Prediction of prognosis and treatment response in ovarian cancer patients from histopathology images using graph deep learning: a multicenter retrospective study

Biopsy image-based deep learning for predicting pathologic response to neoadjuvant chemotherapy in patients with NSCLC

Neoadjuvant chemotherapy (NAC) is a widely used therapeutic strategy for patients with resectable non-small cell lung cancer (NSCLC). However, individual responses to NAC vary widely among patients, limiting its effective clinical application. In this study, we propose a weakly supervised deep learning model, DeepDrRVT, which integrates self-supervised feature extraction and attention-based deep multiple instance learning, to improve NAC decision making from pretreatment biopsy images. DeepDrRVT demonstrated superior predictive performance and generalizability, achieving AUCs of 0.954, 0.872 and 0.848 for complete pathologic response, and 0.968, 0.893 and 0.831 for major pathologic response in the training, internal validation and external validation cohorts, respectively. The DeepDrRVT digital assessment of residual viable tumor correlated significantly with the local pathologists' visual assessment (Pearson r = 0.98, 0.80, and 0.59; digital/visual slope = 1.0, 0.8 and 0.55) and was also associated with longer disease-free survival (DFS) in all cohorts (HR = 0.455, 95% CI 0.234-0.887, P = 0.018; HR = 0.347, 95% CI 0.135-0.892, P = 0.021 and HR = 0.446, 95% CI 0.193-1.027, P = 0.051). Furthermore, DeepDrRVT remained an independent prognostic factor for DFS after adjustment for clinicopathologic variables (HR = 0.456, 95% CI 0.227-0.914, P = 0.027; HR = 0.358, 95% CI 0.135-0.949, P = 0.039 and HR = 0.419, 95% CI 0.181-0.974, P = 0.043). Thus, DeepDrRVT holds promise as an accessible and reliable tool for clinicians to make more informed treatment decisions prior to the initiation of NAC.

Foundation model-driven multimodal prognostic prediction in patients undergoing primary surgery for high-grade serous ovarian cancer

High-grade serous ovarian cancer (HGSOC) presents challenges in prognostic prediction. This study aimed to develop a universal foundation model-driven multimodal model (FoMu model) to assess the prognosis of HGSOC patients. We conducted a retrospective cohort study involving 712 eligible patients across four centers, collecting clinical, MRI, and hematoxylin and eosin (H&E)-stained whole slide images (WSIs) data. Pre-trained radiological and pathological foundation models were employed for feature precoding. Subsequently, we introduced unimodal and cross-modal adaptive aggregation networks to comprehensively model the features derived from each modality. Our findings revealed that both unimodal and cross-modal FoMu models exhibited superior and stable predictive capabilities for overall survival (OS) and progression-free survival (PFS). In summary, our study successfully developed a FoMu model that effectively integrates multimodal data to assess the prognoses of HGSOC patients, highlighting its potential for improving individualized patient management and clinical decision-making in future applications.

An Overview of Artificial Intelligence in Gynaecological Pathology Diagnostics

Background: The advent of artificial intelligence (AI) has revolutionised many fields in healthcare. More recently, it has garnered interest in terms of its potential applications in histopathology, where algorithms are increasingly being explored as adjunct technologies that can support pathologists in diagnosis, molecular typing and prognostication. While many research endeavours have focused on solid tumours, gynaecological malignancies have nevertheless been relatively overlooked. The aim of this review was therefore to provide a summary of the status quo in the field of AI in gynaecological pathology by encompassing malignancies throughout the entirety of the female reproductive tract rather than focusing on individual cancers. Methods: This narrative/scoping review explores the potential application of AI in whole slide image analysis in gynaecological histopathology, drawing on both findings from the research setting (where such technologies largely remain confined), and highlights any findings and/or applications identified and developed in other cancers that could be translated to this arena. Results: A particular focus is given to ovarian, endometrial, cervical and vulval/vaginal tumours. This review discusses different algorithms, their performance and potential applications. Conclusions: The effective application of AI tools is only possible through multidisciplinary co-operation and training.

