OCR for Alternative Medicine Invoices

AITECA extracts the data needed to process and calculate reimbursement, even when the information is scattered, non-standardized, or partially handwritten. Typically, the extraction covers:

• Practitioner identification: last name, specialty, RPPS/ADELI number when present on the document

• Patient/insured person information: last name, first name, identifying details useful for matching the claim

• Financial data: billed amounts, totals, consistency of amounts

• Document metadata: document date, invoice number, presence of a signature

The extraction process does not rely on a fixed format or template, but rather on an understanding of the content tailored to the variability of alternative medicine invoices.

Rather than relying on a single reading technology, AITECA is based on the orchestration of several complementary extraction technologies, each deployed according to the actual content of the document. Specifically, the processing combines, among other things:

• Deep learning-based OCR, including capabilities dedicated to handwriting, to read invoices that combine different writing styles, freehand writing, and varying image qualities.

• NLP (natural language processing), used to understand the meaning of the text and identify key information (dates, amounts, identities), even in the absence of a standardized structure or vocabulary.

• Full-text analysis, to capture all useful information in unstructured documents.

• Healthcare-specific business rules and controls, applied during extraction to secure and standardize the extracted data.

These technologies are dynamically orchestrated: depending on the actual nature of the document, the system activates and combines the most relevant components.

This approach ensures robust and reliable extraction, even when dealing with highly heterogeneous invoices, without relying on a fixed model or a predefined format.

The AITECA IDP does more than just extract data: each piece of information is validated before being processed by reimbursement management tools.

Validation checks focus in particular on:

• The completeness of the document (whether essential fields are present or missing)

• Data format: standardization of amounts, dates, addresses, etc.

• Regulatory and healthcare industry rules: validity and consistency of professional identifiers (RPPS, ADELI, FINESS, as applicable), the document’s date of issue, and the presence of a signature when required.

Each check is traceable and auditable, which secures the review processes and facilitates post-processing audits.

Invoices for non-contracted medical services represent an identified risk (errors, inconsistencies, fraud attempts). The processing therefore incorporates mechanisms designed to detect document-related and contextual anomalies. This includes, in particular:

Detection of anomalies and document tampering:

• Altermetry®: detection of alterations to an original document (e.g., name, date, or amount altered)

• Data validation and consistency checks

Detection of AI-generated fake documents:

• Verification based on C2PA provenance signatures, when present

• Detection using AI models specifically trained on synthetic documents (generative AI)

External verifications:

• Verification of the professional’s existence via official registries (FINESS, RPPS, ADELI, etc.)

• Company checks: legal existence, length of operation, insolvency proceedings, entities under supervision

Each invoice can be assigned a confidence level, allowing sensitive cases to be automatically routed for appropriate processing prior to reimbursement.

• 100% hosted in France by an ISO 27001-certified partner, protected from the Cloud Act (public cloud, SecNumCloud, or on-premises, depending on your needs)
• End-to-end encryption (TLS 1.3 + AES-256) and strict data compartmentalization
• Automatic deletion of documents after processing
• Your data is never used to train our AI models—a contractual commitment
• Data Processing Agreement (DPA) provided as standard, in compliance with Article 28 of the GDPR
• Designated DPO and comprehensive procedures for data subjects’ rights
• ISO 27001:2022 certification across our entire SaaS scope
• SOC 2 Type II and HDS certifications provided by Microsoft (hosting provider)

Once processing is complete, the data from alternative medicine invoices is delivered in a format that can be directly used by the IT system. Specifically:

• Data structured in JSON format, ready to be used by downstream applications

• Control and confidence metrics associated with the extracted data

• Very short processing time, compatible with digital workflows and near real-time processing

This approach enables the standardization of processing without disrupting existing business tools.