技术领域Technical field
本发明属于医学图像处理技术领域，具体涉及一种基于影像组学的脑膜瘤伽马刀后水肿预测系统。The invention belongs to the technical field of medical image processing, and specifically relates to a post-gamma knife edema prediction system for meningioma based on radiomics.
背景技术Background technique
脑膜瘤是最常见的颅内良性肿瘤，占颅内原发肿瘤的13-26％。尽管大多数脑膜瘤有良好的预后，但由于完全切除仍然有难度，它们常常复发。此外，由于脑膜瘤边界清晰，在磁共振成像定位准确，它是立体定向放射外科(SRS)的理想肿瘤类型。因此，SRS已成为脑膜瘤患者长期抑制肿瘤生长和预防症状加剧的重要治疗策略。SRS治疗的5～10年局部肿瘤控制率可达到87～98％。伽玛刀放射外科治疗(以下简称伽马刀)是应用最广泛的SRS方法。虽然伽马刀是一种推荐的治疗方法，但术后有瘤周放射性水肿的风险，文献报道的发生率范围很广，从2％到50％不等，平均/中位发病时间在3～9个月之间。有些GKS术后水肿可能无症状，但有些水肿可导致头痛、恶心、共济失调、其他神经症状，甚至死亡。在较为严重的情况下，这些症状可能需要激素治疗，甚至需要外科切除病灶。这些水肿带来的不良反应的发生率和伽马刀治疗的必要性，是临床医师需要衡量的。因此预测术后水肿的发生概率对于临床决策有重要意义。一些文献已报道了脑膜瘤患者伽马刀术后水肿风险增加的相关因素。潜在因素包括更高的剂量、更大的肿瘤体积、颅底(尤其是近矢状面)位置、治疗前水肿的存在等。Meningiomas are the most common benign intracranial tumors, accounting for 13-26% of primary intracranial tumors. Although most meningiomas have a good prognosis, they often recur because complete resection is still difficult. In addition, meningiomas are ideal tumor types for stereotactic radiosurgery (SRS) because of their clear boundaries and accurate positioning on magnetic resonance imaging. Therefore, SRS has become an important treatment strategy for meningioma patients to inhibit tumor growth and prevent symptom exacerbation in the long term. The 5-10-year local tumor control rate of SRS treatment can reach 87-98%. Gamma Knife radiosurgery (hereinafter referred to as Gamma Knife) is the most widely used SRS method. Although Gamma Knife is a recommended treatment method, there is a risk of peritumoral radiation edema after surgery, and the incidence reported in the literature ranges widely, from 2% to 50%, with a mean/median onset time of 3 to 9 months. Some GKS postoperative edema may be asymptomatic, but some edema can cause headaches, nausea, ataxia, other neurological symptoms, and even death. In more severe cases, these symptoms may require hormone therapy or even surgical resection of the lesion. The incidence of adverse reactions caused by these edema and the necessity of gamma knife treatment are what clinicians need to weigh. Therefore, predicting the probability of postoperative edema is important for clinical decision-making. Some literature has reported factors related to the increased risk of edema after gamma knife surgery in patients with meningioma. Potential factors include higher doses, larger tumor volume, skull base (especially near sagittal plane) location, and the presence of edema before treatment.
近年来，影像组学在一些疾病预后方面显示出巨大的潜力。影像组学是通过从图像中提取大量描述图像各个方面的定量特征来实现的。这种高维数据是对传统统计技术的挑战，如线性回归、Cox回归等。同时，当存在高删失率时，Cox回归的性能将不可靠。相比之下，近年来发展迅速的机器学习方法能够有效地处理高维问题。在各种机器学习模型中，随机生存森林(RSF)是一个非参数模型，它在包含事件发生时间的数据包含大量协变量时，可以获得了很高的预测性能。一些研究已经证实了RSF在多个领域的良好性能，但在脑膜瘤患者在伽马刀术后水肿预测中使用RSF还未有报道。In recent years, radiomics has shown great potential in the prognosis of some diseases. Radiomics is achieved by extracting a large number of quantitative features from images that describe various aspects of the image. This kind of high-dimensional data is a challenge to traditional statistical techniques, such as linear regression, Cox regression, etc. At the same time, the performance of Cox regression will be unreliable when there is a high censorship rate. In contrast, machine learning methods, which have developed rapidly in recent years, can effectively handle high-dimensional problems. Among various machine learning models, Random Survival Forest (RSF) is a non-parametric model that can achieve high predictive performance when the data containing event time contains a large number of covariates. Some studies have confirmed the good performance of RSF in multiple fields, but the use of RSF in predicting edema after gamma knife surgery in meningioma patients has not been reported.
发明内容Summary of the invention
本发明的目的是提供一种基于影像组学的脑膜瘤伽马刀后水肿预测系统。可基于脑膜瘤患者的基本临床信息、影像学特征，以及影像组学特征，建立能够快速预测伽马刀后水肿发生率及发生时间的工具。The purpose of the present invention is to provide a radiomics-based system for predicting edema after gamma knife surgery for meningioma. Based on the basic clinical information, imaging features, and radiomics features of meningioma patients, a tool that can quickly predict the incidence and time of edema after gamma knife surgery can be established.
