requiring as much as 10 full-time workers contributing 78 hours day by day within the Ladies’s Well being unit alone, which created operational bottlenecks and scalability constraints. Automating healthcare KIE introduced challenges: checkbox fields required distinguishing between marking types (checkmarks, X’s, handwritten marks); paperwork contained ambiguous visible parts like overlapping marks or content material spanning a number of fields; extraction wanted contextual understanding to distinguish scientific distinctions and find info throughout various doc codecs. We labored with Myriad to develop an automatic KIE resolution, implementing the next optimization strategies to deal with extraction complexity.
Enhanced OCR configuration for checkbox recognition
To handle checkbox identification challenges, we enabled Amazon Textract’s specialised TABLES and FORMS options on the GenAI IDP Accelerator portal as proven within the following picture, to enhance OCR discrimination between chosen and unselected checkbox parts. These options enhanced the system’s capacity to detect and interpret marking types present in medical kinds.
We enhanced accuracy by incorporating visible cues into the extraction prompts. We up to date the prompts with directions comparable to “search for seen marks in or across the small sq. packing containers (✓, x, or handwritten marks)” to information the language mannequin in figuring out checkbox alternatives. This mixture of enhanced OCR capabilities and focused prompting improved checkbox extraction in medical kinds.
Visible context studying via few-shot examples
Configuring Textract and bettering prompts alone couldn’t deal with advanced visible parts successfully. We carried out a multimodal strategy that despatched each doc photos and extracted textual content from Textract to the inspiration mannequin, enabling simultaneous evaluation of visible format and textual content material for correct extraction selections. We carried out few-shot learning by offering instance doc photos paired with their anticipated extraction outputs to information the mannequin’s understanding of assorted type layouts and marking types. A number of doc picture examples with their right extraction patterns create prolonged LLM prompts. We leveraged the GenAI IDP Accelerator’s built-in integration with Amazon Bedrock’s prompt caching feature to scale back prices and latency. Immediate caching shops prolonged few-shot examples in reminiscence for five minutes—when processing a number of comparable paperwork inside that timeframe, Bedrock reuses cached examples as a substitute of reprocessing them, decreasing each price and processing time.
Chain of thought reasoning for advanced extraction
Whereas this multimodal strategy improved extraction accuracy, we nonetheless confronted challenges with overlapping and ambiguous tick marks in advanced type layouts. To carry out effectively in ambiguous and sophisticated conditions, we used Amazon Nova Premier and carried out Chain of Thought reasoning to have the mannequin assume via extraction selections step-by-step utilizing considering tags. For instance:
Moreover, we included reasoning explanations within the few-shot examples, demonstrating how we reached conclusions in ambiguous instances. This strategy enabled the mannequin to work via advanced visible proof and contextual clues earlier than making closing determinations, bettering efficiency with ambiguous tick marks.
Testing throughout 32 doc samples with various complexity ranges through the GenAI IDP Accelerator revealed that Amazon Textract with Structure, TABLES, and FORMS options enabled, paired with Amazon Nova Premier’s superior reasoning capabilities and the inclusion of few-shot examples, delivered one of the best outcomes. The answer achieved 90% accuracy (identical as human evaluator baseline accuracy) whereas processing paperwork in roughly 1.3 minutes every.
Outcomes and enterprise influence
By our new resolution, we delivered measurable enhancements that met the enterprise objectives established on the venture outset:
Doc classification efficiency:
- We elevated accuracy from 94% to 98% via immediate optimization strategies for Amazon Nova Professional, together with AI-driven immediate engineering, document-format based mostly classification methods, and detrimental prompting.
- We lowered classification prices by 77% (from 3.1 to 0.7 cents per web page) by migrating from Amazon Comprehend to Amazon Nova Professional with optimized prompts.
- We lowered classification time by 80% (from 8.5 to 1.5 minutes per doc) by selecting Amazon Nova Professional to offer a low-latency and cost-effective resolution.
New automated Key Data Extraction efficiency:
- We achieved 90% extraction accuracy (identical because the baseline guide course of): Delivered via a mixture of Amazon Textract’s doc evaluation capabilities, visible context studying via few-shot examples and Amazon Nova Premier’s reasoning for advanced information interpretation.
- We achieved processing prices of 9 cents per web page and processing time of 1.3 minutes per doc in comparison with guide baseline requiring as much as 10 full-time workers working 78 hours day by day per enterprise unit.
Enterprise influence and rollout
Myriad has deliberate a phased rollout starting with doc classification. They plan to launch our new classification resolution within the Ladies’s Well being enterprise unit, adopted by Oncology and Psychological Well being divisions. On account of our work, Myriad will understand as much as $132K in annual financial savings of their doc classification prices. The answer reduces every prior authorization submission time by 2 minutes—specialists now full orders in 4 minutes as a substitute of six minutes resulting from quicker entry to tagged paperwork. This enchancment saves 300 hours month-to-month throughout 9,000 prior authorizations in Ladies’s Well being alone, equal to 50 hours per prior authorization specialist.
