WO2023240819A1 - 一种甲状腺疾病病理分析模块 - Google Patents

一种甲状腺疾病病理分析模块 Download PDF

Info

Publication number
WO2023240819A1
WO2023240819A1 PCT/CN2022/120114 CN2022120114W WO2023240819A1 WO 2023240819 A1 WO2023240819 A1 WO 2023240819A1 CN 2022120114 W CN2022120114 W CN 2022120114W WO 2023240819 A1 WO2023240819 A1 WO 2023240819A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
analysis module
thyroid disease
thyroid
data
Prior art date
Application number
PCT/CN2022/120114
Other languages
English (en)
French (fr)
Inventor
王珣章
杨艳
黄皓辉
邓日强
谢伟东
Original Assignee
广州智睿医疗科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 广州智睿医疗科技有限公司 filed Critical 广州智睿医疗科技有限公司
Publication of WO2023240819A1 publication Critical patent/WO2023240819A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H70/00—ICT specially adapted for the handling or processing of medical references
    • G16H70/60—ICT specially adapted for the handling or processing of medical references relating to pathologies
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the invention belongs to the field of biological information, and specifically relates to a thyroid disease pathological analysis module combination.
  • the digital pathology analysis method of thyroid slices includes: S1: Obtaining a digital pathology image set of thyroid slices; S2: According to the digital pathology image of the thyroid slices Set, use the first image processing method to obtain the first probability data of the thyroid slice; S3: Obtain a hyperspectral image sample set of the thyroid slice; S4: According to the hyperspectral image sample set of the thyroid slice, use the second image
  • the processing method is to obtain the second probability data of the thyroid section; S5: obtain the digital pathology information processing result of the thyroid section according to the first probability data and the second probability data.
  • S1 Obtaining a digital pathology image set of thyroid slices
  • S2 According to the digital pathology image of the thyroid slices Set, use the first image processing method to obtain the first probability data of the thyroid slice
  • S3 Obtain a hyperspectral image sample set of the thyroid slice
  • S4 According to the hyperspectral image sample set of the thyroid slice, use the second image
  • the processing method is to obtain the second probability data of the thyroid section
  • S5
  • the present invention provides a thyroid pathology analysis module.
  • a thyroid pathology analysis module By conducting model training on clinical sample data, an analysis method dependent on electronic equipment is obtained to analyze the pathology of thyroid diseases.
  • module refers to a carrier used to perform detection and analysis functions, including both physical hardware parts and program software parts.
  • model generally refers to a bioinformatics model, which can be used to study the collection, processing, storage, dissemination, analysis and interpretation of biological information.
  • tissue generally refers to biological tissue, which is composed of a group of cells and intercellular substance with similar shapes and identical functions, and is called a tissue.
  • pathology refers to the process and principle of disease occurrence and development. That is, the causes of disease, the pathogenesis, and the structural, functional and metabolic changes and rules of cells, tissues and organs that occur during the disease process.
  • the present invention simulates pathologists' classification of thyroid cancer tissue pathological sections (gold standard: 2017 WHO thyroid tumor classification), and further predicts histological subtypes that may indicate poor prognosis for cases with PTC morphology.
  • Type tall cell subtype, columnar cell subtype, solid type and shoe-nail subtype.
  • this intelligent prediction module also provides a visual quantitative prediction report, which is of great clinical practical translation value.
  • the present invention provides a pathological analysis module for thyroid disease.
  • the described thyroid disease pathology analysis module includes a hardware part and a data analysis program part
  • the hardware part is a computer, and the data analysis part includes a classification preprocessing program and a model program.
  • the described classification preprocessing program is used for WSI image frames and annotations in tissue section data.
  • the model program is one or more of a tile model or an aggregation model.
  • the described thyroid disease pathological analysis module completes the analysis through the following steps:
  • the training, verification and testing methods of the tile model are as follows:
  • the specific parameters of the tile model obtained by training are: learning rate 0.001, epoch 150 rounds, Batch_size is 32, input the original image to obtain the score, binary cross entropy is used as the loss function, and backpropagation is used to update the parameters.
  • the features of the slice tile model are extracted and formed into a heat map of the entire slice. Quantification is performed based on whether counts appear in each area.
  • the quantitative results of the present invention are standardly defined by pathologist annotations. After the slice aggregation model was verified and tested with the validation set data, the DM-pathology thyroid cancer quantitative histological classification and evaluation system was finally formed.
  • the specific parameters of the verified aggregation model in this invention are: learning rate 0.001, momentum super parameter 0.9, epoch 2000 rounds, Batch_size set to 16, and the loss function counts classification error, detection error, and segmentation error optimization.
  • the present invention provides a pathological analysis system for thyroid disease.
  • the pathology analysis system includes the aforementioned pathology analysis module.
  • the pathological analysis system also includes a data acquisition module.
  • the data acquisition module is used to collect tissue slice data that meets preset requirements.
  • the application of the data acquisition module includes the preparation of tissue slices.
  • the invention is suitable for the classification of PTC and the identification of four highly invasive subtypes, and the classification and quantification of highly invasive subtypes reach the level of clinical auxiliary application.
  • the model of the present invention has the following characteristics: first, MIL is used to train the classifier in the tile model, and then the results are aggregated through the learning fusion model to aggregate the prediction scores of each small block in the WSI. On this basis, a new framework was developed, using MIL to train neural networks to obtain semantically rich feature representations, and then using these feature representations for RNN to integrate the information of the entire WSI, and report the final classification results and quantify them. change.
  • Model advantages quantification; high accuracy; more suitable for real-world data (combination of strong supervised learning + weak supervision model).
  • Embodiment 1 A thyroid pathology analysis module
  • the samples include:
  • WSI in Kfb format has been strictly sorted, and the slicing standards should be: complete slices, uniform thickness, proper surface stickers, bright dyeing, clear contrast, neat slices, and clear and correct labels;
  • Dye balancing Find the latent spatial representation of various types of tissues through sparse autoencoders, and then use them as features for clustering to distinguish different tissues for dye balancing.
  • the DM-pathology thyroid cancer quantitative histological classification and evaluation system is finally formed; the main parameters of the final aggregation model are: learning rate 0.001, momentum super parameter 0.9, epoch 2000 rounds, and Batch_size is set to 16.
  • the loss function calculates classification error, detection error, and segmentation error optimization. For multi-classification tasks, a softmax classifier is used to calculate the probability of each type of sample, and finally optimized through the cross-entropy loss function.
  • the tile model (AI quantification) of this embodiment has an accuracy of 94.1% for PTC typing results, a sensitivity of 94.7%, and a specificity of 93.8%; the aggregation model result (AI classification) has an accuracy of 98.2% for PTC typing results.
  • the sensitivity is 99.2% and the specificity is 88.2%.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Biomedical Technology (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Pathology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Image Analysis (AREA)

