MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096142 A) filed by Sr University Warangal on August 08, 2026, for Machine Learning Self-Supervised Learning For Domain Adaptation In Clinical Nlp Applications.
Inventors include Anitha Busari; and Dr. Dadi Ramesh.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: The present disclosure generally relates to a computer-implemented system and method for performing self-supervised learning based domain adaptation for clinical NLP applications. The disclosed system includes a Clinical Data Acquisition Module that acquires structured and unstructured clinical data from Electronic Health Records (EHRs), physician notes, laboratory reports, radiology reports, pathology reports and biomedical knowledge sources. A Clinical Text Preprocessing Module is used to preprocess clinical text to produce standardized clinical text through the following processes: Data normalization, De-identification, Tokenization, Semantic mapping, and Contextual feature extraction. A Deep Clinical Representation Learning Engine is trained using a self-supervised learning objective in the context of unlabeled clinical corpora to learn generalized semantic representations, and a Transformer Based Contextual Embedding Module is used to produce deep contextual embeddings for clinical language understanding. The Domain Adaptation Engine tackles distributional differences between clinical domains of different types by performing adversarial feature alignment, domain-invariant representation learning and adaptable knowledge transfer. A Clinical Knowledge Integration Module enriches the representations in the enriched representations, and a Multi-Task Clinical NLP Prediction Module processes the enriched representations to carry out biomedical knowledge processing tasks such as biomedical named entity recognition, disease classification, diagnosis prediction, medical relation extraction, clinical document summarization, clinical coding and decision support. An Explainable Artificial Intelligence Engine produces interpretable prediction report along with the feature importance and confidence scores, and a Model Evaluation and Optimization Module validates the predictive performance. The generated recommendations are then presented to healthcare professionals via a Clinical Decision Support Interface and the framework is continually refined by incorporating newly gathered clinical data and clinician feedback via a Continuous Learning Module. The disclosed invention enables cross-domain generalization, mitigates the need for manual annotation of clinical datasets, increases prediction accuracy and offers an intelligent, scalable and explainable framework for deployment in a diverse healthcare setting.
Disclaimer: Curated by HT Syndication.
