System and method for identifying, measuring, and assessing the therapeutic response of the extent and distribution of tumour burden in the peritoneal cavity

Abstract

The invention relates to a computer-implemented system and method for identifying, measuring, and assessing the distribution and extent of diseases in peritoneal cavities. The method makes it possible to acquire images of the peritoneal cavity (100), which has been pre-marked with markers (10) affixed via a video laparoscopic procedure, wherein, by means of computer processing, peritoneal disease behaviour can be verified at different stages of therapy and compared with a reference pattern from pathology groups stored in a remote database (40), stored in the cloud, such as a data ecosystem.

Description

SYSTEM AND METHOD TO IDENTIFY, MEASURE AND EVALUATE THE THERAPEUTIC RESPONSE OF THE EXTENSION AND DISTRIBUTION OF TUMOR BURDEN IN THE PERITONEAL SPACE FIELD OF INVENTION [1] This invention patent describes a computer-implemented system and method for identifying, measuring, and evaluating the distribution and extent of diseases in peritoneal cavities. The method allows for obtaining images of the peritoneal cavity, previously demarcated by markers fixed by a videolaparoscopic procedure, which, through computer processing, allows for the verification of the behavior of peritoneal disease at different moments of therapy and comparison with a pattern of groups of pathologies stored in a remote database, stored in the cloud, such as a data ecosystem. BACKGROUND OF THE INVENTION [2] Tumors in the peritoneal space have their extension and positioning verified through imaging tests or open or minimally invasive surgical techniques with direct or indirect visualization, such as videolaparoscopy, open surgery, computed tomography or magnetic resonance imaging. [3] Adequate assessment of the behavior of diseases in the peritoneal space, especially neoplasms, has become a fundamental part of the decision-making process regarding the treatment of pathologies. Adequate assessment of the degree of involvement of these spaces by pathologies of different etiologies represents an additional challenge in measuring the extent of the disease in the space under analysis. Furthermore, monitoring the progression or regression of involvement in the space under analysis uses methods. subjective and imprecise, making it difficult to accurately measure both disease progression and response to different therapies, such as systemic or intracavitary chemotherapy. [4] Peritoneal carcinomatosis is a disease associated with the spread of cancer cells in the peritoneal space. It is frequently related to tumors originating from the digestive tract (colon, stomach, pancreas, and appendix), the gynecological tract (ovaries and fallopian tubes), and primary peritoneal tumors. Peritoneal dissemination of metastases is often classified as stage 4 neoplasia. Thus, the natural history of this form of peritoneal dissemination leads to outcomes such as intestinal obstruction, ascites, and death. Any neoplasm that allows the proliferation of neoplastic cells in the peritoneal space will develop this form of neoplastic dissemination. [5] Accurate estimates of the rate of peritoneal involvement worldwide are scarce in the literature. Data obtained so far show that between 15% and 20% of all colon cancers develop carcinomatosis, regardless of the treatment regimen. Furthermore, between 50% and 60% of gastric cancers present peritoneal dissemination at diagnosis, with the peritoneal route being the primary route of dissemination in ovarian cancers. [6] Currently, the only curative-intent treatment for patients with peritoneal carcinomatosis is cytoreductive surgery and hyperthermic intraperitoneal chemotherapy. However, only 10% of all patients with peritoneal carcinomatosis qualify for these procedures. With the incorporation of new technologies and approaches to peritoneal carcinomatosis, both for appropriate patient selection and response monitoring, to the proposed