*For medical professionals for reference
This study shows that artificial intelligence is in HER2 diagnosis.It shows high accuracy, but there are differences and research heterogeneity, which provides important references for clinical practice and future research.
To accurately diagnose HER2 status is important for patients with breast cancer patients, but pathologists often face subjective challenges in assessing HER2 status.With the rapid development of artificial intelligence (AI) technology, its potential for improving HER2 interpretation accuracy and repetitiveness has attracted much attention.However, there are currently lack of literature for system evaluation for AI algorithms used for HER2 diagnosis.A study announced at the 113rd American and Canadian Society (USCAP) Annual Conference was designed to fill this gap. Through comprehensively retrieving relevant literature and adopting strict quality assessment tools, existing AI algorithms diagnosed in breast cancer HER2 diagnosisThe performance in the middle analysis is performed.The research results show that AI shows high accuracy in identifying HER2 expression, but there are differences in the performance of different HER2 expressions, and there are heterogeneity between research.This study not only provides an important reference basis for clinical practice, but also pointed out the direction for future AIs further research and application of AI in the field of breast cancer.The main content of the research is now sorted out as follows to readers.
Research background
Accurate HER2 diagnosis is essential for HER2 targeted therapy.However, pathologists often show subjectivity in assessing HER2 status, which may cause inconsistency of diagnostic results.The introduction of AI technology is expected to improve the accuracy and repeatability of HER2 interpretation to reduce human errors.Specifically, AI can analyze a large amount of pathological image data and learn to identify the characteristics of different HER2 expression samples, so as to provide more objective and consistent diagnostic results.Nevertheless, the existing literature lacks a system assessment for the AI algorithm used for HER2 diagnosis, which limits the widespread application of AI technology in clinical practice.Therefore, carrying out a comprehensive AI algorithm assessment research, verifying its diagnostic performance under different samples and conditions, is of great significance for promoting the application of AI technology in HER2 diagnosis.
Research Design
In PUBMED, Embase, COCHRANE, and Web of Science databasesFreedom text conducted detailed literature search.The retrieval result covers the computing pathology and science related articles published from the database to the period from the establishment of the database until September 2023.According to the established and excluding standards, seven studies are finally selected (see Figure 1A).The selected studies were evaluated, the Quadas-2 evaluation tool was used, and the results of the results were performed with the help of the REVMAN 5.4 software.In addition, the threshold effect was detected through the Meta-Disc 1.4 software, and the data was summarized and analyzed using the dual-variable hybrid effect model in the STATA 17 software.
Figure 1. Incidential document screening process diagram and bias risk assessment
Research Results
This study summarized seven studies, a total of 6867 HER2 identification tasks.Among them, the two studies use the HER2-Connect algorithm, and the other two are applied to the CNN algorithm. One uses multiple types of logical regression algorithms. The remaining two uses the HER2 4B5 algorithm.Based on the Quadas-2 evaluation standards, most of the studies have shown lower bias risks (see Figure 1B).In terms of distinguishing HER2 0/1+, the sensitivity and specificity of artificial intelligence are 0.98 [0.92-0.99] and 0.92 [0.80-0.97], and no threshold effect (Spearman correlation coefficient: 0.321, P value = 0.482))(See Figure 2A).For the distinction between HER2 2+, sensitivity and specificity are 0.78 [0.50-0.92] and 0.98 [0.93-0.99], and there is no threshold effect (Spearman related coefficient: 0.357, P value = 0.432) (see Figure 2B)EssenceAs for the distinction of HER2 3+, the sensitivity of artificial intelligence is 0.99 [0.98-1.00], the specificity is 0.99 [0.97-1.00], nor did it observe the threshold effect (Spearman related coefficient: -0.500, P value = 0.253)See Figure 2C).Nevertheless, all analysis shows a certain degree of heterogeneity.
Figure 2. Sensitivity-Specific Forest Map
Research summary and summary and summary and summary and study summary and study summary and summary and summary and study summary and summary and summary and study summary and study summary and summary and summary of researchThinking
This study through the system evaluates the performance of AI in the automatic reading of breast cancer HER2 immunohistochemistry, revealing the huge potential of AI technology in improving HER2 diagnosis accuracyEssenceThe research results show that when distinguishing the different expression levels of HER2 (0/1+, 2+, 3+), the AI algorithm shows high sensitivity and specificity, especially in HER2 3+ recognition, the diagnosis of AI, the diagnosis of AIPerformance is close to perfect.However, research also pointed out that the performance of different AI algorithms at different HER2 expression levels is different, as well as heterogeneity between research.These discoveries provide important reference for clinical practice.First of all, the application of AI technology is expected to reduce the subjectivity of pathologists in HER2 state assessment, thereby improving the consistency and accuracy of diagnosis.Secondly, the high accuracy manifestations of the AI algorithm in HER2 3+ means that it has potential clinical value in breast cancer targeted therapy decisions, especially in the occasion that requires rapid and accurate diagnosis to start treatment.
Despite this, the heterogeneity tips found in the study, the current application of AI algorithm in HER2 diagnosis of breast cancer is still limited.This may be related to factors such as the AI algorithm, the differences in the training data set, the quality of the pathological image, and the inappropriate judgment standards used in different studies.Therefore, future research needs to further optimize the AI algorithm, improve its generalization capabilities under different conditions, and verify the stability and reliability of AI algorithms through larger and multi -center clinical trials.
In addition, research should also pay attentionUnder the premise, establish standardized training and verification data sets.Future research should also explore the application potential of AI in other breast cancer -related biomarkers, and how to use AI technologies for joint analysis of multi -logo to provide more comprehensive diagnosis and prognosis of breast cancer.
In short, this study provides a scientific basis for the application of AI in the diagnosis of breast cancer HER2, and also pointed out the shortcomings of the current research and the direction of future development.With the continuous advancement and deepening of technology, AI is expected to play an increasingly important role in precision diagnosis and treatment of breast cancer.
Wonderful information is waiting for you
References:
[1] wu s, li x, miao jx, et al. Performance Evaluation of Artificial Intelligence in Automated Assesses AST CANCER. 2024 USCAP. 249.
Approval Number: CN-146622 Validity to: 2025-01-31
This material is provided by Astrikon for reference only for medical and health professionals
