CN-122024978-A - Tumor treatment strategy decision method based on threshold guidance and related device

CN122024978ACN 122024978 ACN122024978 ACN 122024978ACN-122024978-A

Abstract

The application discloses a threshold-guided tumor treatment strategy decision method and a related device, which aim at the problems of drug resistance, lack of self-adaptability, difficult clinical realization and the like of the existing tumor treatment strategy, and are realized by the following steps: defining a treatment period and a non-treatment period guided by a threshold, constructing a dynamic system model guided by the threshold, carrying out theoretical analysis and numerical simulation of dynamic behaviors, carrying out comparative analysis with other treatment strategies, carrying out theoretical parameter sensitivity analysis, and optimizing an adaptive treatment strategy by reinforcement learning. The method dynamically adjusts the treatment decision through a single threshold strategy, realizes highly personalized and accurate treatment, can effectively cope with heterogeneity of tumor dynamics among patients, obviously prolongs tumor progression time, reduces medicine dosage, reduces toxic and side effects, improves life quality and treatment compliance of patients, and provides a new effective method for tumor treatment.

Inventors

TANG BIAO
WANG NINGJING
LIU YAN
MA KEXIN
YANG ZAI
WANG SHENG

Assignees

西安交通大学

Dates

Publication Date: 20260512
Application Date: 20260122

Claims (10)

1. A threshold guided tumor treatment strategy decision method, comprising the steps of; collecting biomarker level data for dynamic patient monitoring; inputting biomarker level data and corresponding detection time information into a pre-constructed and optimized tumor treatment strategy decision model to obtain a corresponding tumor adaptive treatment strategy; The pre-constructed and optimized tumor treatment strategy decision model is a threshold-guided dynamics system model optimized through reinforcement learning.
2. The threshold guided tumor treatment strategy decision method of claim 1, wherein the threshold guided dynamics system model is specifically as follows: Wherein, the And Respectively representing drug-sensitive tumor cells and drug-resistant tumor cells; And Are the intrinsic growth rates of both cell types; And Representing a competition coefficient for quantifying the competition inhibition intensity of two cells; Representing a corresponding decrease in drug-sensitive cell viability in the case of treatment, parameter k representing environmental load-bearing capacity, the first subscript representing the system, the second subscript representing the cell, And Indicating drug-sensitive tumor cells in the state of treatment And drug-resistant tumor cells Maximum number that can be reached; And Then the maximum number of two cells that can be reached in the non-treated state is indicated; Representing the current time; The time period during which the treatment is performed is indicated, Means a period of time during which no treatment is performed; represents an initial number of tumor cells; indicating that the number of cells is a positive real number.
3. The threshold-guided tumor treatment strategy decision method of claim 2, wherein the tumor treatment strategy decision model comprises a free system and a control system, in particular as follows: Wherein, the Representing differential equations under treatment conditions; representing differential equations in the non-therapeutic state; The system of kinetic behavior differential equations of the free system is as follows: the system of differential equations of the dynamics behavior of the control system is as follows: 。
4. the threshold-guided tumor treatment strategy decision method of claim 1, wherein the tumor treatment strategy comprises: Judging whether the tumor size data exceeds a preset threshold value in a preset fixed time interval When the tumor size exceeds a preset threshold The time period is defined as a treatment time period during which the control system is activated When the tumor size is not exceeded The time period is defined as a non-therapeutic time period during which the free system is enabled Treatment period And non-therapeutic time period The method is characterized by comprising the following steps: Wherein, the Representing a priori treatment period; representing a priori non-therapeutic periods.
5. The threshold-guided tumor treatment strategy decision method of claim 1, wherein optimizing the tumor treatment strategy decision model comprises: establishing a Markov decision process and defining core elements such as states, actions, rewards and state transition probabilities; Constructing a minimum decision frame and comparing different reinforcement learning strategies; And a near-end strategy optimization algorithm is adopted, and key time sequence characteristics are extracted from dynamic changes by means of the double-end output structure and LSTM time sequence modeling capability, so that self-adaptive adjustment is realized.
6. The method for deciding tumor treatment strategy based on threshold guidance according to claim 5, wherein the adoption of the near-end strategy optimization algorithm extracts key time sequence features from dynamic changes by means of a double-end output structure and a long-short-time memory network LSTM time sequence modeling capability, and realizes self-adaptive adjustment, and the method is specifically as follows: Input layer receives environmental status And mapping the dimension to 32 dimensions through a linear transformation layer; Introducing a long-short-time memory network LSTM to process time sequence data of tumor size, wherein the dimension of an input layer is 32, the dimension of an output layer is 64, and the output of the long-short-time memory network LSTM is connected to a shared deep feature processing network, wherein the hierarchy structure of the long-short-time memory network LSTM is 64-128-64-32-16-10-in turn, so that core information related to decision is extracted from advanced features; Performing advantage estimation by adopting a near-end strategy optimization algorithm, namely estimating the relative advantages and disadvantages of a certain action under the current strategy by utilizing an advantage function, and guiding an intelligent agent to select a better action; the 10-dimensional characteristics output by the shared network are input into a strategy network, mapped to 31-dimensional through a linear layer, and output into an action probability distribution through a Softmax activation function 。
7. A threshold guided tumor treatment strategy decision system comprising: the data acquisition module is used for collecting biomarker level data of dynamic monitoring of a patient; the strategy guiding module is used for inputting the biomarker level data and the corresponding detection time information into a pre-constructed and optimized tumor treatment strategy decision model to obtain a corresponding tumor adaptive treatment strategy; The pre-constructed and optimized tumor treatment strategy decision model is a threshold-guided dynamics system model optimized through reinforcement learning.
8. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1-6 when the computer program is executed.
9. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method according to any of claims 1-6.
10. A computer program product, characterized in that it comprises computer instructions, which are read by a processor of a computer device, which computer instructions are executed by the processor of the computer device to carry out the steps of the method according to any one of claims 1 to 6.

Description

Tumor treatment strategy decision method based on threshold guidance and related device Technical Field The application belongs to the technical field of strategy analysis, and relates to a threshold-guided tumor treatment strategy decision method and a related device. Background Cancer has long been a major cause of global morbidity and mortality, and traditional tumor treatment strategies, such as continuous aggressive therapy, while capable of significantly reducing tumor burden in the short term, are prone to drug resistance, resulting in treatment failure and even recurrence and metastasis. This phenomenon indicates that the drug resistance problem cannot be fundamentally solved by simply increasing the drug dosage or prolonging the drug administration time. As a promising alternative, adaptive therapy (ADAPTIVE THERAPY) was proposed and explored in several cancer species (e.g., prostate cancer). The core idea is that the maximum elimination of tumor load is not pursued, but the ecological competition relationship between drug sensitive tumor cells and drug resistant tumor cells is actively utilized through intermittent administration, so that the expansion of the drug resistant cells is inhibited. Through proper medicine reduction or medicine stopping, a certain number of medicine sensitive cells are reserved as far as possible, so that the medicine resistant cells are continuously pressed in the subsequent treatment course, and the treatment period is prolonged. Clinical trials and mathematical modeling studies have shown that adaptive therapy is expected to significantly extend the progression free survival and overall survival of patients by delaying the onset of therapeutic resistance. For quantitative analysis and optimization of adaptive treatment protocols, various mathematical models have been proposed to describe the evolution kinetics of tumors under treatment and the treatment switching mechanism. Typical among these are: 1) The periodic switching system performs periodic switching through presetting fixed treatment period and non-treatment period (such as two weeks of drug administration and three weeks of drug withdrawal), has the advantages of simple structure and easy analysis and realization, but the switching rule is completely based on preset time and does not depend on the real state of a patient (such as tumor size or biomarker level and the like), and lacks self-adaptability. For patients with slower tumor growth or greater individual variability, the fixed period can easily lead to over-or under-treatment, and it is difficult to meet the "individuation, on-demand" principle emphasized by adaptive therapy. 2) State-dependent switching systems (e.g., filippov systems) to avoid unnecessary treatment burden, some studies have employed state-dependent switching frameworks to describe intermittent treatment strategies based on tumor burden by introducing thresholds. Typically, such models are those that switch treatment states based on the current tumor size, i.e., once the tumor size or other relevant indicator reaches a preset threshold, the system switches between "drug administration/drug withdrawal" states immediately. While this approach achieves formal "threshold guidance," it requires frequent or near-continuous monitoring of patient status, which is not readily achieved in clinical practice, and too frequent status switching may also cause "switch jitter" that is detrimental to stable performance of the regimen and patient compliance. The existing tumor treatment strategies have a plurality of problems, and the traditional continuous aggressive therapy can obviously reduce the tumor load in a short period, but easily causes the generation of drug resistance, thereby causing treatment failure and even occurrence of recurrence and metastasis. The existing adaptive therapy modeling framework is either a switching system based on a fixed period, lacks adaptivity, and is prone to over-treatment or under-treatment, or a switching system based on state dependence, requires frequent monitoring of patient states, is difficult to achieve in clinical practice, and may cause "switch jitter". These model frameworks fail to organically combine the standard "regular monitoring" flow with the "threshold guidance" decision logic in clinical practice, and it is difficult to directly guide the real administration decision, so there is an urgent need in the art for a general modeling and analysis framework that both conforms to the clinical actual operational flow and combines periodic and intermittent treatment with threshold guidance strategies. Disclosure of Invention The application aims to solve the problems in the prior art and provides a threshold-guided tumor treatment strategy decision method and a related device. In order to achieve the above purpose, the application is realized by adopting the following technical scheme: In a first aspect, the present application provides a method for