therapies, greater precision in the evaluation of peritoneal metastases has been demanded. [7] The need for precision in measuring the extent, spread, and response of peritoneal metastases has posed increasing challenges to current analog tools, which have limitations. For example, indices such as the Peritoneal Cancer Index (PCI), Fagotti index, and Peritoneal Regression Grading Score (PRGS) are used to analyze peritoneal carcinomatosis. However, these indices are subjective and analog, preventing the precise measurement and calculation of peritoneal metastases in their number, density, volume, and behavior throughout the natural history of neoplasms of different etiologies. [8] Analog tools such as Recist (Response Evaluation Criteria in Solid Tumors) are designed to analyze the response of solid tumors. This tool has helped interpret the role of new treatments, serving as the reference for determining tumor response and the parameter used in phase II and phase III studies. However, Recist has limitations in measuring tumors smaller than 10 mm, cystic lesions, and areas with locoregional treatment (D. Jayne. (2007). Molecular Biology of Peritoneal Carcinomatosis. In: Peritoneal Carcinomatosis: Principles of Management, Springer, ch. 2, pp. 1-10. [Online]. https: / / doi.org / 10.1007 / 978-0-387-48993-3_2), (S. Käbler and J. Jhne. (2007). Staging and Scoring of Peritoneal Carcinomatosis. In: Peritoneal Carcinomatosis: Principles of Management. Springer, ch. 14, pp. 1-10. [Online], https: / / doi.org / 10.1007 / 978-0-387-48993-3_14), (S. Samel and M. Lohr. (2007).Targeted Intraabdominal Chemotherapy for Peritoneal Carcinomatosis. In: Peritoneal Carcinomatosis: Principles of. Management, Springer, ch. 32, pp. 1-10. https: / / doi.org / 10.1007 / 978-0-387-48993-3_32.). All these characteristics determine that peritoneal carcinomatosis is a manifestation considered unmeasurable by Recist, making the tool difficult to apply for the evaluation of peritoneal carcinomatosis treatments. [9] Na tentativa de quantificar a carcinomatose, vários sistemas de pontuação foram desenvolvidos (Fujimoto S, Takahashi M, Mutou T, et al. Improved mortality rate of gastric carcinoma patients with peritoneal carcinomatosis treated with intraperitoneal hyperthermic chemoperfusion combined with surgery. Cancer. 1997; 79: 884- 891 ; Gilly FN, Carry PY, Sayag AC, et al. Regional chemotherapy (with mitomycin C) and intra-operative hyperthermia for digestive cancers with peritoneal carcinomatosis. Hepatogastroenterology. 1994; 41 : 124- 129; Jacquet P, Sugarbaker PH. Clinical research methodologies in diagnosis and staging of patients with peritoneal carcinomatosis. In: PH Sugarbaker, ed. Peritoneal Carcinomatosis: Principles of Management. Boston, MA: Kluwer Academic Publishers; 1996: 359- 374; van der Vange N, van Goethem AR, Zoetmulder FA, et al.Extensive cytoreductive surgery combined with intra-operative intraperitoneal perfusion with cisplatin under hyperthermic conditions (OVHIPEC) in patients with recurrent ovarian cancer: a feasibility pilot. Eur J Surg Oncol. 2000; 26: 663- 668; Goere D, Souadka A, Faron M, et al. Extent of colorectal peritoneal carcinomatosis: attempt to define a threshold above which HIPEC does not offer survival benefit: a comparative study [published online ahead of print January 29, 2015], Ann Surg Oncol; Verwaal VJ, van Tinteren H, van Ruth S, Zoetmulder FA. Predicting the survival of patients with peritoneal carcinomatosis of colorectal origin treated by aggressive cytoreduction and hyperthermic intraperitoneal. chemotherapy. Br J Surg. 2004; 91: 739-746.). Basically, the quantification systems developed involve dividing the abdomen into regions. One of the most widely used scoring systems is the Peritoneal Cancer Index (PCI), which divides the peritoneal cavity and small intestine into 13 different regions. Each region is then given a subjective / scalar score from 0 to 3 based on the amount of tumor present (0 indicates none, while 3 indicates confluent tumor or tumor measuring more than 5 cm).