Prediction of Clavien Dindo Classification ≥ Grade III Complications After Epithelial Ovarian Cancer Surgery Using Machine Learning Methods

Background and Objectives: Ovarian cancer surgery requires multiple radical resections with a high risk of complications. The aim of this single-centre, retrospective study was to determine the best method for predicting Clavien-Dindo grade ≥ III complications using machine learning techniques. Material and Methods: The study included 179 patients who underwent surgery at the gynaecological oncology department of Antalya Training and Research Hospital between January 2015 and December 2020. The data were randomly split into training set n = 134 (75%) and test set n = 45 (25%). We used 49 predictors to develop the best algorithm. Mean absolute error, root mean squared error, correlation coefficients, Mathew's correlation coefficient, and F1 score were used to determine the best performing algorithm. Cohens' kappa value was evaluated to analyse the consistency of the model with real data. The relationship between these predicted values and the actual values were then summarised using a confusion matrix. True positive (TP) rate, False positive (FP) rate, precision, recall, and Area under the curve (AUC) values were evaluated to demonstrate clinical usability and classification skills. Results: 139 patients (77.65%) had no morbidity or grade I-II CDC morbidity, while 40 patients (22.35%) had grade III or higher CDC morbidity. BayesNet was found to be the most effective prediction model. No dominant parameter was observed in the Bayesian net importance matrix plot. The true positive (TP) rate was 76%, false positive (FP) rate was 15.6%, recall rate (sensitivity) was 76.9%, and overall accuracy was 82.2% A receiver operating characteristic (ROC) analysis was performed to estimate CDC grade ≥ III. AUC was 0.863 with a statistical significance of p < 0.001, indicating a high degree of accuracy. Conclusions: The Bayesian network model achieved the highest accuracy compared to all other models in predicting CDC Grade ≥ III complications following epithelial ovarian cancer surgery.

Integrative deep learning and radiomics analysis for ovarian tumor classification and diagnosis: a multicenter large-sample comparative study

PURPOSE

This study aims to evaluate the effectiveness of combining transvaginal ultrasound (US)-based radiomics and deep learning model for the accurate differentiation between benign and malignant ovarian tumors in large-scale studies.

MATERIALS AND METHODS

A multicenter retrospective study collected grayscale and color US images of ovarian tumors. Patients were divided into training, internal, and external validation groups. Models including a convolutional neural networks (CNN), optimal radiomics, and a combined model were constructed and evaluated for predictive performance using area under curve (AUC), sensitivity, and specificity. The DeLong test compared model AUCs with O-RADS and expert assessments.

RESULTS

3193 images from 2078 patients were analyzed. The CNN achieved AUCs of 0.970 (internal) and 0.959 (external), respectively. Optimal radiomic model achieved AUCs of 0.949 (internal) and 0.954 (external), respectively. The combined CNN-radiomics model attained the highest AUC of 0.977 (internal) and 0.972 (external), respectively, outperforming individual models, O-RADS, and expert methods (p < 0.05).

CONCLUSIONS

The combined CNN-radiomics model using transvaginal US images provides more accurate and reliable ovarian tumor diagnosis, enhancing malignancy prediction and offering clinicians a more precise diagnostic tool.

