Skip to main content
Download PDF
Download ePub

Personalized prediction of immunotherapy response in lung cancer patients using advanced radiomics and deep learning

Cancer Imaging volume 24, Article number: 129 (2024) Cite this article

Abstract

Background

Lung cancer (LC) is a leading cause of cancer-related mortality, and immunotherapy (IO) has shown promise in treating advanced-stage LC. However, identifying patients likely to benefit from IO and monitoring treatment response remains challenging. This study aims to develop a predictive model for progression-free survival (PFS) in LC patients with IO based on clinical features and advanced imaging biomarkers.

Materials and methods

A retrospective analysis was conducted on a cohort of 206 LC patients receiving IO treatment. Pre-treatment computed tomography images were used to extract advanced imaging biomarkers, including intratumoral and peritumoral-vasculature radiomics. Clinical features, including age, gene status, hematology, and staging, were also collected. Key radiomic and clinical features for predicting IO outcomes were identified using a two-step feature selection process, including univariate Cox regression and chi-squared test, followed by sequential forward selection. The DeepSurv model was constructed to predict PFS based on clinical and radiomic features. Model performance was evaluated using the area under the time-dependent receiver operating characteristic curve (AUC) and concordance index (C-index).

Results

Combining radiomics of intratumoral heterogeneity and peritumoral-vasculature with clinical features demonstrated a significant enhancement (p < 0.001) in predicting IO response. The proposed DeepSurv model exhibited a prediction performance with AUCs ranging from 0.76 to 0.80 and a C-index of 0.83. Furthermore, the predicted personalized PFS curves revealed a significant difference (p < 0.05) between patients with favorable and unfavorable prognoses.

Conclusions

Integrating intratumoral and peritumoral-vasculature radiomics with clinical features enabled the development of a predictive model for PFS in LC patients with IO. The proposed model’s capability to estimate individualized PFS probability and differentiate the prognosis status held promise to facilitate personalized medicine and improve patient outcomes in LC.

Key points

• Immunotherapy’s unique mechanisms in lung cancer have challenged traditional tumor volume-based response assessment, requiring new prognostic factors and algorithms.

• Advanced radiomics analysis, including intratumoral imaging phenotypes and peritumoral vasculature, comprehensively quantifies the tumor characteristics and angiogenesis in lung cancer.

• Integrating clinical features, advance radiomics, and deep learning models provides a more precise prediction of the response to immunotherapy in lung cancer.

Introduction

Lung cancer (LC) is a prominent contributor to global cancer-related mortality. Immunotherapy (IO) has emerged as a promising therapeutic approach for LC, particularly in advanced-stage patients [1, 2]. Unlike traditional chemotherapy that directly targets and destroys cancer cells, IO harnesses the inherent capabilities of the patient’s immune system to suppress the growth of cancer cells [3]. The most widely studied immunotherapy agents for LC are immune checkpoint inhibitors (ICIs), which block inhibitory signals on T-cells [4]. ICIs have improved overall survival (OS) and quality of life in patients with LC, but not all patients benefit from these therapies [5]. Furthermore, ICIs can cause serious immune-related adverse events such as pneumonitis. Therefore, having predictive biomarkers to identify patients most likely to benefit from IO and to monitor treatment response is critical [6].

Currently, the most used biomarkers for predicting response to ICIs are programmed death ligand 1 (PD-L1) expression on tumor cells. Several clinical trials focusing on PD-L1 expression in LC patients have demonstrated that those with low PD-L1 expression may still benefit from ICI drugs. These individuals commonly receive a combination of chemotherapy and ICI therapy. Conversely, patients with high PD-L1 expression frequently exhibit response rates below 50% when treated solely with ICI monotherapy [7]. This highlights the urgent need for IO predictors in LC patients in addition to the PD-L1 expression [8]. Accurate prediction can facilitate LC management and avoid unnecessary costs and potential adverse effects of IO treatment [9, 10].

Medical imaging combined with deep-learning approaches is increasingly being used to predict the efficacy of lung cancer treatment [11]. AI methods can extract meaningful information from complex data (e.g., imaging data from radiology or histopathology images) to predict response to immunotherapy, either directly or indirectly through surrogate markers. These AI-based techniques have the potential to improve diagnostic accuracy, optimize treatment planning, predict treatment efficacy, and reduce human resource costs in the management of cancer immunotherapy [12, 13]. We expanded upon concepts from prior studies [12, 13] that leveraged diagnostic contrast CT images in developing immunotherapy predictors. This study broke new ground by integrating comprehensive clinical data, encompassing hematology, genetics, and clinical presentation, and fusing advanced imaging features with deep learning models.