[0010] Although still somewhat subjective and analog, these systems allow surgical oncologists to better compare outcomes, including oncological outcomes, morbidity, and mortality, based on the extent of tumor burden. PCI is currently the most widely used tool for evaluating carcinomatosis, demonstrating its prognostic value in carcinomatosis of various etiologies. This fact corroborates the need for more precise identification of tumor distribution and volume in the peritoneal space.

[0011] The prior art describes methods for diagnosing peritoneal carcinomatosis. Document UA99353 describes a method for radiographically diagnosing the severity of peritoneal carcinomatosis in colorectal cancer, by measuring the amount of free fluid in the abdominal cavity in the presence of nodular lesions in the peritoneum measuring 0.5-1.5 cm.

[0012] Document KR20210039078 describes a peptide probe for the diagnosis of peritoneal carcinomatosis that specifically binds to a gastric cancer cell line and can be used as an imaging index in clinical and endoscopic analyses to diagnose abdominal metastases of gastric cancer.

[0013] Document RU2625898 describes a method for scoring the probability of peritoneal carcinomatosis in patients with portal cholangiocarcinoma based on criteria such as TM disease stage (l-IV), metastatic lesion of regional lymph nodes, presence of distant metastatic lesions, lesion of the trunk and branches of the portal vein, presence of ascites.

[0014] The state of the art describes subjective methods for describing the physical and volumetric characteristics of peritoneal carcinomatosis. This subjectivity makes it difficult to identify the behavior of the pathologies, such as measuring and calculating the size of the incidence, the volume of the nodules, and their statistical distribution. Obtaining approximate measurements of the size of the metastases and the large gradient (difference in measurements for each of the variables analyzed) makes it difficult to accurately measure the tumor burden present in the peritoneal cavity. This imprecision and the inference in the measurements lead to inaccuracy in the analysis of the nodules' response to treatments.

[0015] Recent medical approaches have taken into account individual genetic variation, environment, and lifestyle. This precision medicine, or personalized medicine, allows doctors to more accurately predict which treatments and disease prevention strategies will be effective for a specific patient. In this recent trend, doctors use diagnostic information to better understand a patient's illness. This may include genetic testing to determine a patient's predisposition to certain diseases or conditions, or the use of targeted treatments, which are designed to attack diseased cells more effectively and with fewer side effects than traditional treatments.

[0016] In the context of information technology and its digital tools, and the rise of artificial intelligence (AI), the association of correlated data can be an important means of improving cancer diagnosis and treatment. This can aid in diagnoses, eliminate or reduce subjectivity, increase mathematical and statistical accuracy over time, and identify patterns across patients affected by these pathologies. The large volume of data and the possibility of analysis using intelligent algorithms allow for expanded analysis of individual characteristics. The evidence observed in the dataset provides a set of information that impacts individual interventions and can modify the natural history of the individual's pathology. This phenomenon is observed in the analysis of medical images to detect early signs of breast cancer (Subasi, A., Kandpal, A.D., Raj, K.A., & Bagci, U. (2023). Breast cancer detection from mammograms using artificial intelligence.Elsevier. DOI:10.1016 / b978-0-443-18450-5.00005-0.), to detect lung nodules (Zheng, S., van Ooijen, P.M.A., & Oudkerk, M. (2022). Lung Cancer Screening and Nodule Detection: The Role of Artificial Intelligence. Springer International Publishing. DOI:10.1007 / 978-3-030-92087-6_43) and to detect ovarian cancer. Thus, Artificial Intelligence, through methods implemented in computer algorithms, is consolidated as a safe digital technology for the diagnosis and monitoring of tumor progression.

[0017] However, a critical step in the detection and diagnosis of carcinomas involves accurately identifying the region and characteristics of metastatic nodules present in the area of ​​interest. Based on current techniques, some patterns related to shape, size, density, extension, and spectral color evaluation, which may allow us to infer the benign or malignant nature, are not apparent at first glance, impacting the treatment and therapeutic outcome.

[0018] Thus, the present invention describes a system that includes markers implanted in the peritoneal cavity, a unit for acquiring and pre-processing images obtained from the body cavity and a post-processing unit with a computer program that analyzes the data received and compares it with reference parameters stored in a database installed on a cloud server, fed by registered users.

[0019] The computer-based method involves obtaining digital images of the peritoneal cavity, previously demarcated by markers implanted through a laparoscopic procedure. The markers serve as references for determining positioning and sizing, allowing for the acquisition of precise information about the peritoneal cavity through digital analysis of the images, followed by image processing and comparison with data stored on a cloud server, in order to establish a set of information and treatment for individual intervention. SUMMARY

[0020] The invention describes a system and method for measuring the extent of peritoneal cavity involvement, over time and the application of specific treatment, through steps of marking regions of interest to define a reference for obtaining images, pre-processing, analysis, processing and comparison with data from a database with relevant information from other patients to identify patterns of evolution or not of the pathology.

[0021] The invention describes a system whose data is stored encrypted in a remote database, such as a data ecosystem, which associates information on the natural history of the disease and the treatment of each individual.

[0022] The invention describes a computer-implemented method to identify and measure peritoneal cavity involvement using digital mask dissociation and image superimposition to identify geometric, physical and behavioral characteristics, over time, of different pathologies.

