A postoperative care scheme optimization method for liver cancer and an intelligent chip

Abstract

The application discloses a postoperative care scheme optimization method and an intelligent chip for liver cancer, and relates to the technical field of postoperative care for liver cancer. The method comprises the following steps: performing multidimensional detection on a liver cancer tumor tissue sample of a target user to obtain a multidimensional detection data set, determining a tumor immune microenvironment type, and generating an initial postoperative care scheme; obtaining a predicted recurrence and metastasis risk probability of the target user within a preset rehabilitation period; evaluating the initial postoperative care scheme based on baseline state detection information of the target user to generate a care scheme strength and a baseline state score; formulating an adaptive rehabilitation intervention rhythm and an adaptive risk monitoring mechanism according to the predicted recurrence and metastasis risk probability, the care scheme strength and the baseline state score; and performing care on the target user according to the initial postoperative care scheme and dynamically optimizing the postoperative care scheme according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism. The application realizes individualized and accurate adaptation of the postoperative care scheme for liver cancer.

Description

Technical Field

[0001] This invention relates to the field of postoperative care technology for liver cancer, specifically to a method for optimizing postoperative care for liver cancer and a smart chip. Background Technology

[0002] With the rapid development of precision medicine and immunotherapy, postoperative management of liver cancer has shifted from single surgical intervention to multidisciplinary comprehensive rehabilitation. Current technologies increasingly emphasize the crucial role of the tumor immune microenvironment in prognostic assessment and are beginning to explore combining molecular subtyping with postoperative adjuvant therapy. Meanwhile, recurrence risk prediction models based on clinical indicators and standardized postoperative follow-up protocols have become important supports for current postoperative management.

[0003] However, treatment plans tend to be homogeneous, relying heavily on standardized clinical staging and guidelines, neglecting the high heterogeneity within tumors, especially the key differences in the immune microenvironment, and failing to provide targeted interventions for tumors with different immune characteristics. Risk prediction and monitoring models are crude, relying solely on single clinicopathological features to assess recurrence risk, lacking precise models that integrate multi-dimensional biological information, and the follow-up procedures are fixed, making it difficult to match the patient's real-time dynamic risk. Nursing care plans are static and rigid, unable to be dynamically adjusted according to the patient's treatment response, tolerance, and changes in physical condition, making it difficult to carry out individualized stepwise treatment management. Summary of the Invention

[0004] This invention provides a method for optimizing postoperative care for liver cancer and an intelligent chip, aiming to solve the technical problem that existing technologies cannot achieve individualized and precise adaptation of postoperative care for liver cancer.

[0005] In view of the above problems, the present invention provides a method for optimizing the postoperative care plan for liver cancer, comprising: Multidimensional detection datasets are obtained by performing multidimensional detection on liver cancer tumor tissue specimens of target users. Based on the multidimensional detection datasets, the tumor immune microenvironment type is determined, and an initial postoperative care plan is generated. Based on the analysis of the multi-dimensional detection dataset and the initial postoperative care plan, the predicted recurrence and metastasis risk probability of the target user within the preset recovery period is obtained. The initial postoperative care plan is evaluated based on the baseline status detection information of the target user to generate the strength of the care plan, and a baseline status score is generated based on the baseline status detection information. Based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism are developed; Postoperative care was provided to the target user according to the initial postoperative care plan, and the postoperative care plan was dynamically optimized according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism.

[0006] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention provides an intelligent chip for optimizing postoperative care for liver cancer. First, it performs multi-dimensional detection of liver cancer tumor tissue to accurately classify the immune microenvironment, generating a personalized initial care plan, breaking the limitations of traditional homogeneous plans. Then, it combines the detection data with the initial plan to accurately predict the risk of recurrence and metastasis. Simultaneously, it generates plan strength and scores based on baseline status, providing a scientific basis for adaptive adjustments. Furthermore, it establishes an appropriate rehabilitation intervention rhythm and monitoring mechanism to address the problem of inadequate risk monitoring. Finally, through dynamic optimization, it overcomes the static rigidity of care plans, achieving personalized and precise management throughout the entire postoperative period for liver cancer, improving the accuracy and effectiveness of care, reducing the risk of recurrence and metastasis, and improving the long-term quality of life for patients. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a flowchart illustrating an optimization method for postoperative care of liver cancer provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an intelligent chip for optimizing postoperative care for liver cancer, provided in an embodiment of the present invention. The components represented by each number in the attached diagram are explained below: Initial care plan generation module 11, relapse risk probability prediction module 12, baseline evaluation module 13, intervention rhythm and monitoring mechanism formulation module 14, dynamic care and plan optimization module 15. Detailed Implementation

[0009] This invention provides a method for optimizing postoperative care for liver cancer and a smart chip, which addresses the technical problem that existing technologies struggle to achieve individualized and precise adaptation of postoperative care plans for liver cancer.

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0011] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0012] Examples, such as Figure 1 As shown, the present invention provides a method for optimizing postoperative care for liver cancer, the method comprising: S100: Multidimensional detection is performed on liver cancer tumor tissue specimens from the target user to obtain a multidimensional detection dataset. Based on this dataset, the tumor immune microenvironment type is determined, and an initial postoperative care plan is generated. The composition and spatial distribution of immune cells, tumor cells, and stromal cells in the tumor tissues of different liver cancer patients vary significantly. The expression and distribution of key signaling molecules also differ, directly determining the tumor's invasiveness, sensitivity to treatment, and postoperative recurrence risk. Traditional intervention strategies are based solely on clinical staging, failing to match the individualized needs of patients with different immune microenvironment characteristics. Therefore, this step, through multidimensional detection of liver cancer tumor tissue specimens, accurately defines the tumor immune microenvironment type, providing a core basis for generating a personalized initial care plan, thus breaking the limitation of homogeneous care plans from the source.

[0013] Step S100 in the method provided in this embodiment of the invention includes: According to the preset detection process, the surgically removed liver cancer tumor tissue specimens of the target user are subjected to multi-dimensional detection to obtain a multi-dimensional detection dataset. The multi-dimensional detection dataset includes immune cell composition, immune cell spatial distribution, tumor cell composition, tumor cell spatial distribution, stromal cell composition, stromal cell spatial distribution, key signal molecule composition, and key signal molecule spatial distribution. The multidimensional detection dataset is compared with several standard detection datasets in the pre-constructed detection data-immune microenvironment type comparison library. The immune microenvironment type corresponding to the standard detection dataset with the highest comprehensive similarity is selected as the tumor immune microenvironment type. An initial postoperative care plan is obtained based on the matching of the tumor immune microenvironment type. The postoperative care plan includes medical intervention, metabolic and nutritional support, physiological function rehabilitation and psychoneuroimmune regulation.

[0014] The immune microenvironment types include the immune-inflammatory type-functional activation subtype, the immune-inflammatory type-functional depletion subtype, the immune-rejection type-matrix barrier subtype, the immune-rejection type-vascular abnormality subtype, the immune-desert type-immune neglect subtype, and the immune-desert type-active inhibition subtype.

[0015] First, following a pre-defined detection procedure, the surgically removed liver cancer tumor tissue specimens of the target user undergo multi-dimensional detection to obtain a multi-dimensional detection dataset. This dataset includes data on immune cell composition, spatial distribution of immune cells, tumor cell composition, spatial distribution of tumor cells, stromal cell composition, spatial distribution of stromal cells, key signaling molecule composition, and spatial distribution of key signaling molecules. The pre-defined detection procedure refers to a standardized, pre-established detection process. The multi-dimensional detection dataset is a complete set of data obtained through multi-dimensional detection, reflecting the composition and spatial distribution of various cells and key signaling molecules in the tumor tissue. Specifically, it includes eight dimensions: immune cell composition, immune cell spatial distribution, tumor cell composition, spatial distribution of tumor cells, stromal cell composition, spatial distribution of stromal cells, key signaling molecule composition, and spatial distribution of key signaling molecules. Following the pre-defined detection procedure, the surgically removed liver cancer tumor tissue specimens of the target user undergo full-dimensional detection, collecting data from the above eight dimensions to form the multi-dimensional detection dataset.

