Method and system for evaluating treatment means for thyroid cancer in the elderly based on multi-modal omics
By using a multimodal omics assessment system, combined with ultrasound images and underlying health conditions of elderly patients, treatment options for thyroid cancer in the elderly can be evaluated, solving the individualized challenge of thyroid cancer management in the elderly and achieving precision diagnosis and treatment as well as cost savings.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV
- Filing Date
- 2022-10-26
- Publication Date
- 2026-06-12
Smart Images
Figure CN115620861B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of medical image processing technology, specifically relating to a method and system for evaluating treatment methods for thyroid cancer in the elderly based on multimodal omics. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] The incidence and detection rate of thyroid cancer are rapidly increasing, making it the fastest-growing malignant tumor. Thyroid cancer has become a common tumor worldwide, primarily differentiated thyroid carcinoma (DTC), especially papillary thyroid carcinoma (PTC), whose incidence generally increases with age. It is generally believed that age is related to the prognosis of thyroid cancer, with older patients experiencing a worse prognosis. Currently, the management of thyroid cancer in the elderly population has not received sufficient attention. The epidemiological and clinical characteristics of thyroid cancer in the elderly differ from those in ordinary adults, especially given the higher prevalence of underlying diseases, poorer prognosis, higher surgical risks, and a higher incidence of adverse reactions to thyroid-stimulating hormone (TSH) suppression therapy. These factors significantly increase the difficulty of managing thyroid cancer in the elderly compared to ordinary adults. Balancing the benefits and risks of treatment decisions is a clinical challenge in managing elderly patients with thyroid cancer. Despite the abundance of guidelines and consensus statements, no authoritative institution has yet issued guidelines or recommendations for the standardized diagnosis and treatment of thyroid cancer in the elderly.
[0004] Currently, there are three main treatment methods for thyroid diseases: active surveillance, thermal ablation, and surgery. Active surveillance (AS) is used because PTC (papillary thyroid microcarcinoma) has a high incidence, low mortality, and slow progression. Several centers have implemented AS for low-risk PTC, especially PTC, particularly PTC (papillary thyroid microcarcinoma, PTMC). This involves actively monitoring the disease after diagnosis until significant progression occurs, at which point surgery is considered. Thermal ablation is widely used in the treatment of tumors, including liver cancer, kidney cancer, bone cancer, and soft tissue tumors of the breast and head and neck. Elderly individuals often have poor underlying health conditions, and those with low-risk PTMC are more suitable for AS, where thermal ablation may have greater application value. However, it should be noted that elderly patients with thyroid cancer often have low differentiation rates and high rates of lymph node and distant metastasis; therefore, the risk of tumor residue, lymph node metastasis, and tumor recurrence should be rigorously assessed before ablation.
[0005] Currently, surgery remains the traditional first-line treatment for thyroid cancer. However, whether elderly thyroid cancer patients should undergo surgery and the choice of surgical method should be based on a comprehensive assessment of disease severity and prognosis. Summary of the Invention
[0006] To address the aforementioned issues, this invention proposes a method and system for evaluating treatment options for thyroid cancer in the elderly based on multimodal omics. This invention comprehensively analyzes the ultrasound images and basic physical condition of the elderly, and by weighing the advantages and disadvantages of three different treatment methods, it provides patients with more individualized and precise treatment options, while saving unnecessary costs and surgical treatments, which has significant clinical significance and economic benefits.
[0007] According to some embodiments, the first aspect of the present invention provides a multimodal omics-based evaluation system for the treatment of thyroid cancer in the elderly, employing the following technical solution:
[0008] A multimodal omics-based evaluation system for thyroid cancer treatment in the elderly includes:
[0009] The data acquisition module is configured to acquire patient ultrasound images, basic patient information, and basic patient disease information;
[0010] The image processing module is configured to determine the risk level of thyroid cancer based on the patient's ultrasound images;
[0011] The treatment risk assessment module is configured to assess the risk of treatment based on the risk level of thyroid cancer, the patient's basic information, and the patient's underlying disease information.
[0012] The treatment determination module is configured to determine the treatment method based on the risk assessment results of the treatment method.
[0013] Furthermore, the patient's basic information includes height, weight, age, basic physical examination information, and lifestyle habits.
[0014] Furthermore, the patient's underlying medical condition information refers to the patient's own underlying medical condition.
[0015] Furthermore, the determination of the risk level of thyroid cancer based on the patient's ultrasound images includes:
[0016] The lesion segmentation results are obtained by segmenting the patient's ultrasound images using a depth segmentation network.