Abstract

本发明提供一种甲状腺疾病病理分析模块，包括硬件部分和数据分析程序部分；硬件部分为一台计算机，数据分析部分包括分类预处理程序和模型程序，模型程序为tile模型或aggregation模型中的一种或多种。本发明适用于PTC的分类达到临床辅助应用级别。

Description

一种甲状腺疾病病理分析模块 技术领域
本发明属于生物信息领域，具体涉及一种甲状腺疾病病理分析模块组合。
背景技术
甲状腺癌是最为常见的一种内分泌恶性肿瘤。HE切片形态学观察是病理诊断甲状腺癌的关键手段。构成比高达95％的甲状腺癌为分化型甲状腺癌(differentiated thyroid cancer，DTC)，其中甲状腺乳头状癌(papillary thyroid carcinoma，PTC)占分化型甲状腺癌的85％。PTC是指甲状腺滤泡上皮细胞起源、具有特征性PTC细胞核特征的恶性上皮性肿瘤。根据组织学特征，PTC分为14个亚型，侵袭性形态特征、甲状腺外浸润和淋巴结转移等提示肿瘤复发风险高。因此，甲状腺乳头状癌的早期确诊及恶性指标判定，对制定及时合理的治疗方案、挽救患者的生命具有重要意义。而临床工作中符合资质的病理医生的严重不足，导致其工作负担繁重，不但增加漏诊、误诊的风险，还延长病例报告发放时间。
中国专利202111242026.3中公开了一种甲状腺切片的数字病理信息处理方法，所述甲状腺切片的数字病理分析方法包括：S1：获取甲状腺切片的数字病理图像集；S2：根据所述甲状腺切片的数字病理图像集，利用第一图像处理方法，得到所述甲状腺切片的第一概率数据；S3：获取甲状腺切片的高光谱图像样本集；S4：根据所述甲状腺切片的高光谱图像样本集，利用第二图像处理方法，得到所述甲状腺切片的第二概率数据；S5：根据所述第一概率数据和所述第二概率数据，得到所述甲状腺切片的数字病理信息处理结果。但其操作步骤较为复杂，有进一步优化的空间。
中国专利201810318306.X中公开了一种甲状腺肿瘤超声图像识别方法及其装置，所述方法包括：选取甲状腺肿瘤超声图像中的肿瘤区域并扩增一定边缘范围后切割，进行良恶性标注，将切割下来的图像组成训练集；用训练集训练选定的卷积神经网络形成甲状腺肿瘤超声图像识别模型；获取待识别的甲状腺肿瘤超声图像，选取肿瘤区域并扩增一定边缘范围后，用所述甲状腺肿瘤超声图像识别模型进行良恶性识别。本发明所述方法及其装置用于辅助医生对甲状腺肿瘤的良恶性进行诊断，在甲状腺超声图像肿瘤良恶性检测试验中取得了超过90％的准确率，这对临床实际诊断具有重大的参考意义。但其仅能用于前期诊断，对于更为细致的病理诊断不适用，有进一步优化的空间。
发明内容
为了解决上述问题，本发明提供了一种甲状腺病理分析模块，通过对临床样本数据进行模型训练，获得了一种依附于电子设备的分析方法，对甲状腺疾病的病理进行分析。
本发明中，“模块”是指用于发挥检测分析作用的载体，既包括实体的硬件部分，也包括程序类软件部分。
本发明中，“模型”一般指生物信息学模型，可用于研究生物信息的采集、处理、存储、传播、分析和解释。
本发明中，“组织”一般指生物组织，由形态相似、功能相同的一群细胞和细胞间质组合起来，称为组织。
本发明中，“组织切片”一般指取一定大小的病变组织，制成病理切片，本领域通常将病变组织包埋在石蜡块里，用切片机切成薄片，染色后用显微镜进一步检查病变和病变的发生发展过程，最后作出病理诊断。
本发明中，“病理”即疾病发生发展的过程和原理。也就是疾病发生的原因、发病原理和疾病过程中发生的细胞、组织和器官的结构、功能和代谢方面的改变及其规律。