These measurable enhancements have reworked Myriad’s operations, as their engineering management confirms:
“Partnering with the GenAIIC emigrate our Clever Doc Processing resolution from AWS Comprehend to Bedrock has been a transformative step ahead. By bettering each efficiency and accuracy, the answer is projected to ship financial savings of greater than $10,000 per 30 days. The crew’s shut collaboration with Myriad’s inner engineering crew delivered a high-quality, scalable resolution, whereas their deep experience in superior language fashions has elevated our capabilities. This has been a superb instance of how innovation and partnership can drive measurable enterprise influence.”
– Martyna Shallenberg, Senior Director of Software program Engineering, Myriad Genetics
Conclusion
The AWS GenAI IDP Accelerator enabled Myriad’s fast implementation, offering a versatile framework that lowered growth time. Healthcare organizations want tailor-made options—the accelerator delivers intensive customization capabilities that permit customers adapt options to particular doc sorts and workflows with out requiring intensive code modifications or frequent redeployment throughout growth. Our strategy demonstrates the facility of strategic immediate engineering and mannequin choice. We achieved excessive accuracy in a specialised area by specializing in immediate design, together with detrimental prompting and visible cues. We optimized each price and efficiency by choosing Amazon Nova Professional for classification and Nova Premier for advanced extraction—matching the correct mannequin to every particular process.
Discover the answer for your self
Organizations trying to enhance their doc processing workflows can expertise these advantages firsthand. The open source GenAI IDP Accelerator that powered Myriad’s transformation is offered to deploy and take a look at in your setting. The accelerator’s easy setup course of lets customers shortly consider how generative AI can remodel doc processing challenges.
When you’ve explored the accelerator and seen its potential influence in your workflows, attain out to the AWS GenAIIC crew to discover how the GenAI IDP Accelerator will be custom-made and optimized on your particular use case. This hands-on strategy ensures you can also make knowledgeable selections about implementing clever doc processing in your group.
In regards to the authors
Priyashree Roy is a Information Scientist II on the AWS Generative AI Innovation Heart, the place she applies her experience in machine studying and generative AI to develop progressive options for strategic AWS clients. She brings a rigorous scientific strategy to advanced enterprise challenges, knowledgeable by her PhD in experimental particle physics from Florida State College and postdoctoral analysis on the College of Michigan.
Mofijul Islam is an Utilized Scientist II and Tech Lead on the AWS Generative AI Innovation Heart, the place he helps clients sort out customer-centric analysis and enterprise challenges utilizing generative AI, massive language fashions (LLM), multi-agent studying, code technology, and multimodal studying. He holds a PhD in machine studying from the College of Virginia, the place his work targeted on multimodal machine studying, multilingual pure language processing (NLP), and multitask studying. His analysis has been revealed in top-tier conferences like NeurIPS, Worldwide Convention on Studying Representations (ICLR), Empirical Strategies in Pure Language Processing (EMNLP), Society for Synthetic Intelligence and Statistics (AISTATS), and Affiliation for the Development of Synthetic Intelligence (AAAI), in addition to Institute of Electrical and Electronics Engineers (IEEE) and Affiliation for Computing Equipment (ACM) Transactions.
Nivedha Balakrishnan is a Deep Studying Architect II on the AWS Generative AI Innovation Heart, the place she helps clients design and deploy generative AI functions to unravel advanced enterprise challenges. Her experience spans massive language fashions (LLMs), multimodal studying, and AI-driven automation. She holds a Grasp’s in Utilized Information Science from San Jose State College and a Grasp’s in Biomedical Engineering from Linköping College, Sweden. Her earlier analysis targeted on AI for drug discovery and healthcare functions, bridging life sciences with machine studying.
Martyna Shallenberg is a Senior Director of Software program Engineering at Myriad Genetics, the place she leads cross-functional groups in constructing AI-driven enterprise options that remodel income cycle operations and healthcare supply. With a novel background spanning genomics, molecular diagnostics, and software program engineering, she has scaled progressive platforms starting from Clever Doc Processing (IDP) to modular LIMS options. Martyna can also be the Founder & President of BioHive’s HealthTech Hub, fostering cross-domain collaboration to speed up precision drugs and healthcare innovation.
Brode Mccrady is a Software program Engineering Supervisor at Myriad Genetics, the place he leads initiatives in AI, income programs, and clever doc processing. With over a decade of expertise in enterprise intelligence and strategic analytics, Brode brings deep experience in translating advanced enterprise wants into scalable technical options. He holds a level in Economics, which informs his data-driven strategy to problem-solving and enterprise technique.
Randheer Gehlot is a Principal Buyer Options Supervisor at AWS who makes a speciality of healthcare and life sciences transformation. With a deep give attention to AI/ML functions in healthcare, he helps enterprises design and implement environment friendly cloud options that tackle actual enterprise challenges. His work entails partnering with organizations to modernize their infrastructure, allow innovation, and speed up their cloud adoption journey whereas guaranteeing sensible, sustainable outcomes.
Acknowledgements
We want to thank Bob Strahan, Kurt Mason, Akhil Nooney and Taylor Jensen for his or her important contributions, strategic selections and steering all through.