[0017] The tumor area is calculated based on the lesion segmentation results, and the regularity of the shape is analyzed.
[0018] Based on whether the shape is regular and the location of the tumor, it can be determined whether it is a high-risk type of tumor.
[0019] Furthermore, the determination of whether a tumor is high-risk based on its shape and location specifically involves:
[0020] If the shape is irregular and the tumor is located in a position where it is prone to leakage, it is identified as a high-risk tumor.
[0021] If the tumor is regular in shape and not located in a position where it is prone to extravasation, it is identified as a non-high-risk tumor and a treatment approach of active observation is adopted.
[0022] Furthermore, the risk level of thyroid cancer is determined based on the patient's ultrasound images using a pre-trained classification convolutional neural network, which includes an area calculation module, a shape attention module, and a position attention module.
[0023] The area calculation module uses a deep segmentation network, specifically Unet or DeepLab basic segmentation network.
[0024] The shape attention module sets a circle by calculating the area, and then calculates the proportion of the tumor area to the circle. The larger the proportion, the more regular the shape, and the smaller the proportion, the more irregular the shape.
[0025] The position attention module is based on reinforcement learning. It sets up danger points and trains the user to move a certain distance. When the user moves to a danger point, it provides positive feedback, and when the user moves to a danger point, it provides negative feedback.
[0026] Furthermore, the risk assessment of treatment methods based on the risk level of thyroid cancer, patient baseline information, and patient baseline disease information includes:
[0027] If the risk level of thyroid cancer is high-risk, then determine the size of the tumor.
[0028] If the tumor area is too large, determine whether the patient has high-risk factors based on the patient's basic information;
[0029] If high-risk factors are present, determine whether the patient has any high-risk underlying diseases based on the patient's underlying disease information;
[0030] If high-risk diseases are present, a proactive observation approach should be adopted.
[0031] Furthermore, if the tumor is too small, thermal ablation is used as a treatment method.
[0032] Furthermore, if no high-risk disease exists, surgical treatment is employed.
[0033] According to some embodiments, a second aspect of the present invention provides an evaluation method for assessing treatment methods for thyroid cancer in the elderly based on multimodal omics, as described in the first aspect, employing the following technical solution:
[0034] The assessment method of the multimodal omics-based assessment system for thyroid cancer treatment in the elderly, as described in the first scheme, includes:
[0035] Collect patient ultrasound images, basic patient information, and information on the patient's underlying diseases;
[0036] Determining the risk level of thyroid cancer based on patient ultrasound images;
[0037] Risk assessment of treatment options based on thyroid cancer risk level results, patient baseline information, and patient underlying disease information;
[0038] The treatment method is determined based on the risk assessment results of the treatment options.
[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0040] This invention provides a method and system for evaluating treatment options for thyroid cancer in the elderly based on multimodal omics. It comprehensively analyzes the ultrasound images and basic physical condition of the elderly, and by weighing the advantages and disadvantages of three different treatment methods, it provides patients with more individualized and precise treatment options, while saving unnecessary costs and surgical treatments, which has significant clinical significance and economic benefits. Attached Figure Description
[0041] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0042] Figure 1 This is a structural block diagram of a multimodal omics-based evaluation system for thyroid cancer treatment in the elderly, as described in an embodiment of the present invention.
[0043] Figure 2 This is a flowchart illustrating the structural block diagram of a multimodal omics-based evaluation system for thyroid cancer treatment in the elderly, as described in an embodiment of the present invention.
[0044] Figure 3 This is a schematic diagram of the classification convolutional neural network described in an embodiment of the present invention. Detailed Implementation
[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0046] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0047] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0048] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.
[0049] Example 1
[0050] like Figure 1 As shown, this embodiment provides a multimodal omics-based evaluation system for thyroid cancer treatment in the elderly. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly includes:
[0051] The data acquisition module is configured to acquire patient ultrasound images, basic patient information, and basic patient disease information;
[0052] The image processing module is configured to determine the risk level of thyroid cancer based on the patient's ultrasound images;
[0053] The treatment risk assessment module is configured to assess the risk of treatment based on the risk level of thyroid cancer, the patient's basic information, and the patient's underlying disease information.
[0054] The treatment determination module is configured to determine the treatment method based on the risk assessment results of the treatment method.
[0055] The patient's basic information includes height, weight, age, basic physical examination information, and lifestyle habits. Basic physical examination information includes test results such as blood routine and urine routine tests.
[0056] The patient's underlying medical conditions are the patient's own underlying medical conditions.
[0057] The method of determining the risk level of thyroid cancer based on patient ultrasound images includes:
[0058] The lesion segmentation results are obtained by segmenting the patient's ultrasound images using a depth segmentation network.
[0059] The tumor area is calculated based on the lesion segmentation results, and the regularity of the shape is analyzed.
[0060] Based on whether the shape is regular and the location of the tumor, it can be determined whether it is a high-risk type of tumor.
[0061] It is understandable that deep segmentation networks include, but are not limited to, mainstream segmentation network frameworks such as FCN, GAN, and Unet.
[0062] Specifically, determining whether a tumor is high-risk based on its shape and location involves:
[0063] If the shape is irregular and the tumor is located in a position where it is prone to leakage, it is identified as a high-risk tumor.
[0064] If the tumor is regular in shape and not located in a position where it is prone to extravasation, it is identified as a non-high-risk tumor and a treatment approach of active observation is adopted.
[0065] The risk level of thyroid cancer is determined based on the patient's ultrasound images using a pre-trained classification convolutional neural network, which includes an area calculation module, a shape attention module, and a position attention module.
[0066] The area calculation module uses a deep segmentation network, specifically Unet or DeepLab basic segmentation network.
[0067] After segmenting the tumor using a deep segmentation network, the total number of pixels occupied by the segmented tumor is calculated, and the area of the tumor is calculated based on the correspondence between the number of pixels and the actual distance in the ultrasound machine.
[0068] The shape attention module sets a circle by calculating the area, and then calculates the proportion of the tumor area to the circle. The larger the proportion, the more regular the shape, and the smaller the proportion, the more irregular the shape.
[0069] The position attention module is based on reinforcement learning. It sets up danger points and trains the user to move a certain distance. When the user moves to a danger point, it provides positive feedback, and when the user moves to a danger point, it provides negative feedback.
[0070] The risk assessment of treatment methods based on thyroid cancer risk level results, patient baseline information, and patient underlying disease information includes:
[0071] If the risk level of thyroid cancer is high-risk, then determine the size of the tumor.
[0072] If the tumor area is too large, determine whether the patient has high-risk factors based on the patient's basic information;
[0073] If high-risk factors are present, determine whether the patient has any high-risk underlying diseases based on the patient's underlying disease information;
[0074] If high-risk diseases are present, a proactive observation approach should be adopted.
[0075] If the tumor is too small, thermal ablation is used as a treatment method.
[0076] If no high-risk disease is found, surgical treatment will be used.
[0077] Specifically, the data acquisition module is used to collect the patient's ultrasound image information, basic personal information, and basic disease information.
[0078] The data processing module is used to segment the location of lesions based on ultrasound images, calculate the area of the lesions, and analyze whether the shape is regular.
[0079] The model training module is used to train models for intelligent decision-making methods for thyroid cancer treatment in the elderly based on multimodal omics.
[0080] The results output module is used to output treatment recommendations, including active observation, thermal ablation, and surgical treatment.
[0081] As one or more implementation methods, the ultrasound acquisition device in data acquisition module 1 includes, but is not limited to, a handheld ultrasound device. Unlike the traditional ultrasound device consisting of a main unit and a probe, the main unit is miniaturized to a small circuit board built into the probe, making it essentially a single "probe" equivalent to an ultrasound machine. It can be displayed on a mobile phone or tablet with an ultrasound app installed, and the image is transmitted to the phone / tablet via the probe's built-in Wi-Fi. Basic personal information collection includes, but is not limited to, blood tests, urine tests, and medical consultations. Basic disease information collection includes, but is not limited to, retrieving past medical records and other examination methods.
[0082] The data processing module is used to process and analyze the various collected data. A data processing model based on an interpretable neural network is designed. This model, through comprehensive analysis of ultrasound information and basic information, obtains a classification result: high-risk tumors or non-high-risk tumors. The specific steps are as follows:
[0083] Firstly, for doctors, analyzing whether a thyroid tumor is high-risk involves considering several factors: whether the tumor area is excessively large, whether the tumor shape is regular, and whether the tumor is located in an area prone to spread. Based on this, a classification convolutional neural network (an interpretable neural network) is set up with three modules: an area calculation module, a shape attention module, and a location attention module. These three modules are trained together, using the doctor's thinking process as a guide, to train the neural network, thereby providing a basis for determining whether a thyroid tumor is high-risk and offering an interpretable explanation for the black-box classification method of the neural network.