本发明通过开发深度学习预测系统模块模拟病理医生对甲状腺癌组织病理切片进行分类(金标准：2017年WHO甲状腺肿瘤分类)，对形态学为PTC的病例，进一步预测可能提示不良预后的组织学亚型：高细胞亚型、柱状细胞亚型、实体型及鞋钉亚型，并根据病理医生临床工作中，对应肿瘤成分达不到某一亚型的诊断标准，注明提示不良预后的组织学亚型所占比例的要求，本智能预测模块同样给出可视化定量预测报告，极具临床实用转化价值。
一方面，本发明提供了一种甲状腺疾病的病理分析模块。
所述的甲状腺疾病病理分析模块包括硬件部分和数据分析程序部分；
所述的硬件部分为一台计算机，所述的数据分析部分包括分类预处理程序、模型程序。
所述的分类预处理程序用于组织切片数据中的WSI图像画格和标注。
所述的模型程序为tile模型或aggregation模型中的一种或多种。
所述的甲状腺疾病病理分析模块通过以下步骤完成分析：
数据收集与清洗→数据扫描与预处理→数据在线标注与坐标输出→tile模型分析(二类四型：分别对应三类五型标签)→aggregation模型分析→报告输出。
所述的tile模型的训练验证及测试方法如下：
以现有基本网络模型为基础，结合病理图像识别的特点，充分考虑层数、深度及超参数调优，搭建多种卷积神经网络模型(包括se_resnet101、se_resnet152、inception_v4、senet154等)，输入上述的数据进行训练、验证及测试，比较各模型测试集的预测准确率，获得效果最优的分类模型。
本发明中，训练得到的tile模型的具体参数为：学习率0.001，epoch 150轮，Batch_size为32，输入原始图像得到score，二元交叉熵作为损失函数，反向传播进行参数更新。
所述的aggregation模型的训练验证及测试方法如下：
切片tile模型的特征被提取并形成整张切片热图，根据每个区域是否出现计数进行定量，以病理医生标注对本发明的定量结果进行标准定义。切片aggregation模型经验证集数据验证和测试后，最终形成DM-pathology甲状腺癌定量组织学分类评估系统。
本发明中验证后的aggregation模型的具体参数为：学习率0.001，momentum超参0.9，epoch 2000轮，Batch_size设置为16，损失函数计分类误差、检测误差、分割误差优化。
另一方面，本发明提供了一种甲状腺疾病的病理分析系统。
所述的病理分析系统包括前述的病理分析模块。
所述的病理分析系统还包括数据采集模块。
所述的数据采集模块用于采集符合预设要求的组织切片数据。
所述的数据采集模块应用时包括组织切片的制备。
本发明的有益效果：
本发明适用于PTC的分类及4个高侵亚型的识别，高侵亚型分类与定量达到临床辅助应用级别。
本发明的模型具有以下特点：首先在tile模型中使用MIL对分类器进行训练，然后将其结果通过学习融合模型，对WSI中每个小块的预测分数进行聚合。在此基础上开发了一个新的框架，利用MIL来训练神经网络从而得到语义丰富的特征表示，再将这些特征表示用于RNN，以整合整个WSI的信息，并报告出最终的分类结果并定量化。
模型优势：定量化；准确度高；更为适合真实世界的数据(强监督学习+弱监督模型相结合)。
附图说明
图1为本发明的技术路线图。
图2为小标签标记的WSI热图。
具体实施方式
下面结合具体实施例，对本发明作进一步详细的阐述，下述实施例不用于限制本发明，仅用于说明本发明。以下实施例中所使用的实验方法如无特殊说明，实施例中未注明具体条件的实验方法，通常按照常规条件，下述实施例中所使用的材料、试剂等，如无特殊说明，均可从商业途径得到。
实施例1一种甲状腺病理分析模块
训练样本来源：某三甲医院病理科切片样本。
本实施例中，样本包括：
大样本(WSI)950张(格式为Kfb，GB)；
大样本中切割出小样本量6万(病理原图Patch，格式为Jpg，KB)。
通过以下步骤完成分析：