Graph neural networks in histopathology: Emerging trends and future directions

Histopathological analysis of whole slide images (WSIs) has seen a surge in the utilization of deep learning methods, particularly Convolutional Neural Networks (CNNs). However, CNNs often fail to capture the intricate spatial dependencies inherent in WSIs. Graph Neural Networks (GNNs) present a promising alternative, adept at directly modeling pairwise interactions and effectively discerning the topological tissue and cellular structures within WSIs. Recognizing the pressing need for deep learning techniques that harness the topological structure of WSIs, the application of GNNs in histopathology has experienced rapid growth. In this comprehensive review, we survey GNNs in histopathology, discuss their applications, and explore emerging trends that pave the way for future advancements in the field. We begin by elucidating the fundamentals of GNNs and their potential applications in histopathology. Leveraging quantitative literature analysis, we explore four emerging trends: Hierarchical GNNs, Adaptive Graph Structure Learning, Multimodal GNNs, and Higher-order GNNs. Through an in-depth exploration of these trends, we offer insights into the evolving landscape of GNNs in histopathological analysis. Based on our findings, we propose future directions to propel the field forward. Our analysis serves to guide researchers and practitioners towards innovative approaches and methodologies, fostering advancements in histopathological analysis through the lens of graph neural networks.

Explainable AI-based feature importance analysis for ovarian cancer classification with ensemble methods

Introduction

Ovarian Cancer (OC) is one of the leading causes of cancer deaths among women. Despite recent advances in the medical field, such as surgery, chemotherapy, and radiotherapy interventions, there are only marginal improvements in the diagnosis of OC using clinical parameters, as the symptoms are very non-specific at the early stage. Owing to advances in computational algorithms, such as ensemble machine learning, it is now possible to identify complex patterns in clinical parameters. However, these complex patterns do not provide deeper insights into prediction and diagnosis. Explainable artificial intelligence (XAI) models, such as LIME and SHAP Kernels, can provide insights into the decision-making process of ensemble models, thus increasing their applicability.

Methods

The main aim of this study is to design a computer-aided diagnostic system that accurately classifies and detects ovarian cancer. To achieve this objective, a three-stage ensemble model and a game-theoretic approach based on SHAP values were built to evaluate and visualize the results, thus analyzing the important features responsible for prediction.

Results and Discussion

The results demonstrate the efficacy of the proposed model with an accuracy of 98.66%. The proposed model's consistency and advantages are compared with single classifiers. The SHAP values of the proposed model are validated using conventional statistical methods such as the p-test and Cohen's d-test to highlight the efficacy of the proposed method. To further validate the ranking of the features, we compared the p-values and Cohen's d-values of the top five and bottom five features. The study proposed and validated an AI-based method for the detection, diagnosis, and prognosis of OC using multi-modal real-life data, which mimics the move of a clinician approach with a demonstration of high performance. The proposed strategy can lead to reliable, accurate, and consistent AI solutions for the detection and management of OC with higher patient experience and outcomes at low cost, low morbidity, and low mortality. This can be beneficial for millions of women living in resource-constrained and challenging economies.

Denoised recurrence label-based deep learning for prediction of postoperative recurrence risk and sorafenib response in HCC

BACKGROUND

Pathological images of hepatocellular carcinoma (HCC) contain abundant tumor information that can be used to stratify patients. However, the links between histology images and the treatment response have not been fully unveiled.

METHODS

We trained and evaluated a model by predicting the prognosis of 287 non-treated HCC patients postoperatively, and further explored the model's treatment response predictive ability in 79 sorafenib-treated patients. Based on prognostic relevant pathological signatures (PPS) extracted from CNN-SASM, which was trained by denoised recurrence label (DRL) under different thresholds, the PPS-based prognostic model was formulated. A total of 78 HCC patients from TCGA-LIHC were used for the external validation.

RESULTS

We proposed the CNN-SASM based on tumor pathology and extracted PPS. Survival analysis revealed that the PPS-based prognostic model yielded the AUROC of 0.818 and 0.811 for predicting recurrence at 1 and 2 years after surgery, with an external validation reaching 0.713 and 0.707. Furthermore, the predictive ability of the PPS-based prognostic model was superior to clinical risk indicators, and it could stratify patients with significantly different prognoses. Importantly, our model can also stratify sorafenib-treated patients into two groups associated with significantly different survival situations, which could effectively predict survival benefits from sorafenib.