[0016] For example, a tumor tissue specimen from a 45-year-old male patient was examined: Immune cell composition: CD8 was detected by flow cytometry. + T cells accounted for 12%, CD4 + T cells accounted for 8%, macrophages for 15%, and NK cells for 3%, quantified as a 4-dimensional vector: [0.12, 0.08, 0.15, 0.03]. Spatial distribution of immune cells: CD8+ was obtained through spatial transcriptome sequencing. + T cells mainly accumulate in the periphery of the tumor parenchyma, with sparse distribution in the tumor center. Based on CD8 in the periphery... + T cells account for a significant portion of total CD8+. +The proportion of T cells was quantified as a 1D vector: [0.85]. Tumor cell composition: Pathological section analysis showed that highly differentiated hepatocellular carcinoma cells accounted for 70%, moderately differentiated cells for 30%, and no poorly differentiated cells. This was quantified as a 3D vector based on high, moderate, and low differentiation: [0.7, 0.3, 0.0]. Spatial distribution of tumor cells: Immunohistochemical localization showed that tumor cells were distributed in nests with tightly packed cells within the nests. The density score based on the nest structure was quantified as a 1D vector: [0.9]. Stromal cell composition: Flow cytometry analysis showed that fibroblasts accounted for 25% and endothelial cells for 10%. This was quantified as a 2D vector based on fibroblasts and endothelial cells: [0.25, 0.10]. Spatial distribution of stromal cells: Spatial transcriptome sequencing showed that fibroblasts were mainly distributed at the junction of tumor parenchyma and normal tissue, forming a thin barrier. The integrity score based on the barrier was quantified as a 1D vector: [0.7]. Key signaling molecule composition: HPLC-MS / MS analysis revealed PD-1 expression levels of 2.5 ng / mL, PD-L1 expression levels of 3.1 ng / mL, and VEGF expression levels of 1.8 ng / mL, quantified into a 3D vector [2.5, 3.1, 1.8]. Spatial distribution of key signaling molecules: Immunohistochemical analysis showed that PD-1 and PD-L1 were mainly located at the periphery of the tumor parenchyma, specifically on CD8+. + T cell surface expression and VEGF expression primarily around endothelial cells are analyzed. The proportions of each signaling molecule expressed on the target cell surface relative to the total expression are quantified into a 3-dimensional vector: [0.88, 0.90, 0.92]. These eight data types are sequentially concatenated and integrated to form the patient's multi-dimensional detection dataset, with a total dimension of 4+1+3+1+2+1+3+3=18 dimensions.

[0017] Secondly, the multi-dimensional detection dataset was compared with several standard detection datasets in the pre-constructed detection data-immune microenvironment type control library. The immune microenvironment type corresponding to the standard detection dataset with the highest comprehensive similarity was selected as the tumor immune microenvironment type. The immune microenvironment types include: immune-inflammatory type - functional activation subtype, immune-inflammatory type - functional exhaustion subtype, immune-rejection type - matrix barrier subtype, immune-rejection type - vascular abnormality subtype, immune-desert type - immune neglect subtype, and immune-desert type - active inhibition subtype. Under the immune-inflammatory type, the key distinguishing features of the functional activation type are high expression of granzyme B and interferon-γ, and mature tertiary lymphoid structures; the key distinguishing features of the functional exhaustion type are high expression of inhibitory receptors such as PD-1, TIM-3, and LAG-3, and T cell inactivation. Under the immune-rejection type, the key distinguishing features of the matrix barrier type are enrichment of cancer-associated fibroblasts and dense collagen deposition; the key distinguishing features of the vascular abnormality type are disordered vascular structure and high expression of rejection molecules by endothelial cells. In the immune desert type, the key distinguishing feature of the immune neglect type is low tumor mutation burden and defective antigen presentation mechanism; the key distinguishing feature of the active suppression type is rich in suppressive cells such as M2 tumor-associated macrophages and myeloid suppressor cells. The detection data – immune microenvironment type reference library – refers to a pre-constructed standardized database containing standard detection datasets corresponding to six types of tumor immune microenvironment. Each standard dataset includes the standard reference range and distribution characteristics of the aforementioned eight dimensions. The comprehensive similarity refers to the weighted summation of the similarity results of each dimension after comparing the target user's multi-dimensional detection dataset with the eight dimensions of a standard detection dataset in the reference library. The weights are preset as follows: immune cell-related dimension weight 0.3, tumor cell-related dimension weight 0.2, stromal cell-related dimension weight 0.2, and key signal molecule-related dimension weight 0.3. The composition and spatial distribution sub-dimensions under each dimension each account for 50% of the weight of that dimension. Similarity traversal comparison refers to the process of comparing the target user's dataset with all standard datasets in the reference library one by one across all eight dimensions.

[0018] All standard detection datasets were extracted from the detection data-immune microenvironment type reference library. Following the order of immune cell composition, immune cell spatial distribution, tumor cell composition, tumor cell spatial distribution, stromal cell composition, stromal cell spatial distribution, key signaling molecule composition, and key signaling molecule spatial distribution, the target user dataset was compared with each standard dataset along its corresponding dimension using a cosine similarity algorithm. Similarity values ​​ranged from 0 to 1, with values ​​closer to 1 indicating a higher match. The similarity results across the eight dimensions were weighted and summed according to preset weights to obtain the comprehensive similarity between the target user dataset and the corresponding standard dataset. After traversing all standard datasets, the immune microenvironment type corresponding to the standard dataset with the highest comprehensive similarity was selected as the tumor immune microenvironment type for the target user.

[0019] For example, standard datasets for six immune microenvironment types are extracted from a pre-constructed detection data-immune microenvironment type control library. Taking the immune inflammation-functional exhaustion subtype standard dataset as an example, its eight-dimensional standard range is: CD8 + T cells account for 10%-15%, CD8 + T cells were mainly distributed at the tumor periphery; well-differentiated cancer cells accounted for 60%-80%; tumor cells were distributed in nests; fibroblasts accounted for 20%-30%; fibroblasts were distributed at the tumor periphery; PD-1 expression level was 2.0-3.0 ng / mL; PD-1 was mainly found at the tumor periphery. + T cell surface expression. The target user dataset was compared sequentially with this standard dataset: immune cell composition similarity 0.95, immune cell spatial distribution similarity 0.92, tumor cell composition similarity 0.98, tumor cell spatial distribution similarity 0.96, stromal cell composition similarity 0.93, stromal cell spatial distribution similarity 0.94, key signaling molecule composition similarity 0.97, and key signaling molecule spatial distribution similarity 0.95. The weighted summation of the overall similarity is: 0.95×0.15+0.92×0.15+0.98×0.1+0.96×0.1+0.93×0.1+0.94×0.1+0.97×0.15+0.95×0.15=0.9495. After traversing the remaining five standard datasets, the overall similarity scores were calculated as follows: immune-inflammatory type - functional activation subtype 0.42, immune-rejection type - matrix barrier subtype 0.38, immune-rejection type - vascular abnormality subtype 0.35, immune-desert type - immune-neglect subtype 0.29, and immune-desert type - active inhibition subtype 0.31. Since the immune-inflammatory type - functional exhaustion subtype had the highest overall similarity, the tumor immune microenvironment type of this patient was determined to be the immune-inflammatory type - functional exhaustion subtype.

[0020] Finally, an initial postoperative care plan is obtained based on the tumor immune microenvironment type matching. This plan includes medical intervention, metabolic and nutritional support, physiological function rehabilitation, and psychoneuroimmune regulation. The initial postoperative care plan is a basic care plan pre-designed based on the tumor immune microenvironment type, encompassing multi-dimensional intervention measures and including four categories: medical intervention, metabolic and nutritional support, physiological function rehabilitation, and psychoneuroimmune regulation. Medical intervention includes targeted drug therapy and minimally invasive intervention; metabolic and nutritional support includes personalized diet and nutritional supplements; physiological function rehabilitation includes exercise training and organ function protection; and psychoneuroimmune regulation includes emotional intervention and stress management. A pre-constructed database of the correspondence between immune microenvironment type and initial care plan is used. Different types correspond to differentiated care measures. For example, for the immune-inflammatory type-functional exhaustion subtype, the focus is on strengthening immune activation-related interventions. Based on the tumor immune microenvironment type determined in the above steps, a suitable care plan is extracted from the database as the initial postoperative care plan for the target user.