(1)数据收集与清洗：Kfb格式的WSI，经过严格的入排，切片标准：应满足切片完整、厚薄均匀、表贴得当、染色鲜艳、对比分明、切片整洁、标签清楚端正；
(2)数据扫描与预处理：对步骤(1)清洗后的组织切片进行数据扫描得到WSI图像，组织切片数据中的WSI图像画格；
(3)数据在线标注与坐标输出：对WSI图像中每个单元格(512*512像素)进行二分类标注，图像标记倍数为20×或40×，具体二分类标注的方法分两步：病理医生对WSI取图，将图生成坐标，小标签标记；算法在线对WSI取图，病理医生对小标签标记，导出矢量坐标用于建模。小标签标记的WSI示例如图2。
本实施例中病理原图Patch用于tile模型的训练，具体数量为：训练集42000，验证集12000，测试集6000，总样本量60000；WSI用于aggregation模型的训练，具体数量为：训练集580，验证集166，测试集84(内部)+120(外部)，总样本量950。
(4)tile模型(二类四型：分别对应三类五型标签)：
染色均衡：通过稀疏自编码器寻找各类组织的潜在空间表征，然后将其作为特征用于聚类，以区分不同的组织从而进行染色均衡。
染色分解：利用H&E染色通道提取纹理特征。通过利用颜色反卷积矩阵，将RGB图像转换为H通道图像和E通道图像实现染色分解。
以已有基本网络模型为基础，结合病理图像识别的特点，充分考虑层数、深度及超参数调优，搭建多种卷积神经网络模型(包括se_resnet101、se_resnet152、inception_v4、senet154 等)，输入数据进行训练、输入验证集数据进行验证并输入测试集数据进行测试，比较各模型测试集的预测准确率，获得效果最优的分类模型。具体模型的主体参数设置为：学习率0.001，epoch 150轮，Batch_size为32，输入原始图像得到score，采用softmax分类器计算每类样本的概率，最后通过交叉熵损失(cross-entropy loss)函数优化，二元交叉熵作为损失函数
Figure PCTCN2022120114-appb-000001
反向传播进行参数更新；
(5)aggregation模型：切片tile模型的特征被提取并形成整张切片热图，根据每个区域是否出现计数进行定量，以病理医生标注对本发明的定量结果进行标准定义。
使用上述骨架分类网络形成tile模型对tile进行特征提取，在aggregation模型引入特征拼接方法实现整张WSI原图识别预测实现分类与定量分级，对应不同的骨架模型分别表示为Feature1，Feature2，Feature3，Feature4，其中Feature1＝(Feature11，Feature12，Feature13，…Feature1n)；以此类推，Featureall＝{Feature1，Feature2，Feature3，Feature4}。见图1。
切片aggregation模型经验证集数据验证和测试后，最终形成DM-pathology甲状腺癌定量组织学分类评估系统；最终aggregation模型的主体参数为：学习率0.001，momentum超参0.9，epoch 2000轮，Batch_size设置为16，损失函数计分类误差、检测误差、分割误差优化。对于多分类任务，采用softmax分类器计算每类样本的概率，最后通过交叉熵损失(cross-entropy loss)函数优化。
Figure PCTCN2022120114-appb-000002
报告输出。
本实施例的tile模型(AI定量)对于PTC分型结果准确率可达94.1％，灵敏度94.7％，特异度93.8％；aggregation模型结果(AI分类)对于PTC分型结果准确率可达98.2％，灵敏度99.2％，特异度88.2％。
以上结果中准确率、灵敏度和特异度基于测试集数据进行统计。

