Pathology-interpretable radiomic model for predicting clinical outcome in patients with osteosarcoma: a retrospective, multicentre study

doi:10.21203/rs.3.rs-4942890/v1

Download PDF

Research Article

Pathology-interpretable radiomic model for predicting clinical outcome in patients with osteosarcoma: a retrospective, multicentre study

https://doi.org/10.21203/rs.3.rs-4942890/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background: Osteosarcoma is the most prevalent primary malignant bone tumor. Radiomic models demonstrate promise in globally evaluating the prognosis of osteosarcoma; however, they lack biological interpretability. We aimed to develop a radiomic model using MRI to predict disease-free Survival (DFS) in osteosarcoma patients, and to provide underlying pathobiology of the model.

Methods: This retrospective study included 270 patients (training set, n=166; external test set 1, n=56; external test set 2, n=48) with surgically treated and histology-proven osteosarcoma from 14 tertiary centres. A total of 1130 radiomic features were extracted from pre-treatment MRI. After dimensionality reduction, radiomic model was built on the training set and tested on the external test sets. Radiomics interpretability study leveraged the Hematoxylin and eosin (H&E) and Immunohistochemistry (IHC) stained whole slide images (WSIs) of patients from the testing sets. Ten types of nuclear morphological features were extracted from each nucleus in H&E WSIs and aggregated into 150 patient-level features. Moreover, five immune- and hypoxia-related IHC biomarkers—CD3, CD8, CD68, FOXP3, and CAIX—were quantified from IHC WSIs. The correlation between the radiomic features and histopathologic biomarkers was assessed using Spearman correlation analysis.

Results: The radiomic model including 12 features yielded a time-dependent AUC of 0.916 (95% CI: 0.893-0.939), 0.802 (95% CI: 0.763-0.840), and 0.895 (95% CI: 0.869-0.920) in the training set, external test set 1, and external test set 2, respectively. All 12 radiomic features exhibited significant correlations with 109-133 cellular features, totaling 1460 (81.1%) pairs. In detail, there were 574 pairs with absolute coefficient r (|r|) between 0 and 0.1, 516 pairs between 0.1 and 0.2, 241 pairs between 0.2 and 0.3, 99 pairs between 0.3 and 0.4, and 30 pairs exceeding 0.4. Six radiomic features were correlated with CAIX (|r| = 0.03-0.17), 10 features with CD3 (|r| = 0.02-0.71), eight features with CD8 (|r| = 0.05-0.42), nine features with FOXP3 (|r| = 0.01-0.55), 11 features with CD8 / FOXP3 ratio (|r| = 0.004-0.74), and 11 features with CD68 (|r| = 0.02-0.47).

Conclusions: The MRI-based radiomic model effectively predicts DFS in osteosarcoma patients. The correlation strength between radiomic features and histopathologic biomarkers varies.

Osteosarcoma

Magnetic Resonance Imaging

Radiomics

Hematoxylin and eosin

Immunohistochemistry

Disease-Free Survival

Osteosarcoma, a primary malignant bone tumor that typically arises in the metaphyseal regions near growth plates, is the most common and aggressive bone malignancy in children and adolescents [1]. Despite its relatively low incidence, with age-standardized rates of 2.9 per 1,000,000 for males and 2.2 per 1,000,000 for females, its highly invasive nature and tendency for metastasis pose significant therapeutic challenges [2]. The standard treatment paradigm, involving neoadjuvant chemotherapy (NAC) followed by surgery and adjuvant therapies, has markedly improved survival outcomes [3]. Yet, a notable proportion of patients with localized osteosarcoma experience recurrence risks, with 5-year survival rates of 23%-29% [4]. Thus, identifying novel biomarkers for disease progression is critical to advance osteosarcoma treatment. Developing a biomarker-based model for early prediction of disease progression can inform personalized treatment decisions, thereby improving individual patient survival rates.

