MUMBAI, India, April 10 -- Intellectual Property India has published a patent application (202411070898 A) filed by All India Institute Of Medical Sciences; and Indian Institute Of Technology Delhi, New Delhi, on Sept. 19, 2024, for 'd-master: mask annealed transformer for unsupervised domain adaptation in breast cancer detection from mammograms.'

Inventor(s) include Dr. Krithika Rangarajan; Dr. Chetan Arora; Tajamul Ashraf; Richa Gabha; and Mohit Gambhir.

The application for the patent was published on April 10, under issue no. 15/2026.

According to the abstract released by the Intellectual Property India: "The present disclosure relates to field of Artificial Intelligence (AI) and medical imaging that discloses a method and apparatus for training a student model for domain-adaptive object detection of a subject. The apparatus processes a source dataset and a target dataset using a transformer-based student-teacher framework comprising a student model and a teacher model. Further, the apparatus generates pseudo-labels for the target dataset using the teacher model. Thereafter the apparatus filters the pseudo-labels using an adaptive confidence refinement process configured to progressively increase a confidence threshold. Furthermore, the apparatus updates parameters of the teacher model based on an exponential moving average of parameters of the student model. Finally, the apparatus trains the student model for domainadaptive object detection of a subject, using a combination of supervised loss from the source dataset, unsupervised loss from pseudo-labels of the target dataset, and adversarial loss."

Disclaimer: Curated by HT Syndication.