[0023] The invention describes a system and method that allows associating images and information from populations with similar characteristics for detailed and accurate description of physical, geometric, distribution and behavior characteristics of the pathology under analysis, in order to document and identify individual or collective characteristics of the pathologies.

[0024] The invention describes a minimally invasive marking method that makes it possible to accurately identify the region and characteristics of metastatic nodules present in the area of ​​interest, allowing for the accurate obtaining of physical and geometric characteristics and behavior throughout the therapeutic process, based on the identification of patterns that may not be apparent at first glance.

[0025] The invention describes a method that allows obtaining data about the total number of nodules in the area of ​​interest, which may indicate a more complex and widespread condition or a more localized condition, from data related to the volume and percentage of nodules in certain diameter ranges.

[0026] The invention describes a method that allows comparing the physical and geometric characteristics of the detected nodules and the changes in response to established treatments, measuring and quantifying the magnitude of the observed response, in order to provide comparative reports of the individual and collective analysis, based on the data stored in an ecosystem.

[0027] The invention describes a method that can indicate the severity or extent and nature (benign or malignant) from the spectral evaluation of the color and total area of ​​nodules. BRIEF DESCRIPTION OF THE FIGURES

[0028] Figure 1 presents the schematic representation of the system, highlighting the basic components defined as marker (10), image acquisition unit (20), pre-processing unit (30) and remote processing unit (40).

[0029] Figure 2A shows details of the marker, figure 2B shows details of the top surface of the marker and figure 2C shows an example of a diagram of the positioning of the markers in the peritoneal cavity.

[0030] Figure 3A shows the representation of the marker placed in the cartridge, figure 3B shows the marker being moved from the internal region of the cartridge by applying manual pressure to the rod and figure 3C shows the marker in position for application in the peritoneal cavity.

[0031] Figure 4A shows the flowchart of the computer-implemented method to identify, measure and evaluate the therapeutic response of the extent and distribution of the tumor burden in the peritoneal space, highlighting the steps from applying the marker in the peritoneal space to sending the pre-processed images (PPI) to the unit. remote processing and Figure 4B shows the steps of the method executed in the remote processing unit.

[0032] Figure 5A shows the image obtained after validation of the pre-processing system; Figure 5B shows the definition of circular regions on the image obtained by plotting approximately circular regions obtained after calculating the Gaussian difference between the different pixels of the images of layer zero and layer X; Figure 5C shows the thresholding technique in the processed image to identify regions of interest characterized by a certain homogeneity of color intensity (Oncological Risk Pixel Set - CPRO). DETAILED DESCRIPTION OF THE INVENTION

[0033] For the purposes of the present invention, the following terms are conceptualized:

[0034] "Area of ​​interest (100a)” comprises an extension of an individual's peritoneal cavity, demarcated by at least one marker (10), where visual exploration will be carried out by means of obtaining digital images;

[0035] Original image (OI) comprises the digital image obtained from the area of ​​interest together with the marker;

[0036] Peritoneal Digital Localization (PDL) comprises the position of the areas and elements of interest in relation to the marker (10) in the obtained images.

[0037] Oncological risk pixel set (CPRO) comprises an area with a certain homogeneity of color intensity of the pixels that correspond to peritoneal nodules or risk of active neoplasia.

[0038] The system to identify, measure and evaluate the therapeutic response of the extent and distribution of tumor burden in space peritoneal, object of the present invention, comprises at least one marker (10) implanted in the peritoneal cavity (100) of an individual. The marker (10) constitutes a reference in the area of ​​interest (100a) of the peritoneal cavity (100) for obtaining digital images. The digital images obtained by an image acquisition unit (20) and referenced by the marker(s) (10) are forwarded to a pre-processing unit (30) and then to a remote unit installed in the cloud (40), such as a data ecosystem. Figure 1 shows the schematic representation of the system object of the present invention.

[0039] The marker (10) comprises a physical structure with a dual function. In a first function, the marker (10) demarcates an area of ​​interest (100a) in the peritoneal cavity (100), constituting a reference for obtaining the parameters obtained through the computer-implemented method. In a second function, the marker (10) presents visual elements that, analyzed by the computer program, allow validating the image obtained (in the context of the present invention called “original image (OI)”) to be processed in subsequent steps of the method.