为了达到上述目的，本发明一方面提供一种基于影像组学的脑膜瘤伽马刀后瘤周水肿预测系统，所述系统包括：In order to achieve the above object, the present invention provides a system for predicting peritumoral edema after gamma knife surgery for meningioma based on radiomics, the system comprising:
临床数据获取单元：Clinical data acquisition unit:
--用以获取行伽马刀治疗的脑膜瘤患者的临床资料、治疗前及治疗后随访的头颅磁共振图像信息，并形成临床数据集；--To obtain the clinical data of meningioma patients undergoing gamma knife treatment, and the brain magnetic resonance image information before and after treatment, and to form a clinical data set;
结局判定单元：Ending determination unit:
--用以从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间，并形成结局数据集；--Used to obtain the occurrence and time of postoperative peritumoral edema from patients' postoperative follow-up cranial magnetic resonance imaging, and form an outcome data set;
影像组学特征提取单元：Radiomics feature extraction unit:
--用以从患者术前头颅磁共振常规序列脑膜瘤区域进行影像组学特征提取及筛选，并形成影像组学特征数据集；--Used to extract and screen radiomics features from the meningioma area of the patient's preoperative cranial MRI routine sequence and form an radiomics feature dataset;
影像学语义特征获取单元：Imaging semantic feature acquisition unit:
--用以解读分析患者术前磁共振常规序列脑膜瘤区域图像获取影像学语义特征，并形成影像学语义特征数据集；--Used to interpret and analyze patients' preoperative magnetic resonance conventional sequence meningioma area images to obtain imaging semantic features, and form an imaging semantic feature data set;
预测模型建立单元：Predictive model building unit:
--用以将患者随机分为训练集和测试集，根据训练集的结局数据，结合其临床数据、影像组学特征数据和影像学语义特征数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证，评估其对脑膜瘤伽马刀术后水肿的发生率及发生时间的预测效果，生成最优模型；--Used to randomly divide patients into training sets and test sets, and establish a series of random survival forest models containing different types of features based on the outcome data of the training set, combined with their clinical data, radiomics feature data and imaging semantic feature data , and verified in the test set to evaluate its prediction effect on the incidence and time of edema after gamma knife surgery for meningioma, and generate the optimal model;
输出单元：Output unit:
--用以输出最优模型得到的脑膜瘤伽马刀后瘤周水肿的发生率及发生时间的评估预测值。--Used to output the estimated prediction value of the incidence and time of peritumoral edema after gamma knife meningioma obtained from the optimal model.
进一步的，临床数据获取单元中，脑膜瘤患者的临床资料包括患者的性别、年龄、病变范围、周边剂量、中心剂量、靶点数、等剂量线，外科手术史和是否分次治疗。Further, in the clinical data acquisition unit, the clinical data of meningioma patients include the patient's gender, age, lesion range, peripheral dose, central dose, number of targets, isodose lines, surgical history and whether fractionated treatment is performed.
更进一步的，所述临床数据获取单元还用以统计临床资料并将变量进行转换；所述转换包括将临床资料中外科手术史和是否分次治疗转换为二分类变量。Furthermore, the clinical data acquisition unit is also used to count clinical data and convert variables; the conversion includes converting the surgical history and whether fractionated treatment in the clinical data into binary variables.
进一步的，影像组学特征提取单元中，所述头颅磁共振常规序列包括T1增强、T2和ADC序列。Further, in the radiomics feature extraction unit, the conventional cranial magnetic resonance sequence includes T1 enhanced, T2 and ADC sequences.
更进一步的，影像组学特征提取单元中，提取的特征包括灰度直方图灰度矩阵(GLSZM)、形状因数(form factor)、Haralick、灰度共生矩阵(GLCM)和游程矩阵(RLM)，最终在每种序列提取若干个影像组学特征。Furthermore, in the radiomics feature extraction unit, the extracted features include grayscale histogram grayscale matrix (GLSZM), form factor (form factor), Haralick, grayscale co-occurrence matrix (GLCM) and run length matrix (RLM). Finally, several radiomic features are extracted from each sequence.
进一步的，影像学语义特征获取单元中，影像学语义特征包括肿瘤位置、边界是否规则、在增强T1磁共振上肿瘤是否均匀强化、肿瘤内有无血管、肿瘤内有无囊肿或坏死成分、硬脑膜尾征以及伽马刀治疗前是否有瘤周水肿。Furthermore, in the imaging semantic feature acquisition unit, the imaging semantic features include the location of the tumor, whether the boundaries are regular, whether the tumor is uniformly enhanced on enhanced T1 magnetic resonance, whether there are blood vessels in the tumor, whether there are cysts or necrosis components in the tumor, and whether the tumor is solid. Meningeal tail sign and whether there is peritumoral edema before gamma knife treatment.
更进一步的，所述影像学语义特征获取单元还用于统计影像学语义特征并将变量进行转换；所述转换包括将肿瘤位置分为是否位于矢旁、是否位于颅底两个二分类变量。Furthermore, the imaging semantic feature acquisition unit is also used to count imaging semantic features and convert variables; the conversion includes dividing the tumor location into two binary variables: whether it is located parasagittal and whether it is located at the skull base.
进一步的，预测模型建立单元中，将患者按7：3随机分为训练集和测试集。Further, in the prediction model building unit, patients were randomly divided into training sets and test sets in a ratio of 7:3.
进一步的，所述预测模型建立单元采用累积/动态时依ROC曲线的曲线下积分面积(iAUC)，评估不同随机生存森林模型对脑膜瘤伽马刀术后水肿的发生的预测效果，生成最优模型。Furthermore, the prediction model building unit uses the integrated area under the curve (iAUC) of the cumulative/dynamic ROC curve to evaluate the prediction effect of different random survival forest models on the occurrence of edema after gamma knife surgery for meningioma and generate the optimal model.
进一步的，所述预测模型建立单元还用于将最优模型的预测风险分数绘制的列线图用于直观地显示水肿在不同时间的发生率。用于辅助指导临床决策。Furthermore, the prediction model building unit is also used to draw a nomogram of the predicted risk scores of the optimal model to intuitively display the incidence of edema at different times, so as to assist in guiding clinical decision-making.
另一方面，本发明还提供一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述计算机程序被所述处理器执行时，使得所述处理器执行如下步骤：On the other hand, the present invention also provides a computer device, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the computer program causes the processor to perform the following steps:
S1、获取行伽马刀治疗的脑膜瘤患者的临床资料、治疗前及治疗后随访的头颅磁共振，并形成临床数据集用于预测系统的建立；S1. Obtain the clinical data of meningioma patients who underwent gamma knife treatment, and the brain magnetic resonance imaging before and after treatment, and form a clinical data set for the establishment of a prediction system;
S2、从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间，并形成结局数据集用于预测系统的建立；S2. Obtain the occurrence and occurrence time of postoperative peritumoral edema from the patient's postoperative follow-up cranial magnetic resonance imaging, and form an outcome data set for the establishment of a prediction system;
S3、从患者术前头颅磁共振常规序列(T1增强、T2、ADC)脑膜瘤区域进行影像组学特征提取及筛选，并形成影像组学特征数据集用于预测系统的建立；S3. Extract and screen radiomics features from the meningioma area of the patient's preoperative cranial magnetic resonance conventional sequence (T1 enhancement, T2, ADC), and form a radiomics feature data set for the establishment of a prediction system;
S4、分析患者术前磁共振常规序列获取影像学特征，并形成影像学语义特征数据集用于预测系统的建立；S4. Analyze the patient's preoperative magnetic resonance routine sequence to obtain imaging features, and form an imaging semantic feature data set for the establishment of a prediction system;
S5、将患者按7：3随机分为训练集合测试集，根据训练集数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证，评估其对脑膜瘤伽马刀术后水肿的发生的预测效果，生成最优模型。S5. Randomly divide the patients into training sets and test sets in a ratio of 7:3, establish a series of random survival forest models containing different types of features based on the training set data, and verify them in the test set to evaluate their effect on meningioma after gamma knife surgery. Predict the occurrence of edema and generate the optimal model.
另一方面，本发明还提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行上述S1-S5步骤。On the other hand, the present invention also provides a computer-readable storage medium storing a computer program. The computer program includes program instructions. When executed by a processor, the program instructions cause the processing to occur. The device performs the above steps S1-S5.
与现有技术相比，本发明具有如下有益效果：Compared with the prior art, the present invention has the following beneficial effects:
1)本发明提供了一种基于影像组学的脑膜瘤伽马刀后瘤周水肿预测系统，包括对各患者术前头颅磁共振常规序列(T1增强、T2、ADC等)肿瘤区域进行影像组学特征提取；将各患者影像组学特征中的冗余特征去除以得到筛选后的重要影像组学特征；结合筛选后的影像组学特征、临床特征及肉眼获取的影像学特征等，根据水肿发生与否及发生时间得到临床影像特征与术后水肿发生率及发生期之间的关系，建立随机生存森林模型用于预测；1) The present invention provides a system for predicting peritumoral edema after gamma knife surgery for meningioma based on radiomics, including extracting radiomics features from the tumor area of conventional cranial magnetic resonance imaging sequences (T1 enhancement, T2, ADC, etc.) of each patient before surgery; removing redundant features from the radiomics features of each patient to obtain important radiomics features after screening; combining the screened radiomics features, clinical features, and imaging features obtained by naked eyes, etc., obtaining the relationship between clinical imaging features and the incidence and period of postoperative edema according to whether edema occurs and when it occurs, and establishing a random survival forest model for prediction;
2)本发明的术后水肿预测系统，通过提取肿瘤区域的影像学特征结合临床特征，具有无创性、可重复、易操作的优点，通过机器学习技术特征，实现了脑膜瘤患者伽马刀术后水肿的发生率，对预后、改善临床决策提供有力支持，实现了医师可根据模型判断的情况决定是否予伽马刀治疗的有益效果。2) The postoperative edema prediction system of the present invention, by extracting the imaging features of the tumor area and combining them with clinical features, has the advantages of being non-invasive, repeatable, and easy to operate. Through the characteristics of machine learning technology, it realizes the incidence of edema after gamma knife surgery in patients with meningioma, provides strong support for prognosis and improvement of clinical decision-making, and achieves the beneficial effect that doctors can decide whether to give gamma knife treatment based on the situation judged by the model.
附图说明Description of drawings
通过阅读参照以下附图对非限制性实施例所作的详细描述，本发明的其它特征、目的和优点将会变得更明显：Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:
图1是本发明一个实施例的基于影像组学的脑膜瘤伽马刀后水肿预测系统应用的流程示意图；FIG1 is a schematic diagram of a flow chart of an application of a system for predicting edema after gamma knife surgery for meningioma based on radiomics according to an embodiment of the present invention;
图2是本发明一个实施例的iAUC最高的模型的时依AUC曲线；Figure 2 is a time-dependent AUC curve of the model with the highest iAUC according to an embodiment of the present invention;
图3中的A是本发明一个实施例的列线图，B是校正曲线；A in Figure 3 is a nomogram of an embodiment of the present invention, and B is a calibration curve;
图4是本发明一个实施例的基于影像组学的脑膜瘤伽马刀后水肿预测系统的系统架构图；Figure 4 is a system architecture diagram of a radiomics-based post-gamma knife edema prediction system for meningioma according to one embodiment of the present invention;
图5是本发明一个实施例的一种电子设备的结构示意图。Figure 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
具体实施方式Detailed ways
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.
在以下的描述中，涉及到“一些实施例”，其描述了所有可能实施例的子集，但是可以理解，“一些实施例”可以是所有可能实施例的相同子集或不同子集，并且可以在不冲突的情况下相互结合。In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or a different subset of all possible embodiments, and Can be combined with each other without conflict.
除非另有定义，本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中所使用的术语只是为了描述本申请实施例的目的，不是旨在限制本申请。Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.