[0021] For example, based on a pre-defined correspondence database, the initial postoperative care plan for the immune-inflammatory type-functional exhaustion subtype is as follows: Medical intervention: PD-1 inhibitors, 3 mg / kg, are started in the first week postoperatively, every two weeks, with monthly immune function checks. This is quantified into a 3-dimensional vector based on PD-1 inhibitor dosage, dosing interval, and frequency of immune function checks: [3.0, 14, 1]. Metabolic and nutritional support: A high-protein diet, with a daily protein intake of 1.8 g / kg, supplemented with Omega-3 fatty acids, avoiding high-sugar and high-fat foods. This is quantified into a 3-dimensional vector based on daily protein intake, Omega-3 supplementation, and dietary restrictions (yes = 1 / no = 0): [1.8, 1.0, 1]. Physiological rehabilitation: Mild aerobic exercise begins in the second week postoperatively, starting with 30 minutes of slow walking daily, gradually increasing to 45 minutes of brisk walking daily, five times a week. The postoperative exercise initiation time, daily basic exercise duration, duration after exercise intensity upgrade, and weekly exercise frequency are quantified into a 4-dimensional vector: [14, 30, 45, 5]. Psychological neuroimmunological regulation: Cognitive behavioral therapy once a week, and 15 minutes of mindfulness meditation daily. This is quantified into a fixed 2-dimensional vector based on the weekly Cognitive Behavioral Therapy frequency and daily mindfulness meditation duration: [1, 15]. Based on this correspondence, the patient's initial postoperative care plan is directly matched, and the vectors of the above four modules are sequentially concatenated and integrated to form the patient's care plan feature vector, with a total dimension of 3+3+4+2=12 dimensions.

[0022] In this embodiment of the invention, precise subtyping of the tumor immune microenvironment is achieved through standardized multi-dimensional detection, overcoming the limitations of traditional methods that rely solely on coarse classification based on clinical staging. Comprehensive multi-dimensional detection and orderly comparison ensure the accuracy of immune microenvironment subtyping, providing core targets for personalized care. The initial postoperative care plan based on subtyping matching specifically covers key modules such as medical intervention and nutritional support, directly addressing the homogenization problem of existing plans. For example, it strengthens the use of PD-1 inhibitors for immune-depleted subtypes, differentiating it from intervention strategies for immune-activated subtypes. Simultaneously, the multi-dimensional detection dataset and initial care plan generated in this step provide fundamental data support for subsequent steps, including recurrence and metastasis risk prediction and dynamic optimization of the care plan.

[0023] S200: Based on the multi-dimensional detection dataset and the initial postoperative care plan, analyze and obtain the predicted recurrence and metastasis risk probability of the target user within the preset recovery period.

[0024] In this embodiment of the invention, the predicted recurrence and metastasis risk probability of the target user within a preset recovery period is obtained by analyzing the multi-dimensional detection dataset and the initial postoperative care plan. Existing liver cancer postoperative recurrence and metastasis risk assessments largely rely on single clinicopathological features such as tumor size and differentiation degree, failing to incorporate the multi-dimensional biological information of the tumor immune microenvironment and the synergistic influence of the postoperative care plan. This results in insufficient accuracy in risk prediction and cannot provide precise evidence for individualized care strategies. This step integrates multi-dimensional detection data and care plans from historical users to train an accurate risk prediction model, outputting the recurrence and metastasis risk probability of the target user within a preset recovery period. This addresses the problem of the coarse nature of existing risk prediction models and provides quantitative indicators for the adaptation and optimization of subsequent care plans.

[0025] Step S200 in the method provided in this embodiment of the invention includes: Based on the postoperative care records of historical users with liver cancer, several multidimensional detection datasets and several postoperative care plans were collected. With the time span of the preset recovery cycle as a constraint, the proportion of recurrence and metastasis risk events of historical users under different multidimensional detection datasets and postoperative care plans was statistically analyzed as the sample recurrence and metastasis risk probability, and the recurrence and metastasis risk probabilities of several samples were obtained. Using the multi-dimensional detection dataset of several samples and the postoperative care plan of several samples as input data, and using the recurrence and metastasis risk probability of several samples as supervision label, a deep learning model is trained until convergence to generate a recurrence and metastasis risk prediction model. The multi-dimensional detection dataset and the initial postoperative care plan are input into the recurrence and metastasis risk prediction model, which outputs the predicted recurrence and metastasis risk probability of the target user within a preset recovery period.

[0026] First, based on historical users' post-hepatitis B (HBC) cancer post-operative care records, several multi-dimensional detection datasets and several post-operative care plans were collected. Constrained by the preset recovery period, the proportion of recurrence and metastasis risk events among historical users under different multi-dimensional detection datasets and post-operative care plans was statistically analyzed as the sample recurrence and metastasis risk probability, resulting in several sample recurrence and metastasis risk probabilities. Historical users' HBC cancer post-operative care records refer to the complete records of multi-dimensional tumor tissue detection data, post-operative care plans, and recurrence and metastasis events within the recovery period for HBC patients who have completed the preset recovery period. The preset recovery period refers to a pre-defined risk assessment time span, such as 12 months. The sample recurrence and metastasis risk probability refers to the proportion of historical users corresponding to a certain type of multi-dimensional detection dataset and post-operative care plan who experienced recurrence and metastasis events within the preset recovery period.

[0027] For example, from the postoperative care records of liver cancer patients in historical users, cases matching the tumor type of the target user are selected, and corresponding multi-dimensional detection datasets and postoperative care plans for the samples are collected. With a preset recovery period as the time constraint, the number of recurrence and metastasis events under each type of sample dataset and sample care plan is counted, and the event proportion is calculated as the recurrence and metastasis risk probability of that type of sample. For example, the preset recovery period is set to 12 months, and the historical care records of 1000 patients after hepatocellular carcinoma surgery are collected: 200 samples matching the immune-inflammatory-functional exhaustion subtype in the multi-dimensional detection dataset and whose postoperative care plan is PD-1 inhibitor, high-protein diet, and light aerobic exercise are selected; among these 200 samples, 70 users have experienced recurrence and metastasis events within 12 months; the recurrence and metastasis risk probability of the sample = 70 / 200 = 35%, that is, the recurrence and metastasis risk probability corresponding to this type of sample is 35%, which is quantified as a 1-dimensional vector: [0.35].

[0028] Secondly, using the aforementioned multi-dimensional detection datasets and postoperative care plans for several samples as input data, and the recurrence and metastasis risk probabilities of the aforementioned samples as supervision labels, a deep learning model is trained until convergence, generating a recurrence and metastasis risk prediction model. Input data refers to the feature vector obtained by integrating the multi-dimensional detection datasets and postoperative care plans for the samples. Supervision labels refer to the recurrence and metastasis risk probabilities of the samples obtained in the previous steps. The deep learning model employs a multi-feature fusion neural network model, including a feature encoding layer, a convolutional feature extraction layer, and a fully connected output layer. Training to convergence means that the model's loss function stabilizes below a preset threshold, such as a mean squared error ≤ 0.01, and the prediction error meets clinical accuracy requirements. The integrated features of the multi-dimensional detection datasets and postoperative care plans for several samples are used as input data, with the corresponding recurrence and metastasis risk probabilities of the samples used as supervision labels. This data is input into the deep learning model for iterative training until the model's loss function stabilizes and the prediction accuracy meets the target, generating a recurrence and metastasis risk prediction model.

[0029] For example, using 1000 historical samples as the training set, the 18-dimensional multi-dimensional detection data features and 12-dimensional care plan features of each sample are integrated into a 30-dimensional feature vector as the model input data, with the corresponding 1-dimensional sample relapse and metastasis risk probability as the supervision label. A multi-feature fusion neural network model is employed. First, the 30-dimensional input feature vector is standardized through a feature encoding layer. Then, two convolutional layers with a kernel size of 3×3 are used to extract multi-dimensional feature association patterns. Finally, a fully connected layer containing 64 neurons maps the output probability value. With a learning rate of 0.001, after 50 iterations of training, the model's mean squared error decreases to 0.008 ≤ the preset convergence threshold of 0.01, indicating training convergence and the generation of a relapse and metastasis risk prediction model. The trained relapse and metastasis risk prediction model can receive the 30-dimensional feature vector (integrated from the target user's 18-dimensional multi-dimensional detection data features and 12-dimensional care plan features) as input and output a 1-dimensional predicted relapse and metastasis risk probability.

[0030] Finally, the multi-dimensional detection dataset and the initial postoperative care plan are input into the recurrence and metastasis risk prediction model, which outputs the predicted recurrence and metastasis risk probability of the target user within a preset recovery period. The predicted recurrence and metastasis risk probability refers to the probability value output by the model that the target user will experience a recurrence and metastasis event within the preset recovery period. The multi-dimensional detection dataset and the initial postoperative care plan of the target user are integrated into a feature vector, which is then input into the recurrence and metastasis risk prediction model to directly obtain the output probability value, i.e., the predicted recurrence and metastasis risk probability of the target user. For example, integrating the multi-dimensional detection dataset and the initial postoperative care plan of the 45-year-old patient in S100 into a feature vector and inputting it into the trained recurrence and metastasis risk prediction model results in an output of 35%, meaning that the predicted recurrence and metastasis risk probability of this user within a 12-month preset recovery period is 35%.