Magnetic resonance imaging (MRI) has become an indispensable tool for staging, therapeutic response evaluation, and posttreatment surveillance of primary bone tumors. MRI also facilitates personalized treatment prediction and monitoring, serving as an indicator of treatment response or prognosis [5]. While advanced MRI techniques have improved preoperative diagnostic accuracy, the reliance on radiologists' expertise and the limitations of MRI sequences in capturing tumor heterogeneity highlight the need for precise and objective imaging biomarkers. In recent years, radiomic analysis, involving the high-throughput extraction of quantitative image features, has shown potential in predicting the response to NAC in osteosarcoma [6–8]. Serveral studies also have shown that the radiomic models outperformed clinical models in predicting progression of osteosarcoma [9–11]. Nevertheless, these studies have been limited to single-center or dual-center settings, lacking independent validation of model generalizability.

While radiomic models typically rely on a multitude of computational features, the biological meaning of the features often remains elusive, which can hinder model interpretability [12]. A connection between quantitative imaging features and biological processes through interdisciplinary technologies will bridge the gap between proof-of-concept and clinical applicability [13]. Recently, the biological rationale of radiomics has been explored via genomic analysis, histopathological methods, or other experimental assays [12]. Among these, histopathological data is more readily accessible in clinical practice. Hematoxylin and eosin (H&E) and Immunohistochemistry (IHC) stained whole slide images (WSIs) provide valuable insights into the tumor microenvironment (TME), including tissue architecture, cellular morphology, as well as hypoxia and immune factors [14, 15]. Previous studies have elucidated the correlation between histology-depicted TME and clinical outcomes [16–18]. Integrating radiomics with digital pathology can enhance the interpretability of the former by linking imaging features with pathobiology. This approach not only strengthens the reliability of radiomic models but also offers clinicians more understandable predictive insights.

In this current study, we aimed to develop and validate a radiomic model for predicting individual disease-free survival (DFS) in osteosarcoma patients using MRI data from multicentre datasets. The performance of the radiomic model was benchmarked against clinical model, with its robustness evaluated across geographically independent test sets. Furthermore, we also interpret the radiomic model using TME heterogeneity represented by H&E- and IHC-stained WSIs.

Ethics approval

This study was approved by the ethics committee of each participating hospital. Informed consent was waived because of the retrospective nature of the study and the analysis used anonymous clinical and pathological data.

Study design

The overall study design is outlined in Fig. 1. We developed and validated an MRI-based radiomic model to predict DFS in osteosarcoma patients using datasets from multicentres. The generalization ability of radiomic model was validated on two independent external test sets. This work was carried out following the Methodological Radiomics Score (METRICS), a quality scoring tool for radiomics research endorsed by EuSoMII [19]. In addition, we investigated the biological significance of radiomic features using H&E- and IHC-stained WSIs from patients in the external test sets. Ten types of cell-level morphological features were automatically extracted from each nucleus in H&E WSIs and aggregated into a total of 150 patient-level features. Furthermore, hypoxia-related and immune-related TME biomarkers were obtained through IHC analysis. Spearman’s rank correlation analysis was conducted to evaluate the relationship between radiomic features and histopathological features.

Patient cohort and data acquisition

This retrospective study screened 896 osteosarcoma patients from 14 tertiary centres between January 2004 and January 2024. Inclusion criteria were as follows: 1) patients with histopathologically confirmed osteosarcoma based on the surgical specimen; 2) patients who underwent surgery with or without NAC; 3) patients who underwent contrast-enhanced MRI scans one or two weeks prior to treatment; 4) patients who had not received prior antitumor therapy before MRI examinations; and 5) patients without concurrent malignancies. Exclusion criteria were as follows: 1) patients lacking contrast-enhanced MRI images or with insufficient image quality (e.g., artifacts, low signal-to-noise ratio, and low resolution); 2) patients with less than six months of follow-up without disease progression. Finally, 270 patients were included. Figure 2 dispalys the patient recruitment pathway.

The distribution of case numbers across different centers was as follows: Centre 1 (n = 120), Centre 2 (n = 34), Centre 3 (n = 12), Centre 4 (n = 56), Centre 5 (n = 10), Centre 6 (n = 9), Centre 7 (n = 8), Centre 8 (n = 6), Centre 9 (n = 5), Centre 10 (n = 5), Centre 11 (n = 2), Centre 12 (n = 1), Centre 13 (n = 1), and Centre 14 (n = 1). The division of different datasets was based on the geographical distribution of centers. The training set for developing the predictive model included 166 patients from Centres 1 to 3 (101 males and 65 females; mean age, 21.1 years ± 13.2). Independent test set 1 comprised 56 patients from Centre 4 (39 males and 17 females; mean age, 21.4 years ± 11.5), while independent test set 2 comprised 48 patients from Centres 5 to 14 (27 males and 21 females; mean age, 21.4 years ± 12.0).

Baseline clinical, radiological, and pathological data were collected, including age, gender, tumor location, tumor stage, maximum tumor diameter, tumor volume, presence of pathological fracture, and pathological subtypes. All MRI images were acquired from the Picture Archiving and Communication System. Detailed MRI acquisition parameters can be found in the Supplementary Table S1. Tumor volume was calculated as follows: V = π/6 × L × W × H, where L, W, and H represent the length, width, and height of the tumor measured on MRI, respectively [20]. The tumor was staged according to American Joint Committee on Cancer (AJCC) Staging System for Bone Tumors (Eighth Edition) [21]. The presence of pathological fracture at diagnosis was determined by consensus among three experienced radiologists (with more than 10 years of experience). The endpoint was DFS, which was defiend as the time from surgery to disease progression or death. Data collection and evaluation were completed in January 2024.

MRI image normalization, tumor segmentation, and radiomic feature extraction

Contrast-enhanced fat-suppressed T1-weighted images (CE-FS-T1WI) were processed using the N4 bias correction algorithm in 3D Slicer software to mitigate MRI field distortions caused by radiofrequency field inhomogeneity and inherent MR equipment effects. A grayscale normalization algorithm was applied to standardize MRI intensity values to [0, 255], reducing grayscale variations among MRI images from different patients, acquisition times, and parameter settings.

Tumors in the training set were independently segmented by two radiologists (each with 5 years of experience in musculoskeletal imaging) using ITK-SNAP software (www.itksnap.org). All regions of interest (ROI) were manually delineated slice by slice to encompass each gross tumor volume on CE-FS-T1WI. Due to potential inter-observer variability in tumor segmentation, segmentations were evaluated by a senior radiologist specializing in musculoskeletal imaging with 15 years of experience. Final segmentation in the test sets was produced by a single reader, to better reflect clinical practice. Finally, the voxels of the ROI were resampled using a cubic spline interpolation algorithm (1*1*1 mm³), ensuring standardized analysis.

Using the open-source Pyradiomics v3.0 package, radiomic features were extracted from original images and preprocessed images using Laplacian of Gaussian (Log) filter and wavelet transformation. The majority of these features adhere to the definitions outlined by the Imaging Biomarker Standardization Initiative (IBSI) [22]. They encompass four categories: 1) Shape-based features (n = 14); 2) Intensity histogram-based features (n = 72); 3) Matrix texture features (n = 300), including the Grey-level co-occurrence matrix (GLCM), Grey-level run length matrix (GLRLM), Grey-level size zone matrix (GLSZM), Grey-level dependence matrix (GLDM), and Neighbourhood grey-tone difference matrix (NGTDM); and 4) Wavelet features (n = 744). A total of 1130 radiomic features were extracted from CE-FS-T1WI (see Supplementary Materials for specific feature names).

Radiomic model development

Prior to feature selection, radiomic feature values were normalized to [0, 1] using the MinMax algorithm. Subsequently, Pearson correlation analysis was applied to initially identify a minimally redundant feature set (with a correlation coefficient threshold set at 0.9). Further feature selection was further performed using Relief algorithm. Various machine learning classifiers, including Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Adaptive Boosting (AdaBoost), Multilayer Perceptron (MLP), and Naive Bayes (NB), were applied. