Method for obtaining pathological diagnosis according to color Doppler ultrasound image of breast nodule by applying artificial intelligence technology

Abstract

The invention relates to a method for obtaining pathological diagnosis according to color ultrasound images of breast nodules by applying an artificial intelligence technology. The breast nodule color ultrasound image-pathological diagnosis system has the functions of image acquisition, image preprocessing, feature extraction, image recognition and the like, and can be used for recognizing, extracting and memorizing color ultrasound image features of breast nodules and internal features of abnormal cells in pathological tissue staining pictures of the breast nodules. Color Doppler ultrasound image data of breast nodules directly correspond to pathological picture data, and the data can be interpreted. A color Doppler ultrasound image database and a pathological picture database of breast nodules are established, deep learning is carried out with sufficient data, and functions are improved. The system is verified and clinically tested to ensure the accuracy of the system. Pathological diagnosis can be displayed after color Doppler ultrasound image data of breast nodules are input into the system. The device has the advantages that pathological diagnosis can be judged when nodules are seen on breast color ultrasound, so that the aim of accurate treatment is fulfilled. Clinical final diagnosis is based on pathological results under a microscope.

Description

Technical Field

[0001] This invention belongs to the field of medical technology and relates to the application of artificial intelligence in the fields of medical imaging and pathological diagnosis. Its essence is to invent a method for deriving pathological diagnosis based on high-resolution color Doppler (color ultrasound) images of breast nodules. Background Technology

[0002] Breast nodules are abnormal growths within breast tissue, usually solitary, and are typically formed by abnormal cell proliferation or changes in breast structure. The vast majority of nodules are benign, such as cysts, fibroadenomas, breast hyperplasia nodules, lipomas, and hamartomas, while about 20% are malignant.

[0003] Breast nodules are often discovered during physical examinations and ultrasound scans. The morphological characteristics of breast nodules on color Doppler ultrasound are important indicators for assessing benignity or malignancy, including shape, borders, margins, internal structure, presence of calcification, echogenicity, aspect ratio, and internal blood flow. Currently, color Doppler ultrasound can be used to classify breast nodules into TI-RADS risk categories. TI-RADS 2 and TI-RADS 3 nodules are highly likely to be benign; while TI-RADS 4 and TI-RADS 5 nodules are highly likely to be malignant. This experience-based classification does not achieve a definitive diagnosis unless a breast nodule biopsy is performed, which is invasive. Other examinations include mammography (which involves radiation) and MRI (magnetic resonance imaging), which is expensive. In some cases, surgery may be performed too early or when intervention is unnecessary, leading to overtreatment. In some cases, the surgical approach is determined intraoperatively based on frozen section pathology results, which is cumbersome and time-consuming. In some cases, discrepancies in pathological classification or subtype between the surgical frozen section and the initial pathology report were only discovered several days after surgery, when the official pathology report was released. In such cases, patients may require a second surgery or additional adjuvant therapy. This impacts treatment outcomes, increases patient suffering, and wastes medical resources. Therefore, it is crucial to further utilize the ultrasound imaging characteristics of breast nodules to determine whether they are benign or malignant, and to decide whether intervention is necessary. This necessitates the development of a new diagnostic solution for breast nodules based on ultrasound imaging and pathology.

[0004] Artificial intelligence (AI) technology is already quite mature. Image recognition technology is formed by processing, analyzing, and understanding images using computer vision technology. By applying this technology to build models and create algorithms, and through steps such as image acquisition, image preprocessing, feature extraction, and image recognition, a medical image AI analysis system has been established and applied to AI-powered color Doppler ultrasound image diagnosis. The AI-powered color Doppler ultrasound-breast nodule diagnostic system shows improvements in both accuracy and sensitivity compared to simple color Doppler ultrasound.

[0005] Traditionally, postoperative tissue specimens are examined by the naked eye using a microscope to observe stained sections and make pathological diagnoses. However, with advancements in computer vision technology, computational pathology has emerged. Studies show that by analyzing digitized images of stained pathological tissues and performing machine learning on large-scale data, artificial intelligence-based pathological diagnostic systems have achieved extremely high accuracy and reliability in diagnosing cancerous tissue specimens.

[0006] However, currently, medical AI imaging diagnostic systems and postoperative pathological diagnosis exist independently, and it is not yet possible to determine the pathological diagnosis simply by seeing a lesion on an imaging image. This is precisely the direction of future development. Therefore, this invention connects an AI-powered color Doppler ultrasound imaging diagnostic system for diagnosing breast nodules with postoperative pathological diagnosis and related AI-powered pathological diagnostic systems for breast nodules. This allows for the matching of ultrasound images of breast nodules with corresponding pathological results, leading to a clear diagnosis and ultimately achieving precise treatment. Summary of the Invention