[0040] The marker (10) has a top surface (11), preferably, but not limited to, a circular surface, and an anchoring means (12) associated with the top surface (11).

[0041] The marker (10) is implanted in the peritoneal cavity (100) by means of a videolaparoscopy procedure. The anatomical location of the marker (10) in the peritoneal space (100) is defined by the user, based on the areas of interest (100a) desired to be explored.

[0042] Preferably, the anchoring element (12) of the marker (10) presents indentations (121), as shown in figure 2A, ensuring the fixation of the marker (10) on the peritoneal wall and the stabilization of the top surface (11).

[0043] The stabilization of the top surface (11) of the marker (10) is particularly relevant as it references the digital images obtained from the area of ​​interest (100a) of the peritoneal cavity (100) and validates the digital image in the pre-processing unit (30).

[0044] Visual elements (11 1), such as color printing, pictograms or similar, are arranged on the top surface (11) of the marker (10). The visual elements (1 11) are recognized in the pre-processing and processing stage of the digital images to determine their peritoneal location and azimuth. The azimuth is the centering element of the Original Image (OI) for the comparison of future images of the same region and the element used to determine the Digital Peritoneal Location (DPL) of the elements of interest in the obtained images. The determination of the intraperitoneal anatomical position of the marker (10) for sectorization, focus, sharpness, angle and ideal distance of image capture, reference for Digital Peritoneal Location (DPL) and reference for calculating area and volume and determining the set of oncological risk pixels (CPRO) in the region of interest (100a).

[0045] Preferably, the top surface (1 1) is divided into four quadrants as shown in Figure 2B.

[0046] Preferably, the marker (10) is positioned in a cartridge (13) equipped with a plunger (131) which, when actuated by a drive rod (132), moves the marker (10) in the axial plane so that the anchoring element (12) is fixed to the peritoneal wall (100).

[0047] An image acquisition unit (20) captures images of the area of ​​interest (100a) of a peritoneal space (100) from the referencing of at least one marker (10).

[0048] The image acquisition unit (20) preferably comprises a videolaparoscopy tower and a rigid scope or rigid endoscope.

[0049] An electronic pre-processing unit (30) includes a computing board that executes a computer program that converts the original image (OI) received from the image acquisition unit (20) to a processing format required by the operating system, validates the original image (OI) and modifies one or more attributes of said original image (OI), obtaining a pre-processed image (PI) in accordance with, but not necessarily corresponding to, the Digital Imaging and Communications in Medicine (DICOM) system.

[0050] The DICOM system represents a set of standards created to standardize the electronic format used for storing and communicating images. With standardization, all types of exams—CT scans, MRIs, X-rays, etc.—are stored in a single format, allowing them to be exchanged between devices of different brands. The creation of this set of standards enables images to be recognized and viewed on any device.

[0051] In the context of the present invention, the term “validate” comprises comparing the image captured from the marker (10), which becomes the reference for positioning and image quality, with a standard image stored in the electronic pre-processing unit (30). The ideal capture position identified from each marker (10) becomes the reference for image acquisition, referencing the positioning for the overlay of future or previous images.

[0052] The peritoneal anatomical position of the marker (10) is loaded into the pre-processing unit (30) and sent to and stored in the data processing unit (40). During a new application for a given patient, the pre-processing unit (30) retrieves the record of the peritoneal anatomical position of the marker (10) stored in the data processing unit (40). Thus, each marker has an identifier, captured in digital images, which allows the reproducibility of its positioning in future events, ensuring that a set of images is obtained to assess the evolution of the pathology.

[0053] In the pre-processing unit (20) an interface is provided for viewing the images captured by the acquisition unit (20) and including data relating to the patient and the pathology.

[0054] For capturing the image by the image pre-processing unit (20), a physical means (21), such as a button or pedal, is provided to trigger image acquisition. Optionally, the image capture is performed semi-automatically, with the aid of positional masks that are generated in the interface of the pre-processing unit (30), said masks constituting templates that allow positioning the image acquisition unit (20) centered on the visual references present on the top surface (11) of the marker (10).

[0055] The human-machine interface enables the loading of data relating to the identification of the individual undergoing treatment and data related to the treatment and the organization of image files to be transferred to the remote processing unit (40).