[0084] The schematic diagram of the classification convolutional neural network (interpretable neural network) is as follows: Figure 3 As shown:
[0085] The network input consists of ultrasound images, segmentation label data annotated by doctors, lesion detection labels, dangerous location point label data, and label data indicating whether it is a high-risk tumor.
[0086] The lesion segmentation module uses deep segmentation networks, including but not limited to basic segmentation networks such as Unet and deeplab;
[0087] Based on the segmented lesion image, the total number of pixels occupied by the segmented tumor is calculated. The tumor area S is calculated based on the correspondence between the number of pixels and the actual distance in the ultrasound machine. The calculated area is assigned an initial weight value w1, and the loss value of the depth segmentation network is l1.
[0088] The shape attention module is trained using a depth detection network based on ultrasound images. The lesion is enclosed by a minimum bounding box. This depth detection network includes, but is not limited to, YOLO series networks, SSD, RetainNet, and Faster R-CNN. The loss value of the depth detection network is l2.
[0089] Then, based on the ratio R of the tumor area S to the lesion box area calculated by the segmentation network, the larger the ratio, the more regular the shape, and the smaller the ratio, the more irregular the shape. The calculated ratio is assigned an initial weight value w2.
[0090] The location attention module is based on reinforcement learning. Doctors set a danger point; the closer to this danger point, the greater the risk of tumor growth. The module uses this danger point as its initial point and trains it by moving it up, down, left, and right. When the danger point moves to the lesion location, positive feedback is given, and the output is distance information D. An initial weight w3 is assigned to distance information D, and the loss value for the location training network is l3.
[0091] The loss value of an interpretable deep network is calculated as follows:
[0092]
[0093] Where y is the actual value, f() is the network prediction value, and i is the number of training samples.
[0094] An interpretable neural network built on three modules can be used to classify tumors by combining diagnostic information from doctors.
[0095] 1. Area calculation is based on the total number of pixels, while morphological analysis refers to whether the shape is regular and whether the position is dangerous.
[0096] 2. The determination of high-risk tumors is based on a comprehensive assessment of three factors. This is a supervised training process. First, experienced doctors label tumors as high-risk. Then, through training, different weight values are assigned to the three factors of area, shape, and location. The final determination is obtained through comprehensive training.
[0097] 3. Once a high-risk tumor has been identified, it must be treated. The decision of whether to treat it with surgery or ablation depends on the size of the lesion. This size can be determined by analyzing the area obtained from lesion segmentation.
[0098] The steps of the entire decision-making model are as follows Figure 2 As shown, the process begins with acquiring ultrasound images, basic personal information, and basic medical history information for the patient. The obtained ultrasound images are then segmented using a depth segmentation network to obtain lesion segmentation results. Based on the segmentation results, the tumor area is calculated, the shape is analyzed for regularity, and the tumor location is determined. A comprehensive analysis is then performed to determine if the tumor is high-risk. Irregularly shaped tumors located in areas prone to extravasation are classified as high-risk. For non-high-risk tumors, the decision is to actively monitor the patient. For high-risk tumors, the size is assessed. For smaller tumors, thermal ablation is recommended. For larger tumors, the presence of high-risk factors such as advanced age and unhealthy habits like smoking and drinking is investigated. If high-risk factors are present, the presence of underlying diseases such as heart disease, hypertension, and diabetes is assessed. If the patient does not have any high-risk underlying diseases, surgical treatment is recommended; otherwise, active monitoring is recommended. Understandably, high-risk factors include: age over 75 years; unhealthy habits such as smoking and drinking; and a family history of cancer.
[0099] Example 2
[0100] This embodiment provides an evaluation method for assessing the treatment of thyroid cancer in the elderly based on multimodal omics, as described in Embodiment 1, using the following technical solution:
[0101] The evaluation method of the multimodal omics-based thyroid cancer treatment evaluation system for the elderly, as described in Example 1, includes:
[0102] Collect patient ultrasound images, basic patient information, and information on the patient's underlying diseases;
[0103] Determining the risk level of thyroid cancer based on patient ultrasound images;
[0104] Risk assessment of treatment options based on thyroid cancer risk level results, patient baseline information, and patient underlying disease information;
[0105] The treatment method is determined based on the risk assessment results of the treatment options.