[0031] In this embodiment of the invention, the process of collecting historical sample data, training the model, and predicting the target user solves the problem of the crudeness of existing relapse and metastasis risk assessment: the deep learning model, which takes a multi-dimensional detection dataset and a care plan as input, breaks through the limitations of a single clinical feature and improves the accuracy of risk prediction; the output quantitative risk probability provides a basis for the subsequent formulation of adaptive rehabilitation intervention rhythm and risk monitoring mechanism, realizing the transformation of risk assessment from qualitative judgment to quantitative prediction, and laying a data foundation for the dynamic optimization of individualized care.

[0032] S300: Based on the baseline status detection information of the target user, evaluate the initial postoperative care plan to generate the care plan strength, and generate a baseline status score based on the baseline status detection information.

[0033] In this embodiment of the invention, the initial postoperative care plan is evaluated based on the baseline status detection information of the target user to generate the strength of the care plan, and a baseline status score is generated based on the baseline status detection information. Existing postoperative care plans for liver cancer do not consider the individual patient's baseline physiological state, but only formulate plans based on tumor characteristics. This easily leads to a mismatch between the strength of the plan and the patient's tolerance; for example, some patients cannot tolerate the plan due to weak organ function, while others have sufficient physical strength but receive insufficient treatment. This step, by obtaining the baseline status detection information of the target user, evaluates the suitability of the initial care plan from multiple dimensions and generates a baseline score of the patient's overall status, overcoming the limitation of existing plans that prioritize the tumor over the patient, and providing a quantitative basis at the patient level for subsequent optimization of the care plan.

[0034] Step S300 in the method provided in this embodiment of the invention includes: Obtain baseline status detection information of the target user, wherein the baseline status detection information includes basic physiological information, organ function indicators, physical and nutritional status, tumor burden baseline, and infection and immune status; According to the preset evaluation dimensions, the initial postoperative care plan is evaluated in multiple dimensions based on the baseline status detection information to obtain the strength of the care plan. The preset evaluation dimensions include organ function tolerance, physical strength, comorbid diseases and immune background. A comprehensive status evaluation of the target user is performed based on the baseline status detection information, and a baseline status score is generated.

[0035] First, baseline status detection information for the target users is acquired. This baseline status detection information includes basic physiological information, organ function indicators, physical and nutritional status, tumor burden baseline, and infection and immune status. Baseline status detection information refers to the basic physiological and health status data of liver cancer patients before the initiation of the care plan after surgery. Specifically, it includes five categories of information: Basic physiological information: patient's age, gender, BMI, and other basic physical characteristics; Organ function indicators: functional test results of key organs such as the liver, kidneys, and heart; Physical and nutritional status: indicators reflecting physical reserves and nutritional levels such as grip strength and serum albumin; Tumor burden baseline: postoperative circulating tumor ctDNA levels, alpha-fetoprotein, and other tumor marker results; Infection and immune status: hepatitis virus load, autoimmune history, peripheral blood immune cell typing, C-reactive protein, and other inflammation and immune-related indicators. Through postoperative clinical testing, laboratory examinations, and medical history collection, the above five categories of baseline status detection information for the target users are collected to form a complete baseline information set.

[0036] For example, baseline testing was performed on this 45-year-old male patient, and the following information was obtained: Basic physiological information: age 45 years, male, BMI 22.3; Organ function indicators: liver function ALT 25 U / L (normal), AST 20 U / L (normal), kidney function creatinine 70 μmol / L (normal); Physical and nutritional status: grip strength 28 kg (normal range), serum albumin 40 g / L (normal); Baseline tumor burden: postoperative ctDNA test negative, alpha-fetoprotein (AFP) 12 ng / mL (slightly higher than the upper limit of normal); Infection and immune status: hepatitis B virus (HBV) load below the detection limit, no history of autoimmune diseases, peripheral blood CD8 + T cell percentage 10%, C-reactive protein 5 mg / L (normal).

[0037] Secondly, based on the baseline status detection information, the initial postoperative care plan is evaluated from multiple dimensions according to preset evaluation dimensions to obtain the strength of the care plan. These preset evaluation dimensions include organ function tolerance, physical strength, and comorbidities and immune background. The preset evaluation dimensions include: organ function tolerance: assessing the metabolic tolerance of the patient's key organs to the initial care plan; physical strength: assessing the adaptability of the patient's overall physiological reserves to the rehabilitation training and nutritional programs within the care plan; and comorbidities and immune background: assessing the specific risks and adaptability of the patient's underlying diseases and immune status to the care plan. The strength of the care plan refers to the quantitative value of the adaptability between the initial care plan and the patient's condition, obtained by weighted summation of the evaluation scores based on the above three dimensions. The value ranges from 0 to 10, with higher scores indicating a better fit between the plan and the patient's condition. For each preset evaluation dimension, the initial care plan is scored from 0 to 10 based on the baseline status detection information, and the average of the scores for the three dimensions is then taken to obtain the strength of the care plan.

[0038] For example, the initial care plan for this patient was evaluated according to the preset dimensions: Organ function tolerance: The patient's liver and kidney functions are normal and can tolerate the metabolism of PD-1 inhibitors, score 8; Physical strength: Grip strength and nutritional status are normal and can be adapted to light aerobic exercise, score 7; Coexisting diseases and immune background: No underlying diseases, immune cells are depleted but the load is stable and can tolerate immunosuppressive drugs, score 7; Care plan strength = (8+7+7) / 3≈7.33, the suitability strength is moderate to high.

[0039] Finally, a comprehensive status evaluation of the target user is performed based on the baseline status detection information to generate a baseline status score. The baseline status score is a quantitative value of the patient's initial postoperative health status obtained after comprehensively evaluating five categories of baseline status detection information. The score ranges from 0 to 10, with higher scores indicating better initial health. Each of the five categories of baseline status detection information is scored from 0 to 10, and the scores are weighted and summed with a weight of 0.2 for each category to obtain the baseline status score. For example, the scores for this patient's five categories of baseline information are as follows: Basic physiological information: moderate age, normal BMI, score 9; Organ function indicators: normal liver and kidney function, score 9; Physical and nutritional status: normal grip strength and albumin, score 8; Tumor burden baseline: negative ctDNA, slightly elevated AFP, score 9; Infection and immune status: low viral load, normal inflammatory markers, score 8. The baseline status score = 9 × 0.2 + 9 × 0.2 + 8 × 0.2 + 9 × 0.2 + 8 × 0.2 = 8.6, indicating that the patient's initial health status is good.

[0040] In this embodiment of the invention, the degree of fit between the initial care plan and the patient is quantified based on the evaluation of the patient's organ function, physical fitness and other dimensions, avoiding the situation where the intensity of the plan does not match the patient's tolerance; the quantified baseline status score clearly presents the patient's initial health level, providing a patient-level quantitative basis for the formulation of subsequent adaptive rehabilitation intervention rhythm and risk monitoring mechanisms, making the optimization of the care plan more in line with the individual's actual condition.

[0041] S400: Develop an appropriate rehabilitation intervention rhythm and risk monitoring mechanism based on the predicted relapse and metastasis risk probability, the intensity of the care plan, and the baseline status score.

[0042] In this embodiment of the invention, an appropriate rehabilitation intervention rhythm and risk monitoring mechanism are established based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score. Existing postoperative rehabilitation intervention rhythms and risk monitoring mechanisms for liver cancer are characterized by a fixed pattern, with uniformly set intervention feedback adjustment cycles and the number of monitoring items. They fail to dynamically adapt to the patient's predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score. This results in excessively long intervention intervals and insufficient monitoring items for high-risk patients, making timely adjustments impossible; while low-risk patients endure excessively short intervention intervals and too many monitoring items, increasing medical burden and psychological anxiety. This step constructs a postoperative care sensitivity compensation index to quantify the patient's comprehensive risk status, thereby dynamically optimizing and generating an appropriate rehabilitation intervention rhythm and risk monitoring mechanism. This addresses the problems of fixed mechanisms and lack of individualized adaptation in existing mechanisms, providing an execution framework for the dynamic optimization of care plans.

[0043] Step S400 in the method provided in this embodiment of the invention includes: Based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, postoperative care sensitivity is identified, and the postoperative care sensitivity compensation index is obtained. Based on the postoperative care sensitivity compensation index, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted to generate an appropriate rehabilitation intervention rhythm and appropriate risk monitoring mechanism.

[0044] First, postoperative care sensitivity is identified based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, and the postoperative care sensitivity compensation index is obtained.