[0056] Among the attributes of the original image (OI) that can be changed in the pre-processing unit (30), we can mention the adjustment of the azimuth, focus, sharpness and centering of the image on the marker (10).

[0057] If the pre-processing unit (30) identifies a non-conformity in relation to the image acquired from the marker (10), a signal is sent to the pre-processing unit interface to obtain new images of the area of ​​interest (100a).

[0058] The set of preprocessed images (PI) of an area of ​​interest (100a) is stored in a labeled file, linking the image and the associated marker (10). The metadata accompanying each image is stored for later retrieval and analysis, aiding in the processing and comparative analysis of the different images of different regions captured at different points in the disease / treatment progression. The preprocessed image (PI) file generated in the preprocessing unit (30) is transferred via a communication protocol to a remote processing unit (40), using encryption.

[0059] The remote processing unit (40) is stored on a cloud server, having a computer program using artificial intelligence techniques and algorithms, which performs the final processing of the images (FPI).

[0060] The method for identifying, measuring and evaluating the therapeutic response of the extent and distribution of the tumor burden in the peritoneal space, described in the present invention and using the detailed system, comprises an initial step of referencing an area of ​​interest (100a) in the peritoneal space (100).

[0061] In the step of referencing the peritoneal space (100), at least one marker (10) is implanted. In a preferred embodiment, for the broad analysis of the peritoneal space (100), a quadrant is preferably defined in the user's abdomen, with four regions of interest (100a) having at least one marker (10) installed in each region (100a).

[0062] In the image acquisition step of the area of ​​interest (100a) referenced by at least one marker (10), the image acquisition unit (20) is positioned in alignment with the marker (10). The alignment visualization is verified in the interface of the pre-processing unit (30) connected to the image acquisition unit (20).

[0063] The images obtained in the area of ​​interest (100a) are forwarded to the pre-processing unit (30) where they are transformed into the processing format required by the processing unit (40).

[0064] In subsequent captures, the original image (OI) obtained at each moment is validated following the pre-processing pattern of the described method. In this step, the original image (OI) undergoes the same pre-processing process until it is sent to the remote processing unit (40) for the final analysis of the parameters of interest in the image.

[0065] The set of pre-processed images (PI) of a given area of ​​interest (100a) is associated with the respective marker (10), constituting an identified file that is transferred to a remote processing unit (40) using communication protocols.

[0066] Image files are made up of personal and confidential information, so the data is represented for standardization of medical information in the DICOM system, but not limited to this format.

[0067] Illustratively, the encrypted digital file follows a pattern, for example, PACID23_0001_T01_CLIP1_R01, where the first alphanumeric sequence (PACID23) is the user identifier, the next numeric sequence (0001_T01) corresponds to the application, the second alphanumeric sequence (CLIP1) comprises the marker identification (10) and the third alphanumeric sequence (R01) comprises the marker application quadrant (10) in the peritoneal area.

[0068] In the processing unit (40), using a computer program that executes artificial intelligence algorithms, the pre-processed images (PPI) are subjected to several processing steps.

[0069] In the first processing step, the algorithm is executed to obtain the initial zero image layer. In this layer, the nodules, areas of interest, and marker (10) are identified. Based on marker (10), a standard comparison mask for layer zero is defined, containing all the oncological risk pixels (CPRO) of interest in the preprocessed image (PPI). In this first step, it is possible to obtain the total number of recognizable nodules in the area of ​​interest (100a).

[0070] In the sequential algorithm of the first processing stage, the preprocessed image (PPI) is converted to black and white and contrast is enhanced. Algorithms such as Gray Level Transformation and / or Histogram Equalization, among others, are preferably used.

[0071] Following the processing algorithm are applied Digital filters are used to create edge detection by highlighting image features. Sobel, Laplacian, or Gaussian algorithms are preferably applied. For digital carcinomatosis identification, the difference of Gaussians (DoG) is calculated between the different pixels of the layer zero images, resulting in images with approximately circular regions, as shown in Figure 4A. From this analysis, a Blob (Binary Large Object) is determined. Once the blobs are characterized and identified, libraries with the function of plotting circular regions on the image are used, as shown in Figure 4B.

[0072] Following the processing algorithm, digital color filters are applied to remove or enhance certain colors in the image. Preferably, the Median Filter and Mean Filter algorithms are used.