Radiomics, a high-throughput method to quantify image traits based on computed tomography (CT), has shown promising results in diagnosis and prognosis for patients with LC. Intratumoral CT radiomics offers valuable insights into tumor-specific characteristics and can potentially enhance the prediction of treatment outcomes in LC patients [14]. Furthermore, tumor cells induce an immune-evasive state by releasing vascular endothelial growth factor, promoting angiogenesis, and supplying nutrients to the tumor [15, 16]. Numerous studies have demonstrated that well-structured blood vessels hinder metastasis formation and improve the delivery of chemotherapy agents to the tumor site [17, 18]. Peritumoral-vasculature radiomics can quantify subtle vascular changes and provide a comprehensive assessment of the tumor microenvironment [19]. Overall, the multi-modality vasculature characteristics offer a promising solution to enhance the management and treatment outcomes of LC patients undergoing IO.

The present study aims to develop a predictive deep-learning model based on advanced radiomics (including intratumoral and peritumor-vasculature features) and clinical features for progression-free survival (PFS) in LC patients with IO. The proposed model could provide clinicians with a non-invasive tool to identify patients most likely to benefit from IO and to monitor treatment response. The outcome of this study might help to reduce the risk of immune-related adverse events by avoiding unnecessary treatment in patients unlikely to benefit from ICIs.

Materials and methods

Patient cohort and image data

The Institutional Review Board of Taipei Veterans General Hospital (TVGH) approved this retrospective study (IRB No.: 2021-09-009BCF) and waived the requirement of informed consent from patients. This study collected data from 212 LC patients who received IO treatment between 2016 and 2020 with monthly follow-up assessments. The inclusion criteria are listed as follows: (1) histological diagnosis of LC based on surgical specimens or tissue biopsies; (2) staging assessment of LC according to the American Joint Committee on Cancer (AJCC) guidelines [20] based on imaging and pathological information; (3) no other neoplastic diseases; (4) complete clinical information, such as smoking status, Eastern Cooperative Oncology Group performance status (ECOG PS) score, applied IO drugs, EGFR mutation status; and (5) adequate quality of contrast-enhanced chest CT data. Finally, 206 LC patients were included in the subsequent analyses (Fig. 1). Disease progression was identified as tumor enlargement, new metastasis, and patient death based on monthly follow-up assessments (including vital signs, blood tests, and CT scans if needed), aligning with the primary objective of our model to forecast responses to IO. Discontinuities in the follow-up procedure were documented as instances of data censoring.

Fig. 1
figure 1

Patient flowchart for this study

Full size image

Among 206 included patients, pre-treatment CT image data were acquired using several different scanners produced by five major manufacturers, including GE, Hitachi, Siemens, Toshiba, and Philips (detailed information in Supplementary Table 1). We did not limit the use of specific manufacturers or models of CT scanners to ensure that the predictive model we developed could tolerate the variability introduced by different machines. The CT scan coverage was similar, ranging from the thoracic inlet to the upper abdomen. CT images were reconstructed with a slice thickness between 1.0 and 5.0 mm. Pixel spacing ranged from 0.5 to 1.2 mm. Each slice had a matrix size of 512 × 512 pixels with a 16-bit gray-scale resolution. The peak tube voltage was 120 kVp, and the tube current was decided by automatic dose modulation.

In order to delineate the location of the tumors for the radiomics analysis, experienced radiologists and certified pulmonologists collaborated to assess the quality and delineate the region of interest (ROI) in CT scans. For ROI delineation, two specific window settings were applied to the CT images. The soft tissue window (width: 350, level: 50) aided in distinguishing tumors, collapsed lungs, and fluid components like pleural and pericardial effusions. On the other hand, the lung window (width: 1500, level: -600) was used to determine tumor borders accurately. In the cases with multiple tumors, only the primary tumor was selected as the ROI.

Intratumoral radiomics

We applied several preprocessing steps to the CT images, including resolution adjustment to an isotropic pixel size of 1 × 1 × 1 mm³, followed by intensity standardization using Z-score transformation to achieve standardized ranges based on the mean and standard deviation of the image data. Additionally, low-pass (L) and high-pass (H) wavelet filters were applied to the three axes of CT images, resulting in eight decomposed image sets (named LLL, LLH, LHL, LHH, HLL, HLH, HHL, and HHH images). Intratumoral radiomics analysis involved the calculation of histogram, geometry, and texture features, including gray level co-occurrence matrix (GLCM), gray level run length matrix (GLRLM), and local binary pattern (LBP), from eight wavelet-filtered images and the original CT images. The feature extraction process followed the Imaging Biomarker Standardization Initiative (IBSI) guidelines [21]. In total, 593 intratumoral radiomic features were generated for each ROI (Fig. 2a). The image preprocessing, ROI delineation, and intratumoral radiomics analysis were performed using the Multimodal Radiomics Platform, a MATLAB-based graphic user interface (available online: http://cflu.lab.nycu.edu.tw/MRP_MLinglioma.html, accessed on 9 May 2024) [22, 23]. Supplementary Table S2 provides the formulas used for intratumoral radiomics analysis.