CONCLUSIONS

Our prognostic model based on pathology deep learning provided a valuable means for predicting HCC patient recurrence condition, and it could also improve patient stratification to sorafenib treatment, which help clinical decision-making in HCC.

Comparing the Effectiveness of Artificial Intelligence Models in Predicting Ovarian Cancer Survival: A Systematic Review

BACKGROUND

This systematic review investigates the use of machine learning (ML) algorithms in predicting survival outcomes for ovarian cancer (OC) patients. Key prognostic endpoints, including overall survival (OS), recurrence-free survival (RFS), progression-free survival (PFS), and treatment response prediction (TRP), are examined to evaluate the effectiveness of these algorithms and identify significant features that influence predictive accuracy.

RECENT FINDINGS

A thorough search of four major databases-PubMed, Scopus, Web of Science, and Cochrane-resulted in 2400 articles published within the last decade, with 32 studies meeting the inclusion criteria. Notably, most publications emerged after 2021. Commonly used algorithms for survival prediction included random forest, support vector machines, logistic regression, XGBoost, and various deep learning models. Evaluation metrics such as area under the curve (AUC) (18 studies), concordance index (C-index) (11 studies), and accuracy (11 studies) were frequently employed. Age at diagnosis, tumor stage, CA-125 levels, and treatment-related factors were consistently highlighted as significant predictors, emphasizing their relevance in OC prognosis.

CONCLUSION

ML models demonstrate considerable potential for predicting OC survival outcomes; however, challenges persist regarding model accuracy and interpretability. Incorporating diverse data types-such as clinical, imaging, and molecular datasets-holds promise for enhancing predictive capabilities. Future advancements will depend on integrating heterogeneous data sources with multimodal ML approaches, which are crucial for improving prognostic precision in OC.

Benchmarking histopathology foundation models for ovarian cancer bevacizumab treatment response prediction from whole slide images

PURPOSE

Bevacizumab is a widely studied targeted therapeutic drug used in conjunction with standard chemotherapy for the treatment of recurrent ovarian cancer. While its administration has been shown to increase progression-free survival (PFS) in patients with advanced-stage ovarian cancer, the lack of identifiable biomarkers for predicting patient response has been a major roadblock in its effective adoption towards personalized medicine.

METHODS

In this work, we leverage the latest histopathology foundation models trained on large-scale whole slide image (WSI) datasets to extract ovarian tumor tissue features for predicting bevacizumab response from WSIs.

RESULTS

Our extensive experiments across a combination of different histopathology foundation models and multiple instance learning (MIL) strategies demonstrate the capability of these large models in predicting bevacizumab response in ovarian cancer patients with the best models achieving a patient-level balanced accuracy score close to 70%. Furthermore, these models can effectively stratify high- and low-risk patients (p < 0.05) during the first year of bevacizumab treatment.

CONCLUSION

This work highlights the utility of histopathology foundation models to predict response to bevacizumab treatment  from WSIs. The high-attention regions of the WSIs highlighted by these models not only aid the model explainability but also serve as promising imaging biomarkers for treatment prognosis.

Homologous recombination deficiency in ovarian cancer: Global expert consensus on testing and a comparison of companion diagnostics

BACKGROUND

Poly (ADP ribose) polymerase inhibitors (PARPis) are a treatment option for patients with advanced high-grade serous or endometrioid ovarian carcinoma (OC). Recent guidelines have clarified how homologous recombination deficiency (HRD) may influence treatment decision-making in this setting. As a result, numerous companion diagnostic assays (CDx) have been developed to identify HRD. However, the optimal HRD testing strategy is an area of debate. Moreover, recently published clinical and translational data may impact how HRD status may be used to identify patients likely to benefit from PARPi use. We aimed to extensively compare available HRD CDx and establish a worldwide expert consensus on HRD testing in primary and recurrent OC.

METHODS

A group of 99 global experts from 31 different countries was formed. Using a modified Delphi process, the experts aimed to establish consensus statements based on a systematic literature search and CDx information sought from investigators, companies and/or publications.