本发明一个实施例的基于影像组学的脑膜瘤伽马刀后水肿预测系统包括：A radiomics-based post-gamma knife edema prediction system for meningioma according to one embodiment of the present invention includes:
临床数据获取单元：--用以获取行伽马刀治疗的脑膜瘤患者的临床资料、治疗前及治疗后随访的头颅磁共振图像信息，并形成临床数据集；Clinical data acquisition unit: used to obtain clinical data of meningioma patients treated with gamma knife, cranial magnetic resonance image information before and after treatment, and form a clinical data set;
结局判定单元：--用以从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间，并形成结局数据集；Outcome determination unit: - used to obtain the occurrence and time of postoperative peritumoral edema from the patient's postoperative follow-up cranial MRI and form an outcome data set;
影像组学特征提取单元：--用以从患者术前头颅磁共振常规序列脑膜瘤区域进行影像组学特征提取及筛选，并形成影像组学特征数据集；Radiomics feature extraction unit: - Used to extract and screen radiomics features from the meningioma area of the patient's preoperative cranial magnetic resonance conventional sequence, and form a radiomics feature data set;
影像学语义特征获取单元：--用以解读分析患者术前磁共振常规序列脑膜瘤区域图像获取影像学语义特征，并形成影像学语义特征数据集；Imaging semantic feature acquisition unit: - used to interpret and analyze patients' preoperative magnetic resonance conventional sequence meningioma area images to obtain imaging semantic features, and form an imaging semantic feature data set;
预测模型建立单元：--用以将患者随机分为训练集和测试集，根据训练集的结局数据，结合其临床数据、影像组学特征数据和影像学语义特征数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证，评估其对脑膜瘤伽马刀术后水肿的发生率及发生时间的预测效果，生成最优模型；Prediction model building unit: - used to randomly divide patients into training sets and test sets, and establish a series of different types of features based on the outcome data of the training set, combined with its clinical data, radiomics feature data and imaging semantic feature data The random survival forest model was verified in the test set to evaluate its prediction effect on the incidence and time of edema after gamma knife surgery for meningioma and generate the optimal model;
输出单元：--用以输出最优模型得到的脑膜瘤伽马刀后瘤周水肿的发生率及发生时间的评估预测值。Output unit: - Used to output the estimated prediction value of the incidence rate and occurrence time of peritumoral edema after gamma knife meningioma obtained by the optimal model.
图1为本发明该实施例的基于影像组学的脑膜瘤伽马刀后水肿预测系统的应用流程图；如图1所示，应用本发明的脑膜瘤伽马刀后水肿预测系统的预测方法包括以下步骤：Figure 1 is an application flow chart of the post-gamma knife edema prediction system for meningioma based on radiomics in this embodiment of the present invention; as shown in Figure 1, the prediction method of the post-gamma knife edema prediction system for meningioma using the present invention Includes the following steps:
S1：获取行伽马刀治疗的脑膜瘤患者的临床资料、治疗前及治疗后随访的头颅磁共振；S1: Obtain the clinical data of meningioma patients who underwent gamma knife treatment, and the brain magnetic resonance imaging before and after treatment;
S2：从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间；S2: The occurrence and time of postoperative peritumoral edema were obtained from the patients’ postoperative follow-up cranial MRI;
S3：从患者术前头颅磁共振常规序列(T1增强、T2、ADC)脑膜瘤区域进行影像组学特征提取及筛选；S3: Extract and screen radiomic features from the meningioma area of the patient's preoperative cranial magnetic resonance conventional sequence (T1 enhancement, T2, ADC);
S4：通过肉眼分析患者术前磁共振常规序列获取影像学特征；S4: Obtain imaging features by visually analyzing the patient's preoperative magnetic resonance routine sequence;
S5：将患者按7：3随机分为训练集合测试集，根据训练集数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证，评估其对脑膜瘤伽马刀术后水肿的发生的预测效果，生成最优模型。S5: Randomly divide the patients into a training set and a test set in a ratio of 7:3, establish a series of random survival forest models containing different types of features based on the training set data, and verify them in the test set to evaluate their effect on meningioma after gamma knife surgery. Predict the occurrence of edema and generate the optimal model.
在步骤S1中，所述脑膜瘤患者的临床资料包括患者的性别、年龄、病变范围、周边剂量、中心剂量、靶点数、等剂量线，外科手术史和是否分次治疗。In step S1, the clinical data of the meningioma patient includes the patient's gender, age, lesion range, peripheral dose, central dose, number of targets, isodose lines, surgical history and whether fractionated treatment is performed.
在一个具体的实施例中，共纳入445名患者的性别、年龄、病变范围、周边剂量、中心剂量、靶点数、等剂量线，外科手术史和是否分次治疗，共9个临床特征。445例患者的平均年龄为54.7±11.4岁，其中女性343例(77.1％)，男性102例(22.9％)。同时收集此445位患者治疗前及治疗后随访的头颅磁共振图像信息。In a specific embodiment, a total of 9 clinical characteristics of 445 patients were included: gender, age, lesion extent, peripheral dose, central dose, number of targets, isodose lines, surgical history and whether fractionated treatment was performed. The average age of the 445 patients was 54.7±11.4 years, including 343 women (77.1%) and 102 men (22.9%). At the same time, the cranial magnetic resonance image information of these 445 patients before and after treatment was collected.