Fig. 2
figure 2

Feature processing workflow for (a) intratumoral radiomics, (b) peritumoral-vasculature radiomics, and (c) clinical features

Full size image

Peritumoral-vasculature radiomics

After the adjustment of image resolution to an isotropic voxel of 1 × 1 × 1 mm³, a pre-trained U-Net model based on 231 routine CT cases (acquired from different scanner manufacturers) was employed to automatically segment the bilateral lung areas on CT images [24]. Afterward, radiomic features of tumor-associated vasculature were extracted through a stepwise process. First, a multi-scale 3D vessel enhancement filter was utilized to enhance vessel-like structures [25]. Afterward, Otsu’s thresholding method was applied to the vessel enhancement image to extract voxels associated with the vasculature. Morphological operations were subsequently applied to eliminate noise and non-vessel artifacts. Finally, a cubic region, covering a 2.5-cm extension from the tumor volume along three directions, was extracted. A fast-marching algorithm was employed on the segmented vasculature to identify the center lines of vessels and partition the vessel network into distinct branches [26]. Ninety-one vasculature radiomic features were computed, comprising 61 quantitative vessel morphology features (e.g., curvature and torsion) and 30 features describing the organization of the peritumoral vasculature (Fig. 2b). The calculation of vasculature radiomic features is provided in Supplementary Table S3. The proportion of outliers (values located outside the three standard deviations from the mean value) was computed for each normalized radiomic feature, and features with outlier proportions exceeding 20% of the entire dataset were then removed from consideration. The peritumoral-vasculature radiomics analysis was also performed using the Multimodal Radiomics Platform (available online: http://cflu.lab.nycu.edu.tw/MRP_MLinglioma.html, accessed on 9 May 2024) with QuanTAV toolbox (available online: https://github.com/ccipd/QuanTAV, accessed on 9 May 2024).

Prediction models and statistics

In this study, the patient data were first divided into training (55% of cases), validation (15%), and testing (remaining 30%) sets using simple random sampling. Figure 3 shows the workflow to construct the prediction models for the IO response. To identify key radiomic and clinical features for predicting IO outcomes, a two-step feature selection was applied to the training dataset. In the first step, univariate Cox proportional regression for radiomic features and the chi-squared test for clinical features were used to quickly eliminate a large number of redundant features [27, 28], with only those having p < 0.05 advancing to the second step. The second step involved the implementation of a sequential forward selection (SFS) [29] algorithm to further consider feature interactions and eliminate highly correlated features.

Fig. 3
figure 3

The workflow of proposed prediction models for immunotherapy response

Full size image

DeepSurv, a deep learning model built on the Cox proportional hazards (CPH) framework, was employed to predict PFS after IO treatment. This model employed a multilayer perceptron network to embed nonlinear properties into the hazard function, improving the survival prediction [30, 31]. In this study, we used selected radiomic and clinical features as inputs of DeepSurv models. The model output, the estimated hazard function, was further utilized to predict personalized PFS curves for each patient. The architecture and parameter configuration of the proposed DeepSurv model are depicted in Supplementary Figure S1.

Three DeepSurv models were constructed, incorporating different sets of features: (1) clinical features only (e.g., age, sex, hematology, molecular status, metastatic information, and histopathology); (2) clinical and intratumoral radiomic features; and (3) clinical features, intratumoral radiomics, and peritumoral-vasculature radiomics. We further leveraged Shapley Additive Explanations (SHAP) to assess the contribution of individual feature to outcome prediction in DeepSurv models. This process involved sorting and visualizing the importance of each feature in shaping the model’s outcomes [32]. In this study, the hyper-parameters were determined through the grid search method [33]. The selection of hyper-parameters was mainly based on the prediction performance, i.e., concordance index (C-index), in the validation set,. Only when multiple models (with different combinations of hyper-parameters) presented the same C-index value would we select the one with the shortest training time. Supplementary Table S4 presents the selected hyper-parameters based on the grid search.

Patients will be stratified into high-risk and low-risk groups based on the PFS probability predicted by the DeepSurv model at the median PFS time point (3.3 months). Patients with a predicted PFS probability higher than 50% at 3.3 months will be classified as the low-risk group, while those with a predicted PFS probability of 50% or lower will be classified as the high-risk group. A log-rank test was employed to evaluate the statistical differences in Kaplan-Meier (KM) survival curves between the high-risk and low-risk groups. The DeepSurv models were also used to predict progression status at four follow-up time points (i.e., 1, 2, 3, and 6 months after the IO treatment). The performance of DeepSurv models, including time-dependent receiver operating characteristic (ROC) curves, area under the ROC curve (AUC), C-index, sensitivity, and specificity, was evaluated using the testing dataset (remaining 30% of patients). To calculate the C-index, consider all possible pairs of subjects in the dataset. For each pair, determine if the subject with the shorter actual survival time also has a shorter predicted survival time. The C-index is the ratio of concordant pairs to the total number of evaluable pairs [34]:

$$\:C-index=\frac{Number\:of\:concordant\:pairs}{Total\:number\:of\:evaluable\:pairs}$$

To compare performance among three DeepSurv models, we utilized a bootstrap random sampling technique 100 times, followed by the paired t-test of AUC values. Feature selection and training of the DeepSurv models were conducted using the R DeepSurv package (available online: https://rdrr.io/cran/survivalmodels/src/R/deepsurv.R, accessed on 6 May 2023).