RESULTS

Technical information, including analytical and clinical validation, were obtained from 14 of 15 available HRD CDx (7 academic; 7 commercial). Consensus was reached on 36 statements encompassing the following topics: 1) the predictive impact of HRD status on PARPi use in primary and recurrent OC; 2) analytical and clinical validation requirements of HRD CDx; 3) resource-stratified HRD testing; and 4) how future CDx may include additional approaches to help address unmet testing needs.

CONCLUSION

This manuscript provides detailed information on currently available HRD CDx and up-to-date guidance from global experts on HRD testing in patients with primary and recurrent OC.

NETosis Genes and Pathomic Signature: A Novel Prognostic Marker for Ovarian Serous Cystadenocarcinoma

To evaluate the prognostic significance and molecular mechanism of NETosis markers in ovarian serous cystadenocarcinoma (OSC), we constructed a machine learning-based pathomic model utilizing hematoxylin and eosin (H&E) slides. We analyzed 333 patients with OSC from The Cancer Genome Atlas for prognostic-related neutrophil extracellular trap formation (NETosis) genes through bioinformatics analysis. Pathomic features were extracted from 54 cases with complete pathological images, genetic matrices, and clinical information. Two pathomic prognostic models were constructed using support vector machine (SVM) and logistic regression (LR) algorithms. Additionally, we established a predictive scoring system that integrated pathomic scores based on the NETcluster subtypes and clinical signature. We identified four NETosis genes significantly correlated with OSC prognosis, which were functionally associated with immune response, somatic mutations, tumor invasion, and metastasis. Five robust pathomic features were selected for overall survival prediction. The LR and SVM pathomic models demonstrated strong predictive performance for the NETcluster subtype classification through five-fold cross-validation. Time-dependent ROC analysis revealed excellent prognostic capability of the LR pathomic model's score for the overall survival (AUC values of 0.658, 0.761, and 0.735 at 36, 48, and 60 months, respectively), further validated by Kaplan-Meier analysis. The expression levels of NETosis genes greatly affected OSC patients' prognoses. The pathomic analysis of H&E slide pathological images provides an effective approach for predicting both NETcluster subtype and overall survival in OSC patients.

Diagnostic Utility of Metalloproteinases from Collagenase Group (MMP-1, MMP-8 and MMP-13) in Biochemical Diagnosis of Ovarian Carcinoma

BACKGROUND

Ovarian carcinoma (OC) has an unfavorable prognosis due to lack of screening and an asymptomatic course. New diagnostic methods are being sought to enable earlier diagnosis of this condition. The purpose of this study was to determine the diagnostic utility of collagenases (MMP-1, MMP-8 and MMP-13) in the diagnosis of OC compared to HE4 and CA125 and the ROMA.

METHODS

The study group consisted of 120 patients with OC, the control group: 70 patients with benign ovarian lesions (BLs) and 50 healthy women (HS). MMP-1, MMP-8 and MMP-13 were determined by ELISA and HE4 and CA125 by CMIA.

RESULTS

OC patients had higher levels of MMP-1 and MMP-13 compared to the BL and HS groups. MMP-1 (SE: 81.66%; SP: 94%; PPV: 97.02%; NPV: 68.11%; AUC: 0.9625) and MMP-13 (SE: 77.50%; SP: 94%; PPV: 96.875%; NPV: 63.51%; AUC: 0.917) showed similar or higher diagnostic values to routine markers (HE4: SE:85%; SP: 92%; PPV: 96.22%; NPV: 71.875%; AUC: 0.943; CA125: SE: 80%; SP: 98%; PPV: 98.96%; NPV: 67.12%; AUC: 0.909) and the ROMA (SE: 90.83%; SP: 94%; PPV: 97.32%; NPV: 81.03%; AUC: 0.955). Performing combined analyses of individual MMPs and MMPs with ROMA was associated with further increases in diagnostic parameters.

CONCLUSIONS

MMP-1 and MMP-13 have shown preliminary potential as diagnostic markers and auxiliary markers to ROMA in biochemical diagnosis of OC.

Enhanced ovarian cancer survival prediction using temporal analysis and graph neural networks

Ovarian cancer is a formidable health challenge that demands accurate and timely survival predictions to guide clinical interventions. Existing methods, while commendable, suffer from limitations in harnessing the temporal evolution of patient data and capturing intricate interdependencies among different data elements. In this paper, we present a novel methodology which combines Temporal Analysis and Graph Neural Networks (GNNs) to significantly enhance ovarian cancer survival rate predictions. The shortcomings of current processes originate from their disability to correctly seize the complex interactions amongst diverse scientific information units in addition to the dynamic modifications that arise in a affected person`s nation over time. By combining temporal information evaluation and GNNs, our cautioned approach overcomes those drawbacks and, whilst as compared to preceding methods, yields a noteworthy 8.3% benefit in precision, 4.9% more accuracy, 5.5% more advantageous recall, and a considerable 2.9% reduction in prediction latency. Our method's Temporal Analysis factor uses longitudinal affected person information to perceive good-sized styles and tendencies that offer precious insights into the direction of ovarian cancer. Through the combination of GNNs, we offer a robust framework able to shoot complicated interactions among exclusive capabilities of scientific data, permitting the version to realize diffused dependencies that would affect survival results. Our paintings have tremendous implications for scientific practice. Prompt and correct estimation of the survival price of ovarian most cancers allows scientific experts to customize remedy regimens, manipulate assets efficiently, and provide individualized care to patients. Additionally, the interpretability of our version`s predictions promotes a collaborative method for affected person care via way of means of strengthening agreement among scientific employees and the AI-driven selection help system. The proposed approach not only outperforms existing methods but also has the possible to develop ovarian cancer treatment by providing clinicians through a reliable tool for informed decision-making. Through a fusion of Temporal Analysis and Graph Neural Networks, we conduit the gap among data-driven insights and clinical practice, proposing a capable opportunity for refining patient outcomes in ovarian cancer management operations.

Integrated immuno-transcriptomic analysis of ovarian cancer identifies a four-chemokine-dominated subtype with antitumor immune-active phenotype and favorable prognosis

BACKGROUND

Ovarian cancer (OV) is a heterogeneous disease but has traditionally been treated as an immunologically cold malignancy. The relationship between the immune-active cancer phenotype typified by a T helper 1 (Th-1) immune response and clinical outcome in OV remains uncertain.

METHODS

A cohort-scale compendium of transcriptomic data from 2850 OV samples from 19 individual datasets was compiled for integrative immuno-transcriptomic analysis. The immunological constant of rejection was used as a metric to assess the Th-1/cytotoxic response orientation and investigate the clinical-biological significance of immune polarization towards a Th-1 immune response. Single-cell RNA sequencing data from 39 OV samples were analyzed to elucidate the variability of the immune microenvironment, and immunohistochemical validation was performed on 39 samples from the Harbin Medical University Cancer Hospital.

RESULTS

Our results demonstrated the prognostic significance of a Th-1/cytotoxic immune profile within the tumor microenvironment (TME) using the immunological constant of rejection classification to OV samples. Specifically, patients with tumors expressing high levels of ICR markers showed significantly improved survival. A gene panel consisting of four chemokines (CXCL9, CXCL10, CXCL11 and CXCL13) was identified as critical players in mediating the establishment of an active T-cell-inflamed antitumor phenotype. This 4-chemokine signature, which was extensively validated in external multicenter cohorts through transcriptomic profiling and in an independent in-house cohort through immunohistochemistry, introduced a novel immune classification in OV and identified a chemokine-dominated subtype associated with an active antitumor immune phenotype and favorable prognosis. Single-cell transcriptomic analysis revealed that chemokine-dominated tumors increase CXCR3 + NK and T cell recruitment to the TME primarily through the overexpression of macrophage-derived CXCL9/10/11.

CONCLUSIONS

This study provides new insights into understanding immune heterogeneity within the TME and paves the way for tailoring appropriate therapeutic interventions for patients with differing immune profiles.

Automated early detection of acute retinal necrosis from ultra-widefield color fundus photography using deep learning

BACKGROUND

Acute retinal necrosis (ARN) is a relatively rare but highly damaging and potentially sight-threatening type of uveitis caused by infection with the human herpesvirus. Without timely diagnosis and appropriate treatment, ARN can lead to severe vision loss. We aimed to develop a deep learning framework to distinguish ARN from other types of intermediate, posterior, and panuveitis using ultra-widefield color fundus photography (UWFCFP).

METHODS

We conducted a two-center retrospective discovery and validation study to develop and validate a deep learning model called DeepDrARN for automatic uveitis detection and differentiation of ARN from other uveitis types using 11,508 UWFCFPs from 1,112 participants. Model performance was evaluated with the area under the receiver operating characteristic curve (AUROC), the area under the precision and recall curves (AUPR), sensitivity and specificity, and compared with seven ophthalmologists.

RESULTS

DeepDrARN for uveitis screening achieved an AUROC of 0.996 (95% CI: 0.994-0.999) in the internal validation cohort and demonstrated good generalizability with an AUROC of 0.973 (95% CI: 0.956-0.990) in the external validation cohort. DeepDrARN also demonstrated excellent predictive ability in distinguishing ARN from other types of uveitis with AUROCs of 0.960 (95% CI: 0.943-0.977) and 0.971 (95% CI: 0.956-0.986) in the internal and external validation cohorts. DeepDrARN was also tested in the differentiation of ARN, non-ARN uveitis (NAU) and normal subjects, with sensitivities of 88.9% and 78.7% and specificities of 93.8% and 89.1% in the internal and external validation cohorts, respectively. The performance of DeepDrARN is comparable to that of ophthalmologists and even exceeds the average accuracy of seven ophthalmologists, showing an improvement of 6.57% in uveitis screening and 11.14% in ARN identification.

CONCLUSIONS

Our study demonstrates the feasibility of deep learning algorithms in enabling early detection, reducing treatment delays, and improving outcomes for ARN patients.

METnet: A novel deep learning model predicting MET dysregulation in non-small-cell lung cancer on computed tomography images

BACKGROUND

Mesenchymal epithelial transformation (MET) is a key molecular target for diagnosis and treatment of non-small cell lung cancer (NSCLC). The corresponding molecularly targeted therapeutics have been approved by Food and Drug Administration (FDA), achieving promising results. However, current detection of MET dysregulation requires biopsy and gene sequencing, which is invasive, time-consuming and difficult to obtain tumor samples.

METHODS

To address the above problems, we developed a noninvasive and convenient deep learning (DL) model based on Computed tomography (CT) imaging data for prediction of MET dysregulation. We introduced the unsupervised algorithm RK-net for automated image processing and utilized the MedSAM large model to achieve automated tissue segmentation. Based on the processed CT images, we developed a DL model (METnet). The model based on the grouped convolutional block. We evaluated the performance of the model over the internal test dataset using the area under the receiver operating characteristic curve (AUROC) and accuracy. We conducted subgroup analysis on the basis of clinical data of the lung cancer patients and compared the performance of the model in different subgroups.

RESULTS

The model demonstrated a good discriminative ability over the internal test dataset. The accuracy of METnet was 0.746 with an AUC value of 0.793 (95% CI 0.714-0.871). The subgroup analysis revealed that the model exhibited similar performance across different subgroups.

CONCLUSIONS

METnet realizes prediction of MET dysregulation in NSCLC, holding promise for guiding precise tumor diagnosis and treatment at the molecular level.