在步骤S2中，从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间。瘤周水肿定义为T2序列磁共振上脑膜瘤周围的高密度影。水肿的发生定义为术前无瘤周水肿的患者在术后出现新发水肿，或术前有水肿的患者在术后出现水肿的进展。发生时间为伽马刀日期到磁共振发现新发水肿或水肿进展的日期。In step S2, the occurrence and occurrence time of postoperative peritumoral edema are obtained from the patient's postoperative follow-up cranial magnetic resonance imaging. Peritumoral edema is defined as the high-density shadow surrounding the meningioma on T2 sequence magnetic resonance imaging. The occurrence of edema was defined as new postoperative edema in patients without preoperative peritumoral edema, or progression of postoperative edema in patients with preoperative edema. The time of occurrence was from the date of gamma knife to the date of new edema or progression of edema detected on magnetic resonance imaging.
在一个具体的实施例中，43例脑膜瘤患者出现了新发水肿，33例有术前水肿的进展，总归76例(17.1％)。其余369例未发生新发水肿或术前水肿无进展。In one specific example, 43 meningioma patients developed new onset edema, and 33 had progression of preoperative edema, for a total of 76 patients (17.1%). The remaining 369 cases did not develop new edema or had no progression of preoperative edema.
在步骤S3中，从患者术前头颅磁共振常规序列(T1增强、T2、ADC)脑膜瘤区域进行影像组学特征提取及筛选。In step S3, radiomics features are extracted and screened from the meningioma area of the patient's preoperative cranial magnetic resonance conventional sequence (T1 enhanced, T2, ADC).
在一个具体的实施例中，提取的特征包括灰度直方图灰度矩阵(GLSZM)、形状因数(form factor)、Haralick、灰度共生矩阵(GLCM)和游程矩阵(RLM)，最终在每种序列提取了396个影像组学特征(共1188个)。In a specific embodiment, the extracted features include grayscale histogram grayscale matrix (GLSZM), form factor, Haralick, gray level co-occurrence matrix (GLCM) and run length matrix (RLM), and finally 396 imaging omics features (a total of 1188) were extracted in each sequence.
通过variable hunting方法根据特征的重要性进行特征的筛选，最终T1增强序列筛选出16个特征；T2序列筛选出16个特征；ADC图筛选出17个特征。另将T1争取、T2以及ADC图所有特征组合后筛选出22个特征。具体如下表1。The variable hunting method was used to screen features according to their importance. Finally, 16 features were screened out by the T1 enhanced sequence; 16 features were screened out by the T2 sequence; and 17 features were screened out by the ADC chart. In addition, 22 features were selected after combining all the features of T1, T2 and ADC charts. The details are as follows in Table 1.
表1.variable hunting方法根据特征的重要性筛选出的特征Table 1. Features filtered out by the variable hunting method based on feature importance
在步骤S4中，影像医师通过肉眼分析患者术前磁共振常规序列获取影像学特征(即语义特征，semantic features)，包括肿瘤位置，边界是否规则，在增强T1磁共振上肿瘤是否均匀强化，肿瘤内有无血管，肿瘤内有无囊肿或坏死成分，硬脑膜尾征以及伽马刀治疗前是否有瘤周水肿。其中，肿瘤位置分为是否位于矢旁、是否位于颅底两个二分类变量。因此影像学特征共8个。In step S4, the radiologist obtains imaging features (i.e., semantic features) by visually analyzing the patient's preoperative magnetic resonance conventional sequence, including the location of the tumor, whether the boundaries are regular, whether the tumor is uniformly enhanced on the enhanced T1 magnetic resonance, and whether the tumor is evenly enhanced on the enhanced T1 magnetic resonance. Whether there are blood vessels in the tumor, whether there are cysts or necrotic components in the tumor, dural tail sign, and whether there is peritumoral edema before gamma knife treatment. Among them, the tumor location is divided into two binary variables: whether it is located parasagitally and whether it is located at the skull base. Therefore, there are 8 imaging features in total.
在步骤S5中，将患者按7：3随机分为训练集合测试集，根据训练集数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证。In step S5, the patients are randomly divided into a training set and a test set in a ratio of 7:3, and a series of random survival forest models containing different types of features are established based on the training set data and verified in the test set.
在一个具体的实施例中，训练集包括了312例脑膜瘤患者，测试集则包括了133例。训练集及测试集各个基线临床资料之间无显著性差异。In a specific embodiment, the training set includes 312 patients with meningioma, and the test set includes 133. There is no significant difference in the baseline clinical data between the training set and the test set.