Results

Characteristics of enrolled patients

Over 50% of the recruited LC patients were at stage IVB with a median PFS of 3.3 months after the IO treatment. Smokers accounted for more than 55% of the patient cohort. Adenocarcinoma was the predominant histological type (71.3%), and most patients had a wild-type epidermal growth factor receptor (61.1%). Adverse drug reactions among the patients were 18.4%. Detailed clinical characteristics of the 206 recruited NSCLC patients are summarized in Table 1.

Table 1 Characteristics of 206 recruited LC patients receiving immunotherapy
Full size table

Key predictors for prognosis after the IO treatment

A feature selection procedure was employed to identify discriminative features for predicting tumor response. The selected features were utilized to train DeepSurv models. Supplementary Table S5 presents the list of features included in the PFS prediction, along with corresponding hazard ratios, to show the potential influence on the prediction. Visualization results of the SHAP model have been included in Supplementary Figure S2 for additional insight and reference, facilitating further examination of feature contributions to the model. Tumor volume, total protein, and mean corpuscular volume were associated with the PFS. A poorer prognosis was associated with a larger intratumoral local heterogeneity and tumor diameter. Additionally, increased mean torsion on vessel branches surrounding the tumor and overall curvature were identified as features closely associated with poor response to immunotherapy in LC patients.

Prediction performance of DeepSurv models

The distribution of KM curves predicted by DeepSurv models is depicted in Fig. 4. These curves were generated based on three different DeepSurv models with different input features: Model (1) clinical features, Model (2) clinical features and intratumoral radiomics, and Model (3) clinical features, intratumoral radiomics, and peritumoral-vasculature radiomics. Model 2 and Model 3 exhibited significant differences (p < 0.05) in the PFS curves between the high-risk and low-risk groups.

Fig. 4
figure 4

Distribution of KM survival curves predicted by DeepSurv models. The figure displays the KM survival curves for patients in the testing set based on (a) clinical features, (b) clinical + intratumoral radiomics, and (c) clinical + intratumoral + peritumoral-vasculature radiomics

Full size image

We further evaluated the prediction performance of DeepSurv models using the testing dataset at four critical clinical follow-up time points (1, 2, 3, and 6 months). The time-dependent ROC curves displayed in Fig. 5 illustrate the performance of each prediction model. The clinical-based model showed AUCs ranging from 0.68 to 0.77, with a C-index of 0.66. The clinical and intratumoral radiomics-based model demonstrated AUCs between 0.70 and 0.74, with a C-index of 0.68. The model combining clinical, intratumoral, and peritumoral-vasculature radiomics achieved AUCs ranging from 0.76 to 0.80, with a C-index of 0.83. Statistical comparisons of prediction performance between three DeepSurv models are summarized in Table 2. Overall, the DeepSurv model incorporating clinical, intratumoral, and peritumoral-vasculature radiomics significantly outperformed (p < 0.001 in AUCs) two other models for predicting progression risk at each time point.

Fig. 5
figure 5

Time-dependent prediction results for progression-free survival (PFS) after immunotherapy. ROC curves of DeepSurv models based on (a) Clinical Features, (b) Clinical + Intratumoralal Radiomics, and (c) Clinical + Intratumoralal + Peritumoral Radiomics

Full size image
Table 2 Prediction performance and statistical comparisons of PFS among three models
Full size table

Added values to personalized outcome prediction using advanced radiomics

Figure 6 presents the results of intratumoral and peritumoral vascular radiomics in predicting IO. A case with poor IO results (PFS = 0.2 months, Fig. 6a) compared to a good IO result (PFS = 17.3 months, Fig. 6b) showed a higher skewness per branch of peritumoral-vasculature and local heterogeneity within the tumor (Fig. 6c). These vascular characteristics and tumor heterogeneity resulted in poor treatment outcomes. Furthermore, the tumor located near the thoracic wall, as shown in Fig. 6a, presents a poor prognosis.

Fig. 6
figure 6

Representative cases for the IO outcome prediction. The outcome of IO treatment is associated with the characteristics of the peritumoral vasculature and the physical properties of the tumor as observed on CT images. In two cases with similar tumor size, Case 1 (a), with a poorer prognosis, exhibited higher local heterogeneity of the tumor and curvature and tortuosity of the peritumoral vessels compared to Case 2 (b), with a better prognosis. Case 1 (a) also showed growth primarily in the area near the chest wall, with limited distribution of blood vessels around the tumor. These findings suggest that the effects of IO treatment and drug delivery may be influenced by these factors

Full size image

Discussion

The expression of PD-L1 in tumor cells is widely studied and has been approved as a well-known predictor of IO treatment in LC patients. Despite the predictive effect of the PD-L1 biomarker, there are still several limitations to its practical use, including variation over time and tumor locations [35], inconsistent prediction efficacy [36, 37], and unclear interactions with EGFR and ALK mutations [38]. Accordingly, we aimed to identify more reliable predictors of IO outcome to improve clinical management of patients with advanced LC.