[0007] To address the existing problems, this invention provides a solution: applying artificial intelligence technology to create a method for deriving pathological diagnoses from ultrasound images of breast nodules, meaning that the pathological diagnosis can be determined simply by seeing a breast nodule. The invention includes: 1. Designing a software diagnostic system for breast nodule ultrasound images and pathology, solving comprehensive algorithm problems, and enabling deep learning; 2. This diagnostic system can identify the ultrasound image features of breast nodules and the abnormal cell features in their pathological tissue staining images, and establish a correlation between the two; 3. This diagnostic system has the function of forward and reverse interpretation of breast nodule ultrasound image data and pathological tissue image data, as well as a memory function; 4. Establishing a database of breast nodule ultrasound images and a database of pathological tissue staining images, with sufficient data for large-scale machine learning; 5. This diagnostic system has the ability to derive a pathological result upon seeing a breast nodule on ultrasound; 6. The pathological diagnosis made by this system will be displayed in graphical or textual form.

[0008] Compared with existing technologies, the present invention has the following advantages: This diagnostic system is a new diagnostic technology invented by applying artificial intelligence technology to perform deep learning on a large amount of color Doppler ultrasound image data and pathological tissue image data of breast nodules; it can achieve the goal of obtaining pathological diagnosis directly based on the color Doppler ultrasound images of breast nodules; since the development process of lesions within breast nodules is clearly understood, appropriate intervention methods can be adopted to implement precise treatment, which can prevent both missed diagnosis and overtreatment.

[0009] The technical solution is as follows: 1. Establish an artificial intelligence-based ultrasound imaging-pathology diagnostic system for breast nodules. Utilize machine learning, natural language processing, and computer vision algorithms to construct a new algorithm model that associates the ultrasound imaging features of breast nodules with the abnormal cell features in their pathological images, thus forming the ultrasound imaging-pathology diagnostic system for breast nodules; 2. This diagnostic system has functions such as image acquisition, image preprocessing, feature extraction, and image recognition and association. It can identify, extract, and memorize the ultrasound imaging features of breast nodules and the internal features of abnormal cells in their pathological tissue staining images. It can also perform forward and reverse interpretation of the ultrasound imaging data and pathological image data of breast nodules; 3. Different morphological features of breast nodules on ultrasound images will be associated with benign or malignant nodules on pathology. The pathological diagnosis will be displayed simultaneously after the ultrasound imaging data of each patient's breast nodules is input; 4. Establish a database of 100,000 surgically treated breast nodule patients' ultrasound images and corresponding breast nodule data. The system comprises: 1) A database of 100,000 breast nodule ultrasound images. For each breast nodule ultrasound image input, the corresponding patient's pathological image data is input simultaneously. The ultrasound image features of the breast nodule are correlated with the internal features of abnormal cells in the stained pathological tissue images, and deep learning is used to refine its functionality. 2) Validation of the database to test the performance of the diagnostic system: This requires ultrasound images of breast nodules from 10,000 patients and their corresponding clinical pathology reports from patients who have undergone surgery. The system inputs ultrasound images of a patient's breast nodule and verifies whether the displayed pathological diagnosis is consistent with the clinical pathology results, ensuring accuracy even with unseen data. 3) A test database also requires 10,000 patients. Based on the ultrasound images of the patient's breast nodule, the system predicts the pathological diagnosis before surgery and compares it with the postoperative clinical pathology report to evaluate the accuracy of the diagnostic system. 4) The diagnosis made by the AI-powered breast nodule ultrasound image-pathology diagnostic system is presented in graphical or textual form.

[0010] Of course, the final clinical diagnosis is still based on microscopic pathological diagnosis.

[0011] Example Description: When a patient's ultrasound reveals a breast nodule, the ultrasound image data of the nodule is automatically input into the artificial intelligence breast nodule ultrasound image-pathology diagnosis system. Based on the ultrasound image characteristics of the breast nodule, the system will identify the corresponding pathological features, make a pathological diagnosis, and display the results.

Claims

1. A method for deriving pathological diagnosis from high-resolution color Doppler ultrasound images of breast nodules using artificial intelligence technology, characterized by: With the support of artificial intelligence technology, a system can make a pathological diagnosis as soon as a breast nodule is seen on a color Doppler ultrasound.

2. The diagnostic system according to claim 1 is a breast nodule color Doppler ultrasound imaging-pathology diagnostic system formed by constructing a large-scale artificial intelligence model using artificial intelligence machine learning algorithms, natural language processing algorithms, computer vision algorithms, and DeepSeek technologies. This diagnostic system has functions such as image acquisition, image preprocessing, feature extraction, image recognition, and correlation between color Doppler ultrasound images and pathological images.

3. The diagnostic system according to claims 1 and 2 can identify, extract, and memorize the imaging features of breast nodules from color Doppler ultrasound and the internal features of abnormal cells in their pathological tissue staining images. It can also perform forward and reverse interpretation of the color Doppler ultrasound image data and pathological image data of breast nodules. Deep learning is performed using large-scale data, and the results are then detected and validated.

4. The diagnostic system according to the preceding claims will display the pathological diagnosis in graphical or textual form.