Intelligent surgical marker

Inventors

Liu, Xuan • Liu, Yuwei

Assignees

New Jersey Institute of Technology

Abstract

Surgical marker systems and methods for delineating a lesion margin of a subject are provided. An example system includes a handheld probe device configured to capture an optical coherence tomography (OCT) image and a processor coupled to a memory. The handheld probe device includes a handheld probe including a fiber-optic probe assembly and a marker assembly. The processor is configured to: segment, by a neural network, each pixels of the OCT into different tissue-type categories; generate one or more feature vectors based at least in part on the segmented pixels; determine, by a one-class classifier, a boundary location in the OCT image between a normal tissue and an abnormal tissue of the tissue structure; and control the marker assembly to selectively create a visible label on a tissue location of the subject, the tissue location corresponding to the boundary location.

Core Innovation

The invention provides a surgical marker system and methods for delineating a lesion margin of a subject, including a handheld probe device configured to capture an optical coherence tomography (OCT) image and a processor coupled to a memory. The handheld probe device comprises a fiber-optic probe assembly to direct and collect light for OCT imaging, and a marker assembly to create a visible label on the lesion margin. The processor uses a neural network to segment each pixel of the OCT image into different tissue-type categories, generates feature vectors from these segmented pixels, applies a one-class classifier to determine a boundary location between normal and abnormal tissue, and controls the marker assembly to create a visible label at the boundary location.

This invention addresses the clinical need for quantitative, objective, and data-driven delineation of lesion margins to guide surgeons in performing more accurate tissue excision, particularly for procedures such as Mohs micrographic surgery (MMS). Conventional margin determination relies on qualitative and subjective visual assessment, which depends on the surgeon's experience and can result in less accurate outcomes. The system provides contemporaneous and concurrent margin detection and marking, overcoming the limitations of conventional imaging modalities that lack sufficient spatial resolution to identify microscopic pathological features.

The system integrates high-quality OCT imaging with a lightweight, manually and/or automatically scanned probe capable of imaging uneven tissue surfaces. Artificial intelligence algorithms, including a deep convolutional neural network (such as U-Net) for tissue segmentation and a one-class classifier (such as SVM) for detecting anomalies, are employed for automatic tissue classification and tumor margin detection. The system further provides a mechanism for registering tumor margins back to the patient by directly marking the identified boundary on the tissue surface, facilitating precise surgical excision based on objective data.

Claims Coverage

The patent contains two independent claims that define the system and method for delineating a lesion margin using OCT imaging, AI-driven tissue classification, and a marking mechanism.

The inventive features cover a comprehensive surgical marker system and corresponding method that integrate OCT imaging, neural network-based tissue segmentation, feature vector analysis, one-class classification for anomaly detection, and in situ boundary marking, all aimed at accurate lesion margin delineation.

Stated Advantages

Documented Applications