Several studies have indicated that utilizing radiomics and machine learning technologies may benefit LC patients with IO [18, 39, 40]. One previous study proposed that analyzing textural changes in CT scans before and after treatment could provide valuable insights into the tumor microenvironment’s evolution, leading to enhanced prediction of IO response [40]. Another study utilized vessel tortuosity-related features and reported an AUC of 0.67 for predicting treatment outcomes of immune checkpoint inhibitors [41]. Although these findings highlighted the potential of radiomics in IO response prediction, the performance in predicting PFS and OS could only achieve a C-index range of 0.61 to 0.64. Our study showed promising results for the PFS prediction at four follow-up time points (AUC = 0.76 to 0.80) with a C-index of 0.83 by integrating intratumoral radiomics, peritumoral-vasculature radiomics, and clinical features with deep-learning model.

We highlighted significance of tumor heterogeneity and characteristics of the peritumoral-vasculature in predicting PFS in LC patients. Patients with worse PFS tended to exhibit increased tumor volumes and global intratumoral heterogeneity while demonstrating reduced local heterogeneity (Supplementary Table S5). This finding was associated with tumor histopathology, where volumetric and textural features derived from CT images might indicate tumor hypoxia status [42]. Larger tumors were more susceptible to hypoxia and exhibited greater global heterogeneity on CT images, which aligned with tumor survival and treatment resistance [43]. Previous studies suggested that the tumor volume was a potential prognostic indicator for the advanced LC [44, 45]. However, in several cases with similar tumor sizes, different outcomes were observed in our patient cohort. Our findings revealed that elevated mean torsion and overall curvature of the peritumoral-vascular branches were strongly associated with the emergence of IO resistance in advanced LC patients, consistent with prior research [41].

Our results revealed a significant correlation (p < 0.05) between the laboratory blood analysis and PFS following IO in LC patients. High mean corpuscular volume and low total protein levels indicated unfavorable PFS outcomes after IO. Previous studies have reported that elevated mean corpuscular volume of red blood cell is associated with potential folate deficiency, which can disrupt crucial processes such as DNA methylation, replication, synthesis, and repair [46, 47]. Conversely, the decline in total serum protein may lead to pronounced constitutional symptoms and compromise tolerance to intensive treatment [48]. These findings underscored the clinical relevance of these blood test indices as prognostic markers in LC patients. Future investigations integrating vessel-based imaging and laboratory biomarkers may enhance patient stratification and outcome.

While integrating advanced radiomics, hematology, and clinical data can offer a more comprehensive prognostic assessment for patients receiving IO, selecting key features and handling the interaction among features was critical to avoid the risk of overfitting and improve the model performance. The proposed DeepSurv model, which is designed by combining multi-layer perceptron architecture with Cox proportional hazards framework, gained considerable attention for survival prediction [31]. The personalized PFS curves generated by the DeepSurv model provided valuable insights into the likelihood of tumor progression following IO. Since tumor progression can occur at various time points during follow-up, using time-dependent ROC curves became crucial for assessing progression status at critical time intervals, especially in advanced LC cohorts. Most patients included in this study had advanced stage IV LC, with a median PFS of approximately three months. Early personalized survival predictions can aid in subsequent patient management.

This study revealed the personalized prognostic benefit of advanced radiomics and deep-learning models in LC patients treated with IO. However, several limitations of this study should be noted. Firstly, it was a retrospective study based on data from a single hospital, which might limit the generalizability of the findings to a broader population. To establish the reliability and robustness of the radiomics and deep learning risk assessment models, further validation using independent and prospective datasets from multiple institutions is required. Secondly, the PD-L1 testing did not show a predictive effect on the response to IO treatment. This could be attributed to the lack of PD-L1 testing results in nearly half of the patients. The missing data might reduce the predictive capability of PD-L1 expression for IO outcome. To address this limitation, prospective data collection is needed to assess the predictive value of PD-L1 expression for IO treatment response. Furthermore, most patients in this cohort received ICIs after chemotherapy or surgery, as IO was not widely used in the first-line treatment of advanced lung cancer. Consequently, the model performance should be further evaluated where IO is employed as the primary treatment. The presence of mortality as a potential competing risk factor in PFS analysis could influence the interpretability of our findings. Moreover, it should be noted that the stability and robustness of DeepSurv models might benefit from including a larger dataset with more comprehensive patient follow-up data. Finally, establishing standardized protocols, including image processing and feature calculation, for peritumoral-vasculature radiomics analysis is essential to enhance the model robustness and reliability.