Claims (11)

  1. 一种甲状腺疾病病理分析模块，其特征在于，包括硬件部分和数据分析程序部分；所述的硬件部分为一台计算机，所述的数据分析部分包括分类预处理程序和模型程序，所述的模型程序包括tile模型和aggregation模型。
  2. 根据权利要求1所述的甲状腺疾病病理分析模块，其特征在于，所述的tile模型用于小样本分析，aggregation模型用于大样本分析。
  3. 根据权利要求1所述的甲状腺疾病病理分析模块，其特征在于，所述的tile模型的具体参数为：学习率0.001，epoch 150轮，Batch_size为32，输入原始图像得到score，二元交叉熵作为损失函数，反向传播进行参数更新。
  4. 根据权利要求1所述的甲状腺疾病病理分析模块，其特征在于，所述的aggregation模型的具体参数为：学习率0.001，momentum超参0.9，epoch 2000轮，Batch_size设置为16，损失函数计分类误差、检测误差、分割误差优化。
  5. 根据权利要求1所述的甲状腺疾病病理分析模块，其特征在于，所述的甲状腺疾病为甲状腺癌。
  6. 根据权利要求5所述的甲状腺疾病病理分析模块，其特征在于，所述的甲状腺癌为分化型甲状腺癌。
  7. 根据权利要求6所述的甲状腺疾病病理分析模块，其特征在于，所述的分化型甲状腺癌为甲状腺乳头状癌。
  8. 根据权利要求1所述的甲状腺疾病病理分析模块，其特征在于，所述的分类预处理程序用于组织切片数据中的WSI图像画格和标注。
  9. 根据权利要求8所述的甲状腺疾病病理分析模块，其特征在于，所述的标注方法为二分类标注。
  10. 根据权利要求9所述的甲状腺疾病病理分析模块，其特征在于，所述的模块的分析过程包括：数据收集与清洗、数据扫描与预处理、数据在线标注与坐标输出、tile模型分析、aggregation模型分析和报告输出。
  11. 一种甲状腺疾病病理分析系统，其特征在于，包括权利要求1-10任一项所述的病理分析模块，所述的分析系统中还包括数据采集模块，所述的数据采集模块用于采集符合预设要求的组织切片数据。
PCT/CN2022/120114 2022-06-17 2022-09-21 一种甲状腺疾病病理分析模块 WO2023240819A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210690505.X 2022-06-17
CN202210690505.XA CN115132375A (zh) 2022-06-17 2022-06-17 一种甲状腺疾病病理分析模块

Publications (1)

Publication Number Publication Date
WO2023240819A1 true WO2023240819A1 (zh) 2023-12-21

Family

ID=83377664

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/120114 WO2023240819A1 (zh) 2022-06-17 2022-09-21 一种甲状腺疾病病理分析模块

Country Status (2)