[0045] Specifically, postoperative care sensitivity is identified based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, and a postoperative care sensitivity compensation index is obtained, including: The ratio of the predicted relapse and metastasis risk probability to the proportion of historical relapse and metastasis risk events recorded in the historical time zone is used as the first care sensitivity coefficient. The ratio of the intensity of the described care plan to the average intensity of historical care plans recorded in the historical time zone is used as the second care sensitivity coefficient. The ratio of the average historical baseline status score of historical users recorded within the historical time zone to the baseline status score is used as the third care sensitivity coefficient. The first, second, and third care sensitivity coefficients are weighted and fused to generate the postoperative care sensitivity compensation index.

[0046] First, the ratio of the predicted recurrence and metastasis risk probability to the proportion of historical recurrence and metastasis risk events recorded within the historical time zone is used as the first care sensitivity coefficient. The first care sensitivity coefficient quantifies the target user's recurrence risk relative to the historical average; a higher ratio indicates a higher recurrence risk. The historical time zone refers to the follow-up period of historical users matching the target user's clinical characteristics, for example, the past 5 years. The proportion of historical recurrence and metastasis risk events refers to the average proportion of recurrence and metastasis risk events among all liver cancer postoperative patients matching clinical characteristics within the historical time zone during a preset recovery period. The first care sensitivity coefficient is obtained by obtaining the proportion of historical recurrence and metastasis risk events recorded within the historical time zone and dividing the target user's predicted recurrence and metastasis risk probability by this historical proportion.

[0047] For example, with a preset recovery period of 12 months, the proportion of postoperative liver cancer patients with clinical characteristics of hepatocellular carcinoma and immune-inflammatory-functional exhaustion subtype matched within the past 5 years in the historical time zone is 20%; the predicted recurrence and metastasis risk probability of the target user is 35%; the first care sensitivity coefficient = 35% / 20% = 1.75.

[0048] Secondly, the ratio of the intensity of the described care plan to the average intensity of historical care plans recorded within the historical time zone is used as the second care sensitivity coefficient. The second care sensitivity coefficient refers to the ratio of the target user's care plan intensity to the average intensity of historical care plans recorded within the historical time zone. It is used to quantify the target user's care plan fit relative to the historical average level; the larger the ratio, the higher the fit. The average intensity of historical care plans refers to the average intensity of care plans for all postoperative liver cancer patients matching clinical characteristics within the historical time zone. The second care sensitivity coefficient is obtained by obtaining the average intensity of historical care plans recorded within the historical time zone and dividing the target user's care plan intensity by this historical average.

[0049] For example, among postoperative liver cancer patients in the past 5 years who matched the clinical characteristics of hepatocellular carcinoma and the immune-inflammatory-functional exhaustion subtype within the historical time zone, the mean historical care intensity was 6.0; the care intensity of the target user was 7.33; the second care sensitivity coefficient = 7.33 / 6.0 ≈ 1.22.

[0050] Furthermore, the ratio of the mean historical baseline status score of historical users recorded within the historical time zone to the baseline status score is used as the third care sensitivity coefficient. The third care sensitivity coefficient is the ratio of the mean historical baseline status score of historical users recorded within the historical time zone to the baseline status score of the target user. It is used to quantify the target user's baseline status relative to the historical average level; the larger the ratio, the worse the target user's baseline status. The mean historical baseline status score refers to the average baseline status score of all postoperative liver cancer patients matching clinical characteristics within the historical time zone. The mean historical baseline status score recorded within the historical time zone is obtained, and this historical mean is divided by the target user's baseline status score to obtain the third care sensitivity coefficient.

[0051] For example, the mean historical baseline status score of postoperative liver cancer patients matched with clinical characteristics of hepatocellular carcinoma and immune-inflammatory-functional exhaustion subtype within the past 5 years in the historical time zone was 7.5; the baseline status score of the target user was 8.6; the third care sensitivity coefficient = 7.5 / 8.6≈0.87.

[0052] Based on this, the first, second, and third care sensitivity coefficients are weighted and fused to generate a postoperative care sensitivity compensation index. Weighted fusion refers to summing the three care sensitivity coefficients according to preset weights: 0.5 for the first care sensitivity coefficient, 0.25 for the second, and 0.25 for the third, with a total weight of 1. The postoperative care sensitivity compensation index is the quantified value obtained by weighting and fusing the three care sensitivity coefficients. It is used to comprehensively characterize the postoperative care sensitivity and risk status of the target user. The higher the index, the greater the risk and the worse the patient's recovery. The first, second, and third care sensitivity coefficients are weighted and summed according to preset weights to obtain the postoperative care sensitivity compensation index. For example, the postoperative care sensitivity compensation index = 1.75 × 0.5 + 1.22 × 0.25 + 0.87 × 0.25 = 1.3975.

[0053] Secondly, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted based on the postoperative care sensitivity compensation index to generate an appropriate rehabilitation intervention rhythm and appropriate risk monitoring mechanism.

[0054] Specifically, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted based on the postoperative care sensitivity compensation index to generate an adapted rehabilitation intervention rhythm and an adapted risk monitoring mechanism, including: The reciprocal of the postoperative care sensitivity compensation index is used as the rhythm compensation coefficient, and the product of the rhythm compensation coefficient and the initial rehabilitation intervention rhythm is used as the adaptive rehabilitation intervention rhythm, wherein the initial rehabilitation intervention rhythm is the duration of the initial feedback adjustment cycle. An initial risk monitoring mechanism is obtained, wherein the initial risk monitoring mechanism is the number of initial monitoring items, which is half of the total number of preset monitoring items. The monitoring items include at least circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, immune checkpoint molecule expression, cytokine profile, liver function, kidney function, bone marrow function, treatment-specific toxicity markers, and physical status. The product of the postoperative care sensitivity compensation index and the initial number of monitoring items is rounded down to generate the number of adaptive monitoring items, which serves as the adaptive risk monitoring mechanism.

[0055] First, the reciprocal of the postoperative care sensitivity compensation index is used as the rhythm compensation coefficient. The product of the rhythm compensation coefficient and the initial rehabilitation intervention rhythm is used as the adaptive rehabilitation intervention rhythm, where the initial rehabilitation intervention rhythm is the duration of the initial feedback adjustment cycle. The rhythm compensation coefficient, which is the reciprocal of the postoperative care sensitivity compensation index, is used to adjust the duration of the initial rehabilitation intervention rhythm. The larger the index, the smaller the reciprocal, and the shorter the rhythm duration. The initial rehabilitation intervention rhythm refers to the pre-set duration of the initial feedback adjustment cycle, for example, 4 weeks. The adaptive rehabilitation intervention rhythm is the product of the rhythm compensation coefficient and the initial rehabilitation intervention rhythm, i.e., the personalized feedback adjustment cycle duration for the target user. The reciprocal of the postoperative care sensitivity compensation index is calculated to obtain the rhythm compensation coefficient; the rhythm compensation coefficient is multiplied by the initial rehabilitation intervention rhythm to obtain the adaptive rehabilitation intervention rhythm.

[0056] For example, the postoperative care sensitivity compensation index of the target user is 1.3975, and the rhythm compensation coefficient is 1 / 1.3975≈0.7156; the initial rehabilitation intervention rhythm is 4 weeks, and the adapted rehabilitation intervention rhythm is 0.7156×4 weeks≈2.86 weeks, about 2 weeks and 6 days. In actual implementation, it can be adjusted to 3 weeks, which is shorter than the initial 4 weeks, meeting the requirement of short intervals for high-risk patients.

[0057] Secondly, an initial risk monitoring mechanism is obtained. This initial risk monitoring mechanism refers to the number of initial monitoring items, which is half of the total number of preset monitoring items. These monitoring items include at least circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, immune checkpoint molecular expression, cytokine profile, liver function, kidney function, bone marrow function, treatment-specific toxicity markers, and physical performance status. The initial risk monitoring mechanism refers to the pre-set number of initial monitoring items, which is half of the total number of preset monitoring items. The total number of preset monitoring items refers to all monitoring items specified in this step, including at least 11 items: circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, immune checkpoint molecular expression, cytokine profile, liver function, kidney function, bone marrow function, treatment-specific toxicity markers, and physical performance status. The initial number of monitoring items is half of the total number of preset monitoring items. The total number of preset monitoring items is calculated, half of which is rounded down to obtain the initial number of monitoring items, which serves as the initial risk monitoring mechanism. For example, if the total number of monitoring items is preset to 11, the initial number of monitoring items = 11 / 2 = 5.5.

[0058] Finally, the product of the postoperative care sensitivity compensation index and the initial number of monitoring items is rounded down to generate the number of adaptive monitoring items, which serves as the adaptive risk monitoring mechanism. The adaptive risk monitoring mechanism refers to the personalized number of monitoring items for the target user, obtained by rounding down the product of the postoperative care sensitivity compensation index and the initial number of monitoring items. The number of adaptive monitoring items refers to the rounded-down value of multiplying the postoperative care sensitivity compensation index by the initial number of monitoring items; the larger the index, the larger the product, and the more monitoring items. The postoperative care sensitivity compensation index is multiplied by the initial number of monitoring items to obtain the product result; the product result is then rounded down to obtain the number of adaptive monitoring items, which serves as the adaptive risk monitoring mechanism.

[0059] For example, the postoperative care sensitivity compensation index of the target user is 1.3975; the initial number of monitoring items is 5.5, 1.3975×5.5≈7.686, which is rounded down to 7 items; therefore, the adaptive risk monitoring mechanism is adapted to 7 monitoring items, and the specific monitoring items can be selected from 11 items: circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, liver function, physical status, and immune checkpoint molecular expression.

[0060] In this embodiment of the invention, the construction of a sensitivity compensation index and the generation of an appropriate rhythm and monitoring mechanism address the problems of fixed rhythms and lack of individualized adaptation in existing rehabilitation intervention mechanisms and risk monitoring mechanisms. The postoperative care sensitivity compensation index, constructed based on three core quantitative indicators, accurately integrates the patient's recurrence risk, protocol fit, and baseline status, achieving a shift from qualitative judgment to quantitative quantification. The appropriate rehabilitation intervention rhythm and risk monitoring mechanism generated through dynamic index optimization strictly follow the logic of higher risk, shorter intervals, and more comprehensive monitoring. This solves the problems of excessively long intervention intervals and insufficient monitoring for high-risk patients, while avoiding excessive intervention and monitoring burdens for low-risk patients. Simultaneously, the appropriate rhythm and monitoring mechanism generated in this step provide a clear execution framework for the dynamic optimization of the care plan in subsequent steps, ensuring that the optimization of the care plan can be implemented in a timely and accurate manner.

[0061] S500: Perform postoperative care on the target user according to the initial postoperative care plan, and dynamically optimize the postoperative care plan according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism.

[0062] In this embodiment of the invention, the target user is provided with postoperative care according to the initial postoperative care plan, and the postoperative care plan is dynamically optimized according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism. Existing postoperative care plans for liver cancer are fixed long-term once determined, and cannot be dynamically adjusted according to the patient's real-time treatment response, physical tolerance, and risk monitoring results. This results in a static and rigid approach, easily leading to undertreatment of high-risk patients and overtreatment of patients with low tolerance. This step addresses the lack of dynamic adaptability in existing plans by implementing care according to the initial postoperative care plan and relying on the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism to obtain real-time monitoring feedback data, thereby iteratively optimizing the care plan and achieving personalized and precise postoperative care for liver cancer.

[0063] Step S500 in the method provided in this embodiment of the invention includes: Postoperative care was provided to the target user according to the initial postoperative care plan, and the first monitoring feedback data of the target user was obtained at the end of the first feedback adjustment cycle according to the adaptive risk monitoring mechanism, wherein the first feedback adjustment cycle is the first feedback adjustment cycle in the adaptive rehabilitation intervention rhythm. The postoperative care plan for the next adjacent feedback adjustment cycle is optimized based on the first monitoring feedback data, and iterative dynamic optimization is performed according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism until the preset rehabilitation cycle is completed.

[0064] First, the target user is cared for postoperatively according to the initial postoperative care plan. The first monitoring feedback data of the target user is acquired at the end of the first feedback adjustment cycle, which is the first feedback adjustment cycle in the adapted rehabilitation intervention rhythm. The initial postoperative care plan refers to the care plan generated in S100 based on tumor immune microenvironment typing and matching, including four categories: medical intervention, metabolic and nutritional support, physiological function rehabilitation, and psychoneuroimmune regulation. The first feedback adjustment cycle refers to the first feedback adjustment cycle in the adapted rehabilitation intervention rhythm, with a duration equal to the quantified value of the adapted rehabilitation intervention rhythm. The first monitoring feedback data refers to the set of real-time physical status, treatment response, and risk-related data of the target user acquired at the end of the first feedback adjustment cycle according to the adapted monitoring items of the adapted risk monitoring mechanism. The target user is cared for according to the initial postoperative care plan; during the care process, data is collected periodically according to the monitoring items of the adapted risk monitoring mechanism; at the end of the first feedback adjustment cycle, all monitoring data are integrated to form the first monitoring feedback data.

[0065] For example, the initial postoperative care protocol for the target patient was implemented for 3 weeks: Medical intervention: intravenous infusion of the PD-1 inhibitor nivolumab 3 mg / kg every 2 weeks, for a total of 1 infusion; Metabolic and nutritional support: daily protein intake of 1.8 g / kg, supplementation with Omega-3 fatty acids 1 g / day, diet mainly consisting of high-quality protein, avoiding high sugar and high fat; Physical rehabilitation: 30 minutes of slow walking daily, 5 times a week; Psychological and neuroimmunological regulation: cognitive behavioral therapy once a week, 15 minutes of mindfulness meditation daily. The target patient was monitored according to the seven items of the appropriate risk monitoring mechanism, including: circulating tumor DNA, classical tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, liver function, performance status, and immune checkpoint molecule expression. At the end of 3 weeks, the first monitoring feedback data was obtained: Circulating tumor DNA: negative; Classical tumor marker (alpha-fetoprotein): 10 ng / mL (decreased from baseline 12 ng / mL); Imaging examinations: no new lesions were found in the liver, and the surgical area was healing well; Peripheral blood immune cell dynamic typing: CD8 + T cell percentage 11% (1% increase from baseline 10%); Liver function: ALT 23 U / L, AST 19 U / L (both normal); Physical fitness: Grip strength 29 kg (increase from baseline 28 kg), can easily complete 30 minutes of slow walking daily; Immune checkpoint molecule expression: PD-1 expression level 2.3 ng / mL (0.2 ng / mL decrease from baseline 2.5 ng / mL).

[0066] Secondly, the postoperative care plan for the next adjacent feedback adjustment cycle is optimized based on the first monitoring feedback data, and iteratively and dynamically optimized according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism until the preset rehabilitation cycle is completed. The next adjacent feedback adjustment cycle refers to the continuous feedback adjustment cycle after the first feedback adjustment cycle, with a duration consistent with the adaptive rehabilitation intervention rhythm. Iterative dynamic optimization refers to the process of repeatedly acquiring monitoring feedback data, optimizing the care plan, and implementing the new plan after each feedback adjustment cycle, until the preset rehabilitation cycle is completed. The expected target data refers to the ideal state data that the patient should achieve at the end of each feedback adjustment cycle, which is based on the care effects of similar patients in the past and clinical guidelines. The deviation between the first monitoring feedback data and the expected target data of the first feedback adjustment cycle is calculated; the postoperative care plan for the next adjacent feedback adjustment cycle is optimized in a targeted manner according to the type and degree of deviation; the optimized care plan, the adaptive rehabilitation intervention rhythm, and the adaptive risk monitoring mechanism are implemented to acquire the monitoring feedback data for the next cycle; the above process is repeated to iteratively and dynamically optimize the care plan until the preset rehabilitation cycle is completed.

[0067] For example, the expected target data for the first feedback adjustment cycle is: CD8 + The proportion of T cells increased to 13%, and PD-1 expression decreased to 2.0 ng / mL. Compared with the first monitoring feedback data: CD8 + The percentage of T cells was 11%, and the PD-1 expression level was 2.3 ng / mL, indicating a bias towards insufficient enhancement of immune function. Specifically: CD8 + The T-cell percentage did not meet expectations (deviation of 2%), and the decrease in PD-1 expression was insufficient (deviation of 0.3 ng / mL). To address the deviations in immune function improvement, the following adjustments were made: Medical intervention: Maintain the PD-1 inhibitor dose, shortening the dosing interval to once every 10 days; Metabolic and nutritional support: Increase daily protein intake to 2.0 g / kg, and supplement with vitamin D (1000 IU / day); Physiological rehabilitation: Increase exercise intensity to 40 minutes of brisk walking daily, 5 times a week; Psychological and neuroimmunological regulation: Remain unchanged. The second 3-week feedback adjustment cycle was initiated, implementing the optimized care plan. Second monitoring feedback data was obtained based on 7 monitoring items, and the deviation was calculated before further optimization. This process was repeated for the third and fourth 3-week feedback adjustment cycles until the 12-month pre-set recovery cycle was completed.

[0068] In this embodiment of the invention, the core problem of static and rigid postoperative care protocols for liver cancer is solved through an initial protocol execution, real-time monitoring, deviation optimization, and iterative loop. Relying on an adaptive rehabilitation intervention rhythm and risk monitoring mechanism, real-time and accurate monitoring of the patient's treatment response and physical condition is achieved, avoiding the lag of traditional fixed-cycle monitoring. By calculating the deviation between the monitoring data and the expected target data, the care protocol is optimized in a targeted manner, such as increasing protein intake and exercise intensity, achieving dynamic adaptability and effectively addressing the problem of insufficient treatment for high-risk patients. Simultaneously, through iterative dynamic optimization, individualized and precise postoperative care for liver cancer is achieved, improving the effectiveness and safety of care and providing the ultimate execution guarantee for reducing the risk of postoperative recurrence and metastasis and improving long-term quality of life.

[0069] Through the specific implementation methods described above, the embodiments of the present invention achieve the following technical effects: This invention provides a method for optimizing postoperative care for liver cancer patients and an intelligent chip. First, multi-dimensional detection enables precise subtyping of the tumor immune microenvironment and generates a personalized initial care plan, breaking the limitations of homogeneous traditional treatment plans from the outset. Second, a deep learning model accurately predicts the probability of recurrence and metastasis, combining baseline status detection information to generate the intensity of the care plan and a baseline status score, solving the problem of coarse risk prediction models. Third, by constructing a postoperative care sensitivity compensation index, a dynamic adjustment mechanism for rehabilitation intervention and risk monitoring is established, balancing monitoring intensity with medical burden and avoiding the drawbacks of insufficient monitoring for high-risk patients and excessive monitoring for low-risk patients. Finally, relying on the adjusted rhythm and monitoring mechanism to obtain real-time feedback data, the care plan is iteratively optimized through deviation analysis, overcoming the shortcomings of static and rigid care plans. Ultimately, this achieves personalized and precise postoperative care for liver cancer patients, improving the accuracy and effectiveness of care plans, effectively reducing the risk of postoperative recurrence and metastasis, optimizing the allocation of medical resources, and improving the long-term quality of life for patients.

[0070] Example 2, as Figure 2 As shown, this invention provides an intelligent chip for optimizing postoperative care for liver cancer patients. The intelligent chip includes: The initial postoperative care plan generation module 11 is used to perform multi-dimensional detection on the liver cancer tumor tissue specimen of the target user to obtain a multi-dimensional detection dataset, determine the tumor immune microenvironment type based on the multi-dimensional detection dataset, and generate an initial postoperative care plan. The recurrence risk probability prediction module 12 is used to analyze and obtain the predicted recurrence and metastasis risk probability of the target user within a preset recovery period based on the multi-dimensional detection dataset and the initial postoperative care plan. The baseline evaluation module 13 is used to evaluate the initial postoperative care plan based on the baseline status detection information of the target user, generate the care plan strength, and generate a baseline status score based on the baseline status detection information. The intervention rhythm and monitoring mechanism development module 14 is used to develop an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism based on the predicted relapse and metastasis risk probability, the intensity of the care plan and the baseline status score; The dynamic care and plan optimization module 15 is used to provide postoperative care to the target user according to the initial postoperative care plan, and to dynamically optimize the postoperative care plan according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism.

[0071] In one embodiment, the initial care plan generation module 11 is further configured to: According to the preset detection process, the surgically removed liver cancer tumor tissue specimens of the target user are subjected to multi-dimensional detection to obtain a multi-dimensional detection dataset. The multi-dimensional detection dataset includes immune cell composition, immune cell spatial distribution, tumor cell composition, tumor cell spatial distribution, stromal cell composition, stromal cell spatial distribution, key signal molecule composition, and key signal molecule spatial distribution. The multidimensional detection dataset is compared with several standard detection datasets in the pre-constructed detection data-immune microenvironment type comparison library. The immune microenvironment type corresponding to the standard detection dataset with the highest comprehensive similarity is selected as the tumor immune microenvironment type. An initial postoperative care plan is obtained based on the matching of the tumor immune microenvironment type. The postoperative care plan includes medical intervention, metabolic and nutritional support, physiological function rehabilitation and psychoneuroimmune regulation.

[0072] The immune microenvironment types include the immune-inflammatory type-functional activation subtype, the immune-inflammatory type-functional depletion subtype, the immune-rejection type-matrix barrier subtype, the immune-rejection type-vascular abnormality subtype, the immune-desert type-immune neglect subtype, and the immune-desert type-active inhibition subtype.

[0073] In one embodiment, the recurrence risk probability prediction module 12 is further used for: Based on the postoperative care records of historical users with liver cancer, several multidimensional detection datasets and several postoperative care plans were collected. With the time span of the preset recovery cycle as a constraint, the proportion of recurrence and metastasis risk events of historical users under different multidimensional detection datasets and postoperative care plans was statistically analyzed as the sample recurrence and metastasis risk probability, and the recurrence and metastasis risk probabilities of several samples were obtained. Using the multi-dimensional detection dataset of several samples and the postoperative care plan of several samples as input data, and using the recurrence and metastasis risk probability of several samples as supervision label, a deep learning model is trained until convergence to generate a recurrence and metastasis risk prediction model. The multi-dimensional detection dataset and the initial postoperative care plan are input into the recurrence and metastasis risk prediction model, which outputs the predicted recurrence and metastasis risk probability of the target user within a preset recovery period.

[0074] In one embodiment, the baseline evaluation module 13 is further configured to: Obtain baseline status detection information of the target user, wherein the baseline status detection information includes basic physiological information, organ function indicators, physical and nutritional status, tumor burden baseline, and infection and immune status; According to the preset evaluation dimensions, the initial postoperative care plan is evaluated in multiple dimensions based on the baseline status detection information to obtain the strength of the care plan. The preset evaluation dimensions include organ function tolerance, physical strength, comorbid diseases and immune background. A comprehensive status evaluation of the target user is performed based on the baseline status detection information, and a baseline status score is generated.

[0075] In one embodiment, the intervention rhythm and monitoring mechanism formulation module 14 is further used for: Based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, postoperative care sensitivity is identified, and the postoperative care sensitivity compensation index is obtained. Based on the postoperative care sensitivity compensation index, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted to generate an appropriate rehabilitation intervention rhythm and appropriate risk monitoring mechanism.

[0076] Specifically, postoperative care sensitivity is identified based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, and a postoperative care sensitivity compensation index is obtained, including: The ratio of the predicted relapse and metastasis risk probability to the proportion of historical relapse and metastasis risk events recorded in the historical time zone is used as the first care sensitivity coefficient. The ratio of the intensity of the described care plan to the average intensity of historical care plans recorded in the historical time zone is used as the second care sensitivity coefficient. The ratio of the average historical baseline status score of historical users recorded within the historical time zone to the baseline status score is used as the third care sensitivity coefficient. The first, second, and third care sensitivity coefficients are weighted and fused to generate the postoperative care sensitivity compensation index.

[0077] Specifically, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted based on the postoperative care sensitivity compensation index to generate an adapted rehabilitation intervention rhythm and an adapted risk monitoring mechanism, including: The reciprocal of the postoperative care sensitivity compensation index is used as the rhythm compensation coefficient, and the product of the rhythm compensation coefficient and the initial rehabilitation intervention rhythm is used as the adaptive rehabilitation intervention rhythm, wherein the initial rehabilitation intervention rhythm is the duration of the initial feedback adjustment cycle. An initial risk monitoring mechanism is obtained, wherein the initial risk monitoring mechanism is the number of initial monitoring items, which is half of the total number of preset monitoring items. The monitoring items include at least circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, immune checkpoint molecule expression, cytokine profile, liver function, kidney function, bone marrow function, treatment-specific toxicity markers, and physical status. The product of the postoperative care sensitivity compensation index and the initial number of monitoring items is rounded down to generate the number of adaptive monitoring items, which serves as the adaptive risk monitoring mechanism.

[0078] In one embodiment, the dynamic care and scheme optimization module 15 is further configured to: Postoperative care was provided to the target user according to the initial postoperative care plan, and the first monitoring feedback data of the target user was obtained at the end of the first feedback adjustment cycle according to the adaptive risk monitoring mechanism, wherein the first feedback adjustment cycle is the first feedback adjustment cycle in the adaptive rehabilitation intervention rhythm. The postoperative care plan for the next adjacent feedback adjustment cycle is optimized based on the first monitoring feedback data, and iterative dynamic optimization is performed according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism until the preset rehabilitation cycle is completed.