[0073] In a second processing step, layer zero is unfolded into at least one X sublayer, where the pixels obtained by successive scans are processed in order to obtain the segment edges. The number of X sublayers can be determined based on the number of images obtained from the area of ​​interest (100a). At this stage, the computer program recognizes the nodules, the areas of interest (100a), and the marker (10) in the X sublayers. In this second processing step, using the digital reference of the marker (10), the distance and position of the most central pixel (oncological risk pixel (ORP)) of each area of ​​interest (100a) or nodule of interest are determined, using the fixed measurement patterns recognized on the top surface of the marker (10), defining the Peritoneal Digital Location of each of the elements. of interest contained in the image.

[0074] The various X sublayers processed in the first and second stages are stored in the data ecosystem, in the context of the present invention called final processed images (FPI). The final processed images (FPI) are compared to identify any changes in the image pattern.

[0075] The X sublayers are superimposed to determine the presence or absence of previous markings. Each central pixel of each image layer is compared according to its Peritoneal Digital Location, identifying changes in the recognition pattern and in the quantities analyzed in each of the X sublayers. The data obtained are stored in the processing unit's database (40) for analysis and comparison with the data ecosystem, in order to identify trends and / or changes in patterns in relation to cases with the same pathology characteristics, individual and treatment used.

[0076] To calculate the areas of the region of interest or peritoneal nodule, each blob has its radius calculated from the known measurements of the marker (10). From the known measurements of the marker (10), the measurement of the circular region of interest is performed, maintaining the proportional scale between the size of the known object and the size of the circular region of interest.

[0077] As a final step in preprocessed image processing (PPI), the region of interest (100a) is segmented to unify distinct regions with similar specific characteristics. Watershed and Thresholding algorithms are preferably used. In this identification and processing step, the thresholding technique is applied to the processed image (FPI) to identify regions of interest characterized by a certain homogeneity of color intensity, in the context of the present invention called the Oncological Risk Pixel Set - CPRO.

[0078] To perform segmentation, a convolutional neural network (CNN) is used. CNNs are a category of deep neural networks specifically designed to process visual data, such as images. A CNN essentially works by passing input data (image pixels) through nodes. At each node, the data is weighted by adjustable parameters known as weights. These weights are responsible for capturing relevant information from the image during network training. Each node also employs activation functions, which determine how the weighted data is processed. These functions are crucial for defining and identifying important features in images, such as edges, textures, or specific patterns.

[0079] The first layer of a CNN is the convolution layer. In this layer, small filters (also called kernels) scan the original image, performing convolution operations. These operations capture simple patterns, such as edges and textures, in different parts of the image. Each filter generates a "feature map" that highlights the presence of these patterns in the image.

[0080] After the convolutional layer, an activation function, usually ReLU (Rectified Linear Activation), is applied. This function introduces nonlinearity into the network, allowing it to learn complex relationships between the identified patterns.

[0081] The pooling layer aims to reduce the dimensionality of the image, preserving the most important information. relevant. A common method is Max pooling, where the image is divided into regions and only the maximum value of each region is kept, thus reducing the size of the feature maps.

[0082] After the convolutional and pooling layers, the resulting feature maps are flattened into a one-dimensional vector and fed into one or more fully connected (dense) layers. These layers perform matrix operations that learn more complex and abstract representations of the features. The last layer of the network is the output layer, which produces the network's final predictions, depending on the task the CNN is performing.

[0083] To obtain the number of segmented areas, objects (blobs) that correspond (or may correspond) to the nodes or areas of interest in the image are counted. To count objects in images, Blob Analysis methods are used to analyze consistent regions in the image. A complementary technique called Detected Component Analysis can be used to identify and label objects in the segmented image, grouping pixels belonging to the same object into a connected component (a set of connected pixels in the image that share the same characteristics).

[0084] To delimit regions of interest and identify the area present in the segmented image, supervised or unsupervised classification techniques are used. In supervised classification, a labeled training dataset is used to teach the algorithm to identify specific patterns and classify new data into predetermined categories. Unsupervised classification does not require a labeled training dataset. labeled training, automatically grouping data into categories based on common characteristics.

[0085] To analyze the characteristics of carcinomas in segmented images, such as size, shape, and texture, feature classification algorithms, such as the Haralick Method, are used to extract texture information from images using statistical and mathematical techniques. The extracted features are then used in machine learning models, forming a labeled dataset to recognize specific patterns in the images and classify new images into predetermined categories.