Country Link
CN (1) CN115132375A (zh)
WO (1) WO2023240819A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108364006A (zh) * 2018-01-17 2018-08-03 超凡影像科技股份有限公司 基于多模式深度学习的医学图像分类装置及其构建方法
US20180374210A1 (en) * 2015-11-17 2018-12-27 The Board Trustees Of The Leland Stanford Junior University Profiling of Pathology Images for Clinical Applications
CN110009629A (zh) * 2019-04-12 2019-07-12 北京天明创新数据科技有限公司 一种尘肺病筛查系统及其数据训练方法
US20200388029A1 (en) * 2017-11-30 2020-12-10 The Research Foundation For The State University Of New York System and Method to Quantify Tumor-Infiltrating Lymphocytes (TILs) for Clinical Pathology Analysis Based on Prediction, Spatial Analysis, Molecular Correlation, and Reconstruction of TIL Information Identified in Digitized Tissue Images
CN112884724A (zh) * 2021-02-02 2021-06-01 广州智睿医疗科技有限公司 一种用于肺癌组织病理分型的智能判断方法及系统
CN113947607A (zh) * 2021-09-29 2022-01-18 电子科技大学 一种基于深度学习的癌症病理图像生存预后模型构建方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114612664A (zh) * 2022-03-14 2022-06-10 哈尔滨理工大学 一种基于双边分割网络的细胞核分割方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180374210A1 (en) * 2015-11-17 2018-12-27 The Board Trustees Of The Leland Stanford Junior University Profiling of Pathology Images for Clinical Applications
US20200388029A1 (en) * 2017-11-30 2020-12-10 The Research Foundation For The State University Of New York System and Method to Quantify Tumor-Infiltrating Lymphocytes (TILs) for Clinical Pathology Analysis Based on Prediction, Spatial Analysis, Molecular Correlation, and Reconstruction of TIL Information Identified in Digitized Tissue Images
CN108364006A (zh) * 2018-01-17 2018-08-03 超凡影像科技股份有限公司 基于多模式深度学习的医学图像分类装置及其构建方法
CN110009629A (zh) * 2019-04-12 2019-07-12 北京天明创新数据科技有限公司 一种尘肺病筛查系统及其数据训练方法
CN112884724A (zh) * 2021-02-02 2021-06-01 广州智睿医疗科技有限公司 一种用于肺癌组织病理分型的智能判断方法及系统
CN113947607A (zh) * 2021-09-29 2022-01-18 电子科技大学 一种基于深度学习的癌症病理图像生存预后模型构建方法

Also Published As

Publication number Publication date
CN115132375A (zh) 2022-09-30

Similar Documents

Publication Publication Date Title
Landau et al. Artificial intelligence in cytopathology: a review of the literature and overview of commercial landscape
Xue et al. An application of transfer learning and ensemble learning techniques for cervical histopathology image classification
Gertych et al. Machine learning approaches to analyze histological images of tissues from radical prostatectomies
KR102583103B1 (ko) 계산 검출 방법들을 위해 전자 이미지들을 처리하기 위한 시스템들 및 방법들
Mi et al. Deep learning-based multi-class classification of breast digital pathology images
US20190042826A1 (en) Automatic nuclei segmentation in histopathology images
CN107280697A (zh) 基于深度学习和数据融合的肺结节分级判定方法和系统
Pan et al. Mitosis detection techniques in H&E stained breast cancer pathological images: A comprehensive review
CN112101451A (zh) 一种基于生成对抗网络筛选图像块的乳腺癌组织病理类型分类方法
Nofallah et al. Machine learning techniques for mitoses classification
US20210216745A1 (en) Cell Detection Studio: a system for the development of Deep Learning Neural Networks Algorithms for cell detection and quantification from Whole Slide Images
Habtemariam et al. Cervix type and cervical cancer classification system using deep learning techniques
He et al. A review: The detection of cancer cells in histopathology based on machine vision
Zhang et al. Automatic detection of invasive ductal carcinoma based on the fusion of multi-scale residual convolutional neural network and SVM
CN112990214A (zh) 一种医学图像特征识别预测模型
CN113657449A (zh) 一种含噪标注数据的中医舌象腐腻分类方法
CN113012129A (zh) 一种脑切片图像的区域定位及标记神经细胞计数系统及装置
Xu et al. Histopathological tissue segmentation of lung cancer with bilinear cnn and soft attention
Chen et al. VGG16-based intelligent image analysis in the pathological diagnosis of IgA nephropathy
Jing et al. A comprehensive survey of intestine histopathological image analysis using machine vision approaches
Zhang et al. Deep learning-based methods for classification of microsatellite instability in endometrial cancer from HE-stained pathological images
Saito et al. Dawn of the digital diagnosis assisting system, can it open a new age for pathology?
Zhang Classification and diagnosis of thyroid carcinoma using reinforcement residual network with visual attention mechanisms in ultrasound images
Saxena et al. Study of Computerized Segmentation & Classification Techniques: An Application to Histopathological Imagery
WO2023240819A1 (zh) 一种甲状腺疾病病理分析模块

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22946492

Country of ref document: EP

Kind code of ref document: A1