[0079] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0081] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A method for optimizing postoperative care for liver cancer, characterized in that, The method includes: Multidimensional detection datasets are obtained by performing multidimensional detection on liver cancer tumor tissue specimens of target users. Based on the multidimensional detection datasets, the tumor immune microenvironment type is determined, and an initial postoperative care plan is generated. Based on the analysis of the multi-dimensional detection dataset and the initial postoperative care plan, the predicted recurrence and metastasis risk probability of the target user within the preset recovery period is obtained. The initial postoperative care plan is evaluated based on the baseline status detection information of the target user to generate the strength of the care plan, and a baseline status score is generated based on the baseline status detection information. Based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism are developed; Postoperative care was provided to the target user according to the initial postoperative care plan, and the postoperative care plan was dynamically optimized according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism.

2. The method for optimizing postoperative care for liver cancer according to claim 1, characterized in that, Multidimensional detection datasets are obtained by performing multidimensional detection on liver cancer tumor tissue specimens from target users. Based on these datasets, the tumor immune microenvironment type is determined, and an initial postoperative care plan is generated, including: According to the preset detection process, the surgically removed liver cancer tumor tissue specimens of the target user are subjected to multi-dimensional detection to obtain a multi-dimensional detection dataset. The multi-dimensional detection dataset includes immune cell composition, immune cell spatial distribution, tumor cell composition, tumor cell spatial distribution, stromal cell composition, stromal cell spatial distribution, key signal molecule composition, and key signal molecule spatial distribution. The multidimensional detection dataset is compared with several standard detection datasets in the pre-constructed detection data-immune microenvironment type comparison library. The immune microenvironment type corresponding to the standard detection dataset with the highest comprehensive similarity is selected as the tumor immune microenvironment type. An initial postoperative care plan is obtained based on the matching of the tumor immune microenvironment type. The postoperative care plan includes medical intervention, metabolic and nutritional support, physiological function rehabilitation and psychoneuroimmune regulation.

3. The method for optimizing postoperative care for liver cancer according to claim 2, characterized in that, The immune microenvironment types include the immune-inflammatory type-functional activation subtype, the immune-inflammatory type-functional depletion subtype, the immune-rejection type-matrix barrier subtype, the immune-rejection type-vascular abnormality subtype, the immune-desert type-immune neglect subtype, and the immune-desert type-active inhibition subtype.

4. The method for optimizing postoperative care for liver cancer according to claim 1, characterized in that, Based on the analysis of the multi-dimensional detection dataset and the initial postoperative care plan, the predicted recurrence and metastasis risk probability of the target user within the preset recovery period is obtained, including: Based on the postoperative care records of historical users with liver cancer, several multidimensional detection datasets and several postoperative care plans were collected. With the time span of the preset recovery cycle as a constraint, the proportion of recurrence and metastasis risk events of historical users under different multidimensional detection datasets and postoperative care plans was statistically analyzed as the sample recurrence and metastasis risk probability, and the recurrence and metastasis risk probabilities of several samples were obtained. Using the multi-dimensional detection dataset of several samples and the postoperative care plan of several samples as input data, and using the recurrence and metastasis risk probability of several samples as supervision label, a deep learning model is trained until convergence to generate a recurrence and metastasis risk prediction model. The multi-dimensional detection dataset and the initial postoperative care plan are input into the recurrence and metastasis risk prediction model, which outputs the predicted recurrence and metastasis risk probability of the target user within a preset recovery period.

5. The method for optimizing postoperative care for liver cancer according to claim 1, characterized in that, The initial postoperative care plan is evaluated based on the target user's baseline status detection information to generate the care plan strength, and a baseline status score is generated based on the baseline status detection information, including: Obtain baseline status detection information of the target user, wherein the baseline status detection information includes basic physiological information, organ function indicators, physical and nutritional status, tumor burden baseline, and infection and immune status; According to the preset evaluation dimensions, the initial postoperative care plan is evaluated in multiple dimensions based on the baseline status detection information to obtain the strength of the care plan. The preset evaluation dimensions include organ function tolerance, physical strength, comorbid diseases and immune background. A comprehensive status evaluation of the target user is performed based on the baseline status detection information, and a baseline status score is generated.

6. The method for optimizing postoperative care for liver cancer according to claim 1, characterized in that, Based on the predicted relapse and metastasis risk probability, the intensity of the care plan, and the baseline status score, an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism are developed, including: Based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, postoperative care sensitivity is identified, and the postoperative care sensitivity compensation index is obtained. Based on the postoperative care sensitivity compensation index, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted to generate an appropriate rehabilitation intervention rhythm and appropriate risk monitoring mechanism.

7. The method for optimizing postoperative care for liver cancer according to claim 6, characterized in that, Postoperative care sensitivity is identified based on the predicted recurrence and metastasis risk probability, the intensity of the care plan, and the baseline status score, and a postoperative care sensitivity compensation index is obtained, including: The ratio of the predicted relapse and metastasis risk probability to the proportion of historical relapse and metastasis risk events recorded in the historical time zone is used as the first care sensitivity coefficient. The ratio of the intensity of the described care plan to the average intensity of historical care plans recorded in the historical time zone is used as the second care sensitivity coefficient. The ratio of the average historical baseline status score of historical users recorded within the historical time zone to the baseline status score is used as the third care sensitivity coefficient. The first, second, and third care sensitivity coefficients are weighted and fused to generate the postoperative care sensitivity compensation index.

8. The method for optimizing postoperative care for liver cancer according to claim 6, characterized in that, Based on the postoperative care sensitivity compensation index, the initial rehabilitation intervention rhythm and initial risk monitoring mechanism are optimized and adjusted to generate an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism, including: The reciprocal of the postoperative care sensitivity compensation index is used as the rhythm compensation coefficient, and the product of the rhythm compensation coefficient and the initial rehabilitation intervention rhythm is used as the adaptive rehabilitation intervention rhythm, wherein the initial rehabilitation intervention rhythm is the duration of the initial feedback adjustment cycle. An initial risk monitoring mechanism is obtained, wherein the initial risk monitoring mechanism is the number of initial monitoring items, which is half of the total number of preset monitoring items. The monitoring items include at least circulating tumor DNA, classic tumor markers, imaging examinations, peripheral blood immune cell dynamic typing, immune checkpoint molecule expression, cytokine profile, liver function, kidney function, bone marrow function, treatment-specific toxicity markers, and physical status. The product of the postoperative care sensitivity compensation index and the initial number of monitoring items is rounded down to generate the number of adaptive monitoring items, which serves as the adaptive risk monitoring mechanism.

9. The method for optimizing postoperative care for liver cancer according to claim 1, characterized in that, Postoperative care was provided to the target user according to the initial postoperative care plan, and the postoperative care plan was dynamically optimized according to the adaptive rehabilitation intervention rhythm and adaptive risk monitoring mechanism, including: Postoperative care was provided to the target user according to the initial postoperative care plan, and the first monitoring feedback data of the target user was obtained at the end of the first feedback adjustment cycle according to the adaptive risk monitoring mechanism, wherein the first feedback adjustment cycle is the first feedback adjustment cycle in the adaptive rehabilitation intervention rhythm. The postoperative care plan for the next adjacent feedback adjustment cycle is optimized based on the first monitoring feedback data, and iterative dynamic optimization is performed according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism until the preset rehabilitation cycle is completed.

10. A smart chip for optimizing postoperative care after liver cancer surgery, characterized in that, For implementing the method for optimizing postoperative care for liver cancer as described in any one of claims 1-9, the smart chip comprises: The initial postoperative care plan generation module is used to perform multi-dimensional detection on the liver cancer tumor tissue specimens of the target user to obtain a multi-dimensional detection dataset, determine the tumor immune microenvironment type based on the multi-dimensional detection dataset, and generate an initial postoperative care plan. The recurrence risk probability prediction module is used to analyze and obtain the predicted recurrence and metastasis risk probability of the target user within a preset recovery period based on the multi-dimensional detection dataset and the initial postoperative care plan. The baseline evaluation module is used to evaluate the initial postoperative care plan based on the baseline status detection information of the target user, generate the strength of the care plan, and generate a baseline status score based on the baseline status detection information. The intervention rhythm and monitoring mechanism development module is used to develop an appropriate rehabilitation intervention rhythm and an appropriate risk monitoring mechanism based on the predicted relapse and metastasis risk probability, the intensity of the care plan, and the baseline status score. The dynamic care and plan optimization module is used to provide postoperative care to the target user according to the initial postoperative care plan, and to dynamically optimize the postoperative care plan according to the adaptive rehabilitation intervention rhythm and the adaptive risk monitoring mechanism.

Images