Determining a tissue type of a tissue of an animal or human individual

Inventors

Engel, Thomas • Gigler, Alexander Michael • Otte, Clemens • Pastusiak, Remigiusz • Paust, Tobias • Simon, Elfriede • Stütz, Evamaria • Vogl, Stefanie

Abstract

The invention relates to a method for determining a tissue type of a tissue of an animal or human individual, in which method: electromagnetic radiation (26) emitted by a tissue sample (24) of the tissue is sensed (10) by means of a radiation sensor (22), the radiation sensor (22) providing a sensor signal (28) in accordance with the sensed electromagnetic radiation, and the sensor signal (28) is evaluated (12) by means of an evaluation unit (30) in order to determine and output the tissue type. The problem addressed by the invention is that of enabling improved determination of the tissue type. According to the invention, the evaluation unit (30) is a self-learning evaluation unit (30) that is initially trained (14) by means of training data sets (32) on the basis of at least one model, which is based on a method for machine learning, the training of the evaluation unit being conducted by means of such training data sets (32) each comprising a training sensor signal with an associated training tissue type.

Core Innovation

The invention relates to a method for determining a tissue type of a tissue of an animal or human individual by sensing electromagnetic radiation emitted by a tissue sample using a radiation sensor, and evaluating the sensor signal with an evaluation unit to determine and output the tissue type. The evaluation unit is a self-learning evaluation unit initially trained using training data sets based on at least one machine learning model. Each training data set comprises a training sensor signal with an associated training tissue type.

The problem being solved is the difficulty in reliably distinguishing between normal or healthy tissue and abnormal tissue (such as tumor tissue) during surgical procedures, where visual impressions and surgeon expertise are often insufficient for defining tumor boundaries or the presence of abnormal tissue. Conventional methods, including sending removed tissue for time-consuming histological examination during surgery, are slow and may increase risks for the patient. Moreover, spectral differences in radiation emitted by different tissues are very small, with individual and measurement variations making direct association unreliable.

The invention improves tissue type determination by applying a self-learning analysis unit trained with extensive training data, including data from various individuals and tissue types, to analyze spectral sensor signals from tissue samples. This approach allows for more reliable, prompt, and sensitive identification of tissue types during surgery, potentially replacing slow traditional histological examinations and assisting the surgeon intraoperatively. The method can use spectroscopy (e.g., infrared or ultraviolet) and may involve stimulating tissues or marker agents to emit electromagnetic radiation, analyzed by the self-learning unit to determine tissue conditions such as inflammatory, necrotic, benign, or malignant.

Claims Coverage

The patent includes one independent claim that covers a method for determining a tissue type of tissue employing a self-learning analysis unit trained with datasets based on at least one machine learning model with specific features related to training, analysis, and retraining.

The claims define a method employing a self-learning, machine learning-based analysis unit trained initially with broad datasets and additionally with datasets from the same individual before tissue analysis, with post-analysis retraining conditioned on reliability targets, improving tissue type determination and aiding surgical procedures.

Stated Advantages

Documented Applications