本实施例共训练了19个随机生存森林(RSF)模型。这些模型包括一层(模型1-T1、1-T2、1-ADC、1-Rad、1-C、1-S)、两层(模型2-CS、2-CT1、2-CT2、2-CADC、2-CRad、2-ST1、2-ST2、2-SADC、2-SRad)和三层(模型3-CST1、3-CST2、3-CSADC、3-CSRad)模型，分别包含影像组学、临床特征和影像学语义特征的不同组合。每个模型都是基于以下参数进行训练的：模型1-T1:从396个特征中选择的16个T1增强序列的组学特征；模型1-T2:使用相同方法从396个特征中选择的16个T2序列的组学特征；模型1-ADC：从396个提取特征中选择的17个ADC图的组学特征；模型1-Rad：从三个序列提取的1188个特征中选择的22个组学特征；模型1-C：9个临床特征；模型1-S：8个影像学语义特征；模型2-CS：结合17个临床和影像学语义特征；模型2-CT1：临床特征和选择的T1增强组学特征(共25个)；模型2-CT2：临床特征和选择的T2序列组学特征(共25个)；模型2-CADC：临床特征和选择的ADC序列组学特征(共26个)；模型2-CRad：临床特征和1-Rad的22个组学特征；模型2-ST1：语义特征和选择的T1增强组学特征(共24个)；模型2-ST2：语义特征和选择的T2组学特征(共24个)；模型2-SADC：语义特征和选定的ADC组学特征(共25个)；模型2-SRad：语义特征和1-Rad的22个组学特征(30个)；模型3-CST1：临床、语义和选择的T1增强组学特征(33个)；模型3-CST2：临床、语义和选择的T2组学特征(33个)；模型3-CSADC：临床、语义和选择的ADC组学特征(34个)；模型3-CSRad：临床、语义，以及1-Rad的22个组学特征(39个)。In this embodiment, a total of 19 Random Survival Forest (RSF) models were trained. These models include one layer (model 1-T1, 1-T2, 1-ADC, 1-Rad, 1-C, 1-S), two layers (model 2-CS, 2-CT1, 2-CT2, 2- CADC, 2-CRad, 2-ST1, 2-ST2, 2-SADC, 2-SRad) and three-layer (model 3-CST1, 3-CST2, 3-CSADC, 3-CSRad) models, respectively containing radiomics , different combinations of clinical features and imaging semantic features. Each model was trained based on the following parameters: Model 1-T1: omics features of 16 T1-enhanced sequences selected from 396 features; Model 1-T2: 16 features selected from 396 features using the same method Omic features of T2 sequences; Model 1-ADC: Omic features of 17 ADC maps selected from 396 extracted features; Model 1-Rad: 22 groups selected from 1188 features extracted from three sequences model 1-C: 9 clinical features; Model 1-S: 8 imaging semantic features; Model 2-CS: combines 17 clinical and imaging semantic features; Model 2-CT1: clinical features and selected T1 enhanced omics features (25 in total); Model 2-CT2: clinical features and selected T2 sequence omics features (25 in total); Model 2-CADC: clinical features and selected ADC sequence omics features (26 in total) ); Model 2-CRad: clinical features and 22 omics features of 1-Rad; Model 2-ST1: semantic features and selected T1 enhanced omics features (24 in total); Model 2-ST2: semantic features and Selected T2 omics features (24 in total); Model 2-SADC: semantic features and selected ADC omics features (25 in total); Model 2-SRad: semantic features and 22 omics features of 1-Rad (30); Model 3-CST1: clinical, semantic and selected T1 enhanced omics features (33); Model 3-CST2: clinical, semantic and selected T2 omics features (33); Model 3-CSADC : clinical, semantic and selected ADC omics features (34); Model 3-CSRad: clinical, semantic, and 22 omics features of 1-Rad (39).
本实施例采用累积/动态时依ROC曲线的曲线下积分面积(iAUC)，评估不同随机生存森林模型对脑膜瘤伽马刀术后水肿的发生的预测效果，生成最优模型。This embodiment uses the integrated area under the curve (iAUC) of the ROC curve in cumulative/dynamic time to evaluate the prediction effect of different random survival forest models on the occurrence of edema after gamma knife surgery for meningioma, and generates the optimal model.
图2是iAUC最高的7个模型的时依AUC曲线。其中iAUC最高模型为模型3-CST1，该模型iAUC达0.942(95％可信区间:0.939–0.944)，c-index为0.964±0.002,Brier score:为0.131。Figure 2 shows the time-dependent AUC curves of the seven models with the highest iAUC. Among them, the model with the highest iAUC is model 3-CST1. The iAUC of this model reaches 0.942 (95% confidence interval: 0.939–0.944), c-index is 0.964±0.002, and Brier score: is 0.131.
本实施例还提供基于最优模型的预测风险分数绘制的列线图用于直观地显示水肿在不同时间的发生率，用于指导临床决策。见图3。图3A为列线图，提示0.5年、1年、1.5年和2年时水肿的发生率，图3B为不同时间的校正曲线，得到偏差校正估计。This embodiment also provides a nomogram drawn based on the predicted risk score of the optimal model to visually display the incidence of edema at different times and to guide clinical decision-making. See Figure 3. Figure 3A is a nomogram, indicating the incidence of edema at 0.5 years, 1 year, 1.5 years, and 2 years. Figure 3B is the calibration curve at different times to obtain bias-corrected estimates.
本发明通过构建基于影像组学的脑膜瘤伽马刀后瘤周水肿预测随机生存森林模型，选出最优模型，具有无创性、可重复、易操作的优点，可为评估行伽马刀的脑膜瘤患者预后、改善临床决策提供有力支持。The present invention constructs a random survival forest model for predicting peritumoral edema after gamma knife surgery for meningiomas based on radiomics and selects the optimal model. It has the advantages of non-invasiveness, repeatability, and easy operation, and can be used for the evaluation of gamma knife surgery. It provides strong support for the prognosis of meningioma patients and improving clinical decision-making.
图4是本发明一个实施例的脑膜瘤伽马刀后瘤周水肿预测系统架构图。如图4所示，一种基于影像组学的脑膜瘤伽马刀后瘤周水肿预测的初筛诊断系统，该系统实现基于影像组学的脑膜瘤伽马刀后瘤周水肿预测的步骤如上述方法所示，包括：Figure 4 is an architecture diagram of a post-gamma knife meningioma peritumoral edema prediction system according to one embodiment of the present invention. As shown in Figure 4, a preliminary screening diagnostic system for predicting peritumoral edema after gamma knife meningioma based on radiomics. The steps for this system to predict peritumoral edema after gamma knife for meningiomas based on radiomics are as follows: The above methods include:
数据获取单元，用以获取行伽马刀治疗的脑膜瘤患者的临床资料、治疗前及治疗后随访的头颅磁共振；Data acquisition unit, used to obtain clinical data of meningioma patients treated with gamma knife, and cranial MRI before and after treatment;
结局判定单元，用以从患者术后随访头颅磁共振获取术后瘤周水肿的发生情况及发生时间；The outcome determination unit is used to obtain the occurrence and time of postoperative peritumoral edema from the patient's postoperative follow-up cranial magnetic resonance imaging;
(影像组学)特征提取单元，用以从患者术前头颅磁共振常规序列(T1增强、T2、ADC)脑膜瘤区域进行影像组学特征提取及筛选；(Radiomics) feature extraction unit, used to extract and screen radiomics features from the meningioma area of the patient's preoperative cranial magnetic resonance conventional sequence (T1 enhancement, T2, ADC);
(影像学)语义特征获取单元，用以通过肉眼分析患者术前磁共振常规序列获取影像学特征；(Imaging) semantic feature acquisition unit, used to acquire imaging features by visually analyzing the patient's preoperative magnetic resonance conventional sequence;
预测模型建立单元，将患者按7：3随机分为训练集合测试集，根据训练集数据建立一系列包含不同类型特征的随机生存森林模型，并在测试集中进行验证，评估其对脑膜瘤伽马刀术后水肿的发生的预测效果，生成最优模型。The prediction model building unit randomly divides patients into training sets and test sets in a ratio of 7:3, establishes a series of random survival forest models containing different types of features based on the training set data, and verifies them in the test set to evaluate their effect on meningioma gamma. Predict the occurrence of postoperative edema and generate the optimal model.
图5是本发明一个实施例的一种电子设备的结构示意图。如图5所示，本发明一个实施例的一种电子设备包括一个或多个输入设备、一个或多个输出设备、一个或多个处理器和存储器。Figure 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. As shown in FIG. 5, an electronic device according to one embodiment of the present invention includes one or more input devices, one or more output devices, one or more processors and memories.
在本发明一个实施例中，处理器、输入设备、输出设备和存储器可以通过总线或其它方式连接。输入设备、输出设备可以是标准的有线或无线通信接口。In one embodiment of the present invention, the processor, input device, output device and memory may be connected through a bus or other means. Input devices and output devices can be standard wired or wireless communication interfaces.
处理器可以是中央处理模块(Central Processing Unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。The processor can be a central processing module (Central Processing Unit, CPU), which can also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Ready-made field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
存储器可以是高速RAM存储器，也可为非不稳定的存储器，例如磁盘存储器。存储器用于存储一组计算机程序，输入设备、输出设备和处理器可以调用存储器中存储的程序代码。The memory can be high-speed RAM memory or non-volatile memory, such as disk memory. Memory is used to store a set of computer programs, and input devices, output devices, and processors can call the program code stored in the memory.
存储器存储的计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行如上述实施例中所述基于影像组学的脑膜瘤伽马刀后瘤周水肿预测的步骤。The computer program stored in the memory includes program instructions that, when executed by the processor, cause the processor to perform the steps of predicting peritumoral edema after gamma knife meningioma based on radiomics as described in the above embodiments.
本发明的一个实施例还提供一种计算机可读存储介质。该计算机可读存储介质可以是高速RAM存储器，也可为非不稳定的存储器，例如磁盘存储器。该计算机可读存储介质可通过外部计算设备或网络进行连接，以读取该计算机可读存储介质所存储的一组计算机程序。该计算机可读存储介质存储的计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行如上述实施例中所述基于影像组学的脑膜瘤伽马刀后瘤周水肿预测方法的步骤。An embodiment of the present invention also provides a computer-readable storage medium. The computer-readable storage medium may be a high-speed RAM memory or a non-volatile memory such as a magnetic disk memory. The computer-readable storage medium can be connected through an external computing device or a network to read a set of computer programs stored in the computer-readable storage medium. The computer program stored in the computer-readable storage medium includes program instructions, which when executed by the processor cause the processor to perform radiomics-based post-gamma knife peritumoral meningioma surgery as described in the above embodiments. Steps in the edema prediction method.
尽管上面已经示出和描述了本发明的实施例，可以理解的是，上述实施例是示例性的，不能理解为对本发明的限制，本领域的普通技术人员在本发明的范围内可以对上述实施例进行变化、修改、替换和变型。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。Although the embodiments of the present invention have been shown and described above, it can be understood that the above-mentioned embodiments are illustrative and should not be construed as limitations of the present invention. Those of ordinary skill in the art can make modifications to the above-mentioned embodiments within the scope of the present invention. The embodiments are subject to changes, modifications, substitutions and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110913967.9ACN113643805B (en) | 2021-08-10 | 2021-08-10 | Radiomics-based edema prediction system after gamma knife treatment for meningioma |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110913967.9ACN113643805B (en) | 2021-08-10 | 2021-08-10 | Radiomics-based edema prediction system after gamma knife treatment for meningioma |
| Publication Number | Publication Date |
|---|---|
| CN113643805A CN113643805A (en) | 2021-11-12 |
| CN113643805Btrue CN113643805B (en) | 2024-04-02 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110913967.9AActiveCN113643805B (en) | 2021-08-10 | 2021-08-10 | Radiomics-based edema prediction system after gamma knife treatment for meningioma |
| Country | Link |
|---|---|
| CN (1) | CN113643805B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116705296B (en)* | 2023-06-06 | 2024-09-13 | 中国科学院深圳先进技术研究院 | Method and system for risk stratification of GBM patient based on conventional MRI sequence |
| CN117976225B (en)* | 2024-03-05 | 2025-01-21 | 齐鲁工业大学(山东省科学院) | Method, system, storage medium and device for predicting hematoma change probability |
| CN119517388B (en)* | 2025-01-23 | 2025-07-01 | 中国人民解放军空军军医大学 | A multi-task based brain edema progression prediction system and method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815481A (en)* | 2017-01-19 | 2017-06-09 | 中国科学院深圳先进技术研究院 | A kind of life cycle Forecasting Methodology and device based on image group |
| CN111657945A (en)* | 2020-06-16 | 2020-09-15 | 中南大学湘雅医院 | Nasopharyngeal carcinoma prognosis auxiliary evaluation method based on enhanced MRI (magnetic resonance imaging) imaging omics |
| CN112185569A (en)* | 2020-09-11 | 2021-01-05 | 中山大学孙逸仙纪念医院 | Breast cancer patient disease-free survival period prediction model and construction method thereof |
| CN112768072A (en)* | 2021-01-12 | 2021-05-07 | 哈尔滨医科大学 | Cancer clinical index evaluation system constructed based on imaging omics qualitative algorithm |