Claims

CLAIMS:

1. System for identifying, measuring and evaluating the therapeutic response of the extent and distribution of the tumor load in the peritoneal space characterized by comprising: a) at least one marker (10) provided with a top surface (1 1) with visual elements (11 1) and an anchoring means (12) associated with the top surface (1 1).b) an image acquisition unit (20) that captures images of the area of ​​interest (100a) of a peritoneal space (100) from the referencing of at least one marker (10); c) a pre-processing unit (30) that includes a computing board that executes a computer program that converts the original image (OI) received from the image acquisition unit (20), obtaining a pre-processed image (PI); d) a remote processing unit installed in the cloud (40) equipped with a computer program using artificial intelligence techniques and algorithms for processing the pre-processed images (PI) in the pre-processing unit (30) and storing the final processed images (FPI).

2. System, according to claim 1, characterized in that the marker (10) is implanted in the peritoneal cavity (100) by means of a videolaparoscopy procedure.

3. System, according to claim 1, characterized in that the anchoring element (12) of the marker (10) presents indentations (121).

4. System, according to claim 1, characterized by the fact that the marker (10) is positioned in a cartridge (13) equipped with a piston (131) with a drive rod (132) that moves said marker (10) in the axial plane.

5. System, according to claim 1, characterized in that the image acquisition unit (20) comprises a videolaparoscopy tower and a rigid optic or rigid endoscope.

6. System, according to claim 1, characterized in that the pre-processing unit (20) presents an interface for viewing the images captured by the acquisition unit (20) and including data.

7. Method for identifying, measuring and evaluating the therapeutic response of the extent and distribution of tumor burden in the peritoneal space, characterized by comprising the steps: a) referencing an area of ​​interest (100a) in the peritoneal space (100) by implanting at least one marker (10); b) acquiring images of the area of ​​interest (100a) of the peritoneal space (100) based on the implanted marker (10); c) validating the original image (OI) by comparing the captured image of the marker (10) with the reference image of the marker (10); d) modifying one or more attributes of the original image (OI), obtaining a pre-processed image (PI); e) associating the set of pre-processed images (PI) of a given area of ​​interest (100a) to a marker (10), constituting an identified file; f) processing the pre-processed images (PPI) including: f.1) obtaining the initial image layer zero with identification of marker (10), nodules and areas of interest; f.2) pre-processed image (PPI) is converted to black and white and contrast improvement is performed; f.3) digital filters are applied for edge detection by highlighting image features; f.4) the Gaussian difference calculation is performed between the different pixels of the layer zero images; f.5) characterization and identification of blobs; f.6) plotting of circular regions on the layer zero image; f.7) application of digital color filters; f.8) unfolding of layer zero into at least one sublayer X; f.9) determining the distance and position of the most central pixel (oncological risk pixel (PRO)) of each area of ​​interest (100a) or nodule of interest using the fixed measurement patterns recognized on the top surface (11) of the marker (10): g) superposition of the X sublayers with the comparison of each central pixel of each image layer according to its Peritoneal Digital Location; h) storage of the data obtained in the database of the processing unit (40); i) calculation of the area of ​​the blobs from the known measurements of the marker (10); j) segmentation of the region of interest (100a) to unify distinct regions with similar specific characteristics; k) counting of blobs that correspond to the nodules or areas. of interest of the image; I) analysis of the characteristics of the blobs in the segmented image using feature classification algorithms.

8. Method, according to 7, characterized by the fact that the peritoneal space (100) is divided into four regions of interest (100a) with at least one marker (10) installed in each region (100a).

9. Method, according to claim 7, characterized in that the attributes of the original image (OI) to be changed include adjustment of the azimuth, focus, sharpness and centering of the image on the marker (10).

10. Method, according to claim 7, characterized in that the calculation of the areas of the region of interest or peritoneal nodule is calculated from the measurement of the radius of each blob in comparison with the known measurements of the marker (10). 1 1. Method according to claim 7, characterized in that the number of sublayers X is determined based on the number of images obtained from the area of ​​interest (100a).

12. Method, according to claim 7, characterized in that the blob counting includes the identification and labeling of the blobs present in the segmented image, with grouping of the pixels that have the same characteristics.

Images