Conclusions

In conclusion, IO has revolutionized the treatment of LC by targeting immune checkpoints to improve patient outcomes. This study reported the potential of intratumoral and peritumoral-vasculature radiomics with DeepSurv models to predict IO response. Our findings provided insights into tumor heterogeneity, tumor-associated vasculature, and blood analysis as prognostic markers for personalized treatment strategies in advanced LC.

Data availability

Data is provided within the manuscript or supplementary information files. The raw data cannot be publicly available for ethical and legal reasons. However, researchers can submit inquiries for analyzed data to the corresponding authors upon reasonable request.

Abbreviations

LC:

Lung cancer

IO:

Immunotherapy

ICI:

Immune checkpoint inhibitor

OS:

Overall survival

PD-L1:

Programmed death ligand 1

CT:

Computed tomography

PFS:

Progression-free survival

AJCC:

American Joint Committee on Cancer

ROI:

Region of interest

GLCM:

Gray-level co-occurrence matrix

GLRLM:

Gray level run length matrix

LBP:

Local binary pattern

IBSI:

Imaging Biomarker Standardization Initiative

SFS:

Sequential forward selection

SHAP:

SHapley Additive exPlanations

CPH:

Cox proportional hazards

KM:

Kaplan-Meier

ROC:

Receiver operating characteristic

AUC:

Area under the curve

C-index:

Concordance index

EGFR:

Epidermal growth factor receptor

ALK:

Anaplastic lymphoma kinase

References

  1. Gridelli C, et al. Non-small-cell lung cancer. Nat Reviews Disease Primers. 2015;1(1):1–16.

    Google Scholar 

  2. Herbst RS, Morgensztern D, Boshoff C. The biology and management of non-small cell lung cancer. Nature. 2018;553(7689):446–54.

    Article  CAS  PubMed  Google Scholar 

  3. Doroshow DB, et al. Immunotherapy in non–small cell lung cancer: facts and HopesLung cancer immunotherapy: facts and hopes. Clin Cancer Res. 2019;25(15):4592–602.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  4. Jain P, Jain C, Velcheti V. Role of immune-checkpoint inhibitors in lung cancer. Ther Adv Respir Dis. 2018;12:1753465817750075.

    Article  PubMed Central  PubMed  Google Scholar 

  5. Gettinger S, et al. Impaired HLA class I antigen processing and presentation as a mechanism of acquired resistance to immune checkpoint inhibitors in lung CancerAntigen-processing defects and resistance to PD-1 blockade. Cancer Discov. 2017;7(12):1420–35.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  6. Brahmer JR, et al. Management of immune-related adverse events in patients treated with immune checkpoint inhibitor therapy: American Society of Clinical Oncology Clinical Practice Guideline. J Clin Oncol. 2018;36(17):1714–68.

    Article  CAS  PubMed  Google Scholar 

  7. Lesterhuis WJ, Haanen JB, Punt CJ. Cancer immunotherapy–revisited. Nat Rev Drug Discovery. 2011;10(8):591–600.

    Article  CAS  PubMed  Google Scholar 

  8. Xia L, Liu Y, Wang Y. PD-1/PD‐L1 blockade therapy in advanced non‐small‐cell lung cancer: current status and future directions. Oncologist. 2019;24(S1):S31–41.

    Article  PubMed Central  PubMed  Google Scholar 

  9. Kennedy LB, Salama AK. A review of cancer immunotherapy toxicity. Cancer J Clin. 2020;70(2):86–104.

    Article  Google Scholar 

  10. Thom B, et al. The experience of financial toxicity among advanced melanoma patients treated with immunotherapy. J Psychosoc Oncol. 2021;39(2):285–93.

    Article  PubMed  Google Scholar 

  11. Liu M, et al. The value of artificial intelligence in the diagnosis of lung cancer: a systematic review and meta-analysis. PLoS ONE. 2023;18(3):e0273445.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  12. Xie J, et al. Advances in artificial intelligence to predict cancer immunotherapy efficacy. Front Immunol. 2023;13:1076883.

    Article  PubMed Central  PubMed  Google Scholar 

  13. Xu Z, et al. Applying artificial intelligence for cancer immunotherapy. Acta Pharm Sinica B. 2021;11(11):3393–405.

    Article  CAS  Google Scholar 

  14. Chen B et al. Radiomics: an overview in lung cancer management—a narrative review. Annals Translational Med, 2020. 8(18).

  15. Li MO, et al. Transforming growth factor-β regulation of immune responses. Annu Rev Immunol. 2006;24:99–146.

    Article  CAS  PubMed  Google Scholar 

  16. Hendry SA, et al. The role of the tumor vasculature in the host immune response: implications for therapeutic strategies targeting the tumor microenvironment. Front Immunol. 2016;7:621.

    Article  PubMed Central  PubMed  Google Scholar 

  17. Jain RK. Normalizing tumor vasculature with anti-angiogenic therapy: a new paradigm for combination therapy. Nat Med. 2001;7(9):987–9.

    Article  CAS  PubMed  Google Scholar 

  18. Xiao Y, Yu D. Tumor microenvironment as a therapeutic target in cancer. Pharmacol Ther. 2021;221:107753.

    Article  CAS  PubMed  Google Scholar 

  19. Braman N, et al. Novel radiomic measurements of Tumor-Associated vasculature morphology on clinical imaging as a biomarker of treatment response in multiple cancers. Clin Cancer Res. 2022;28(20):4410–24.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  20. Amin MB, et al. The eighth edition AJCC cancer staging manual: continuing to build a bridge from a population-based to a more personalized approach to cancer staging. Cancer J Clin. 2017;67(2):93–9.

    Article  Google Scholar 

  21. Zwanenburg A, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology. 2020;295(2):328–38.

    Article  PubMed  Google Scholar 

  22. Liao C-Y, et al. Enhancement of radiosurgical treatment outcome prediction using MRI radiomics in patients with non-small cell lung cancer brain metastases. Cancers. 2021;13(16):4030.

    Article  PubMed  Google Scholar 

  23. Lu C-F, et al. Machine learning–based Radiomics for Molecular Subtyping of GliomasMachine Learning for Molecular Subtyping of Gliomas. Clin Cancer Res. 2018;24(18):4429–36.

    Article  PubMed  Google Scholar 

  24. Hofmanninger J, et al. Automatic lung segmentation in routine imaging is primarily a data diversity problem, not a methodology problem. Eur Radiol Experimental. 2020;4(1):1–13.

    Article  Google Scholar 

  25. Frangi AF et al. Multiscale vessel enhancement filtering. in Medical Image Computing and Computer-Assisted Intervention—MICCAI’98: First International Conference Cambridge, MA, USA, October 11–13, 1998 Proceedings 1. 1998. Springer.

  26. Sethian JA. A fast marching level set method for monotonically advancing fronts. Proc Natl Acad Sci. 1996;93(4):1591–5.

    Article  CAS  PubMed  Google Scholar 

  27. Wang Y, Zhou C. Feature selection method based on chi-square test and minimum redundancy. in Emerging Trends in Intelligent and Interactive Systems and Applications: Proceedings of the 5th International Conference on Intelligent, Interactive Systems and Applications (IISA2020). 2021. Springer.

  28. Theng D, Bhoyar KK. Feature selection techniques for machine learning: a survey of more than two decades of research. Knowl Inf Syst. 2024;66(3):1575–637.

    Article  Google Scholar 

  29. Mao KZ. Orthogonal forward selection and backward elimination algorithms for feature subset selection. IEEE Trans Syst Man Cybernetics Part B (Cybernetics). 2004;34(1):629–34.

    Article  CAS  Google Scholar 

  30. Lu C-F, et al. A radiomics-based deep learning approach to predict progression free-survival after tyrosine kinase inhibitor therapy in non-small cell lung cancer. Cancer Imaging. 2023;23(1):1–12.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  31. Katzman JL, et al. DeepSurv: personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Med Res Methodol. 2018;18(1):1–12.

    Article  Google Scholar 

  32. Mangalathu S, Hwang S-H, Jeon J-S. Failure mode and effects analysis of RC members based on machine-learning-based SHapley Additive exPlanations (SHAP) approach. Eng Struct. 2020;219:110927.

    Article  Google Scholar 

  33. Bergstra J, Bengio Y. Random search for hyper-parameter optimization. J Mach Learn Res, 2012. 13(2).

  34. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361–87.

    Article  PubMed  Google Scholar 

  35. Topalian SL, et al. Safety, activity, and immune correlates of anti–PD-1 antibody in cancer. N Engl J Med. 2012;366(26):2443–54.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  36. Hellmann MD, et al. Nivolumab plus Ipilimumab in advanced non–small-cell lung cancer. N Engl J Med. 2019;381(21):2020–31.

    Article  CAS  PubMed  Google Scholar 

  37. Carbone DP, et al. First-line nivolumab in stage IV or recurrent non–small-cell lung cancer. N Engl J Med. 2017;376(25):2415–26.

    Article  CAS  PubMed Central  PubMed  Google Scholar 

  38. Gadgeel S, et al. OA09. 03 Pembrolizumab in combination with platinum-based chemotherapy in recurrent EGFR/ALK-positive non-small cell lung Cancer (NSCLC). J Thorac Oncol. 2021;16(10):S863.

    Article  Google Scholar 

  39. Roisman LC, et al. Radiological artificial intelligence-predicting personalized immunotherapy outcomes in lung cancer. NPJ Precision Oncol. 2023;7(1):125.

    Article  Google Scholar 

  40. Gong J, et al. A short-term follow-up CT based radiomics approach to predict response to immunotherapy in advanced non-small-cell lung cancer. Oncoimmunology. 2022;11(1):2028962.

    Article  PubMed  Google Scholar 

  41. Alilou M, et al. A tumor vasculature–based imaging biomarker for predicting response and survival in patients with lung cancer treated with checkpoint inhibitors. Sci Adv. 2022;8(47):eabq4609.

    Article  CAS  PubMed  Google Scholar 

  42. Hammond E, et al. The meaning, measurement and modification of hypoxia in the laboratory and the clinic. Clin Oncol. 2014;26(5):277–88.

    Article  CAS  Google Scholar 

  43. Tunali I et al. Hypoxia-Related radiomics predict immunotherapy response: a multi-cohort study of NSCLC. BioRxiv, 2020: p. 2020.04. 02.020859.

  44. Basaki K, et al. Prognostic factors for survival in stage III non–small-cell lung cancer treated with definitive radiation therapy: impact of tumor volume. Int J Radiation Oncology* Biology* Phys. 2006;64(2):449–54.

    Article  Google Scholar 

  45. Lee P, et al. Metabolic tumor volume is an independent prognostic factor in patients treated definitively for non–small-cell lung cancer. Clin Lung Cancer. 2012;13(1):52–8.

    Article  PubMed  Google Scholar 

  46. Li K-j, et al. High mean corpuscular volume as a predictor of poor overall survival in patients with esophageal cancer receiving concurrent chemoradiotherapy. Cancer Manage Res. 2020;12:7467.

    Article  CAS  Google Scholar 

  47. Kim Y-I. Will mandatory folic acid fortification prevent or promote cancer? Am J Clin Nutr. 2004;80(5):1123–8.

    Article  CAS  PubMed  Google Scholar 

  48. Watanabe T, et al. Pretreatment total serum protein is a significant prognostic factor for the outcome of patients with peripheral T/natural killer-cell lymphomas. Leuk Lymphoma. 2010;51(5):813–21.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

This work was supported by Veterans General Hospitals and University System of Taiwan Joint Research Program (VGHUST113-G1-2-3), and National Science and Technology Council (NSTC 112-2314-B-A49-060, NSTC 113-2314-B-A49-048-MY3). The funding sources had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Funding

This work was supported by Veterans General Hospitals and University System of Taiwan Joint Research Program (VGHUST113-G1-2-3), and National Science and Technology Council (NSTC 112-2314-B-A49-060, NSTC 113-2314-B-A49-048-MY3). The funding sources had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Author information

Author notes

  1. Chien-Yi Liao and Yuh-Min Chen contributed equally to this work.

Authors and Affiliations

  1. Department of Biomedical Imaging and Radiological Sciences, National Yang Ming Chiao Tung University, No. 155, Sec. 2, Linong Street, Beitou District, Taipei, 112, Taiwan

    Chien-Yi Liao, Jyun-Ru Chen & Chia-Feng Lu

  2. Department of Chest Medicine, Taipei Veterans General Hospital, Taipei, Taiwan

    Yuh-Min Chen, Heng-Sheng Chao, Hwa-Yen Chiu & Tsu-Hui Shiao

  3. Institute of Biophotonics, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Yu-Te Wu, Hwa-Yen Chiu, Ting-Wei Wang & Chia-Feng Lu

  4. School of Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Yuh-Min Chen, Hwa-Yen Chiu & Ting-Wei Wang

  5. Brain Research Center, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Yu-Te Wu

  6. Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, USA

    Chien-Yi Liao

Authors
  1. Chien-Yi Liao
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Yuh-Min Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Yu-Te Wu
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Heng-Sheng Chao
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Hwa-Yen Chiu
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Ting-Wei Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Jyun-Ru Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Tsu-Hui Shiao
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Chia-Feng Lu
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Conception and design: C.F. Lu, C.Y. Liao, Y.M. Chen. Acquisition of data: H.S. Chao, H.Y. Chiu, T.H. Shiao, Y.M. Chen. Analysis and interpretation of data: C.F. Lu, C.Y. Liao, T.W. Wang, J.R. Chen. Statistical analysis: C.F. Lu, C.Y. Liao. Drafting the article: C.F. Lu, C.Y. Liao, H.Y. Chiu. Critically revising the article: all authors. Reviewed and approved the submitted version of the manuscript: all authors. Study supervision: C.F. Lu, Y.M. Chen, Y.T. Wu, H.S. Chao.

Corresponding author

Correspondence to Chia-Feng Lu.

Ethics declarations

Ethics approval and consent to participate

The Institutional Review Board of Taipei Veterans General Hospital approved this retrospective study (IRB No.: 2021-09-009BCF) and waived the requirement of acquiring informed consent from patients.

Consent for publication

Not applicable.

Conflict of interest

The authors declare no conflict of interest.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Liao, CY., Chen, YM., Wu, YT. et al. Personalized prediction of immunotherapy response in lung cancer patients using advanced radiomics and deep learning. Cancer Imaging 24, 129 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40644-024-00779-4

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40644-024-00779-4

Keywords

Cancer Imaging

ISSN: 1470-7330

Contact us