[0106] Acquire ultrasound image data to be predicted, basic examination information of the elderly such as height, weight and lifestyle habits, as well as information on underlying diseases such as hypertension and diabetes;
[0107] Based on the ultrasound image data to be predicted, thyroid cancer lesions are segmented and the disease is classified into grades.
[0108] Based on the processed ultrasound image data and basic information, a trained decision model is used to obtain the optimal treatment plan for thyroid cancer in the elderly.
[0109] like Figure 2 As shown, the decision model refers to a model that integrates ultrasound images, basic examination information, and underlying disease information to make a final treatment prediction. This is a classification convolutional neural network (an interpretable neural network).
[0110] A multimodal omics-based intelligent decision-making method and system for thyroid cancer treatment in the elderly, including:
[0111] The data acquisition module is configured to acquire ultrasound image data to be predicted and basic information of the elderly.
[0112] The prediction module is configured to: obtain the prediction result of thyroid disease based on the ultrasound image data to be predicted using a trained decision model; wherein the decision model will comprehensively make decisions based on ultrasound images, basic examination information and basic disease information to obtain the final treatment prediction model.
[0113] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. A multimodal omics-based evaluation system for thyroid cancer treatment in the elderly, characterized in that, include: The data acquisition module is configured to acquire patient ultrasound images, basic patient information, and basic patient disease information; The image processing module is configured to determine the risk level of thyroid cancer based on the patient's ultrasound images; The risk level of thyroid cancer determined based on the patient's ultrasound images is determined using a pre-trained classification convolutional neural network, which includes an area calculation module, a shape attention module, and a position attention module. The area calculation module uses a deep segmentation network, specifically Unet or DeepLab basic segmentation network. The shape attention module sets a circle by calculating the area, and then calculates the proportion of the tumor area to the circle. The larger the proportion, the more regular the shape, and the smaller the proportion, the more irregular the shape. The position attention module is based on reinforcement learning. It sets up danger points and trains the movement distance. When the distance reaches the danger point, it gives positive feedback, and when it does not reach the danger point, it gives negative feedback. The treatment risk assessment module is configured to assess the risk of treatment based on the risk level of thyroid cancer, the patient's basic information, and the patient's underlying disease information. The treatment determination module is configured to determine the treatment method based on the risk assessment results of the treatment method.
2. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 1, characterized in that, The patient's basic information includes height, weight, age, basic physical examination information, and lifestyle habits.
3. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 1, characterized in that, The patient's underlying medical conditions are the patient's own underlying medical conditions.
4. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 1, characterized in that, The method of determining the risk level of thyroid cancer based on patient ultrasound images includes: The lesion segmentation results are obtained by segmenting the patient's ultrasound images using a depth segmentation network. The tumor area is calculated based on the lesion segmentation results, and the regularity of the shape is analyzed. Based on whether the shape is regular and the location of the tumor, it can be determined whether it is a high-risk type of tumor.
5. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 4, characterized in that, The determination of whether a tumor is high-risk based on its shape and location is as follows: If the shape is irregular and the tumor is located in a position where it is prone to leakage, it is identified as a high-risk tumor. If the tumor is regular in shape and not located in a position where it is prone to extravasation, it is identified as a non-high-risk tumor and a treatment approach of active observation is adopted.
6. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 1, characterized in that, The risk assessment of treatment methods based on thyroid cancer risk level results, patient baseline information, and patient underlying disease information includes: If the risk level of thyroid cancer is high-risk, then determine the size of the tumor. If the tumor area is too large, determine whether the patient has high-risk factors based on the patient's basic information; If high-risk factors are present, determine whether the patient has any high-risk underlying diseases based on the patient's underlying disease information; If high-risk diseases are present, a proactive observation approach should be adopted.
7. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 6, characterized in that, If the tumor is too small, thermal ablation is used as a treatment method.
8. The multimodal omics-based evaluation system for thyroid cancer treatment in the elderly as described in claim 6, characterized in that, If no high-risk disease is found, surgical treatment will be used.
9. The evaluation method of the multimodal omics-based thyroid cancer treatment evaluation system for the elderly as described in any one of claims 1-8, characterized in that, include: Collect patient ultrasound images, basic patient information, and information on the patient's underlying diseases; Determining the risk level of thyroid cancer based on patient ultrasound images; Risk assessment of treatment options based on thyroid cancer risk level results, patient baseline information, and patient underlying disease information; The treatment method is determined based on the risk assessment results of the treatment options.