| CN112951406A (en)* | 2021-01-27 | 2021-06-11 | 安徽理工大学 | Lung cancer prognosis auxiliary evaluation method and system based on CT (computed tomography) image omics |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200058125A1 (en)* | 2018-08-14 | 2020-02-20 | Mikhail Teverovskiy | Comparative cancer survival models to assist physicians to choose optimal treatment |
| US10971255B2 (en)* | 2018-09-14 | 2021-04-06 | Zasti Inc. | Multimodal learning framework for analysis of clinical trials |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815481A (en)* | 2017-01-19 | 2017-06-09 | 中国科学院深圳先进技术研究院 | A kind of life cycle Forecasting Methodology and device based on image group |
| CN111657945A (en)* | 2020-06-16 | 2020-09-15 | 中南大学湘雅医院 | Nasopharyngeal carcinoma prognosis auxiliary evaluation method based on enhanced MRI (magnetic resonance imaging) imaging omics |
| CN112185569A (en)* | 2020-09-11 | 2021-01-05 | 中山大学孙逸仙纪念医院 | Breast cancer patient disease-free survival period prediction model and construction method thereof |
| CN112768072A (en)* | 2021-01-12 | 2021-05-07 | 哈尔滨医科大学 | Cancer clinical index evaluation system constructed based on imaging omics qualitative algorithm |
| CN112951406A (en)* | 2021-01-27 | 2021-06-11 | 安徽理工大学 | Lung cancer prognosis auxiliary evaluation method and system based on CT (computed tomography) image omics |
| Title |
|---|
| 低级别胶质瘤的机器学习和影像组学表型:提高生存预测;Y.S.Choi;S.S.Ahn;J.H.Chang;S.G.Kang;E.H.Kim;S.H.Kim;赵海清;;国际医学放射学杂志(第05期);全文* |
| 基于深度学习的时间飞跃法磁共振血管成像对脑动脉瘤自动检测方法研究;耿辰;上海医学(第5期);全文* |
| Publication number | Publication date |
|---|---|
| CN113643805A (en) | 2021-11-12 |
| Publication | Publication Date | Title |
|---|---|---|
| CN113643805B (en) | Radiomics-based edema prediction system after gamma knife treatment for meningioma | |
| US10939874B2 (en) | Deep-learning-based cancer classification using a hierarchical classification framework | |
| US11263793B2 (en) | Method and system for assessing images using biomarkers | |
| Zhang et al. | Radiomics analysis for the differentiation of autoimmune pancreatitis and pancreatic ductal adenocarcinoma in 18F‐FDG PET/CT | |
| US20180242906A1 (en) | Predicting immunotherapy response in non-small cell lung cancer with serial quantitative vessel tortuosity | |
| EP2510468B1 (en) | Diagnostic techniques for continuous storage and joint analysis of both image and non-image medical data | |
| US20210241178A1 (en) | Computationally derived cytological image markers for predicting risk of relapse in acute myeloid leukemia patients following bone marrow transplantation images | |
| WO2021135774A1 (en) | Tumor prediction method and device, cloud platform, and computer-readable storage medium | |
| CN114549463B (en) | A method, system, device and medium for predicting the efficacy of anti-HER-2 treatment for liver metastasis of breast cancer | |
| Pareek et al. | [Retracted] Predicting the Spread of Vessels in Initial Stage Cervical Cancer through Radiomics Strategy Based on Deep Learning Approach | |
| CN114664409B (en) | A method, system, device and storage medium for processing radiomics data of recurrent nasopharyngeal carcinoma | |
| Guo et al. | Machine learning and new insights for breast cancer diagnosis | |
| Liu et al. | A hybrid classification model with radiomics and CNN for high and low grading of prostate cancer Gleason score on mp-MRI | |
| Miao et al. | Application of deep learning and XGBoost in predicting pathological staging of breast cancer MR images. | |
| Amiri et al. | A machine learning approach for prediction of auditory brain stem response in patients after head-and-neck radiation therapy | |
| Jiang et al. | Machine Learning-Based Prediction of Pathological Responses and Prognosis After Neoadjuvant Chemotherapy for Non–Small-Cell Lung Cancer: A Retrospective Study | |
| Liu et al. | Prediction of clinical outcome for high-intensity focused ultrasound ablation of adenomyosis based on non-enhanced MRI radiomics | |
| Choi et al. | identification of CT texture features robust to tumor size variations for normal lung texture analysis | |
| Lee et al. | Radiomics-based sialadenitis staging in contrast-enhanced computed tomography and ultrasonography: a preliminary rat model study | |
| Leitch et al. | Placenta accreta spectrum and hysterectomy prediction using MRI radiomic features | |
| Kulavuz et al. | Breast lesion detection from DCE-MRI using YOLOv7 | |
| Mahmoudiandehkordi et al. | Ovarian Cancer Diagnosis Using Advanced Deep Learning Models with Histopathology Images | |
| Wang et al. | A machine learning model based on placental magnetic resonance imaging and clinical factors to predict fetal growth restriction | |
| US20240161301A1 (en) | Predicting response and prognosis to cdk 4/6 inhibitors based on tumor vascularization and vessel shape | |
| US20220405918A1 (en) | Computational features of tumor-infiltrating lymphocyte (til) architecture |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |