Morpheus removes patient identifiers from greek clinical notes locally
Morpheus: a fully local, defense-in-depth de-identification engine for Greek clinical text - a benchmark evaluation
Computation and Language
Summary
Medical notes written in Greek need to have all personal patient information removed before leaving the hospital to protect privacy. The authors tested a system called Morpheus that automatically removes these identifiers from Greek clinical texts without sending data outside the hospital. They created 500 fake hospital reports with personal details and showed that Morpheus removed over 98% of direct patient information while keeping almost all medical content intact. The system also caught most remaining leaks and was consistent across repeated runs. This is the first evaluation of such a tool for Greek medical text.
What this means in practice
- •For hospital data teams: Protect patient privacy by locally removing personal identifiers from Greek medical documents before sharing or analysis.
- •For healthcare it administrators: Run a fully local, deterministic system to ensure consistent de-identification and detect privacy leaks in Greek clinical notes.
Authors
Filippos-Paschalis Rorris, Nikoleta Stanitsa, Tatiana Rorri, Foteini Rorri, Constantine N. Antonopoulos
Abstract
Background and Objective: Clinical free text cannot leave the hospital without de-identification, yet no de-identification system or benchmark exists for Greek clinical text. We evaluated Morpheus, a fully local de-identification pipeline for Greek clinical documents, on 500 synthetic discharge summaries. Methods: The notes (300 cardiac surgery, 100 cardiology, 100 nephrology) were generated by a frontier large language model (Gemini 3.1 Pro) constrained to authentic Greek hospital document layouts. Notes were pre-annotated at generation; two clinicians then independently reviewed and corrected every note, blinded to each other, with disagreements arbitrated. Primary outcomes were recall over must-redact identifiers and over-redaction of clinical content, measured over must-survive probes. Secondary outcomes: residual leaks flagged by an independent egress-review layer, processing time and reproducibility. Results: Inter-annotator agreement was F1 = 0.998. The de-identification rate was 96.59% (7,666/7,937 identifier instances; 95% CI 96.16-96.96): 98.49% for direct and 93.42% for quasi-identifiers. No patient's own complete identifier survived in any note; of 271 residual instances, 199 (73.4%) concerned institutional identity, chiefly the hospital letterhead. The over-redaction rate was 2.82% (190/6,747 probes; 95% CI 2.45-3.24). The egress guard flagged 2 of 271 residuals, both complete telephone numbers. Two independent runs gave identical verdicts. Conclusions: A fully local, deterministic pipeline removed 98.5% of direct patient identifiers from Greek clinical text while altering under 3% of protected clinical content; most residuals concerned the institution, not the patient. To our knowledge this is the first benchmark evaluation of clinical-text de-identification for Greek and the first quantified evaluation of an independent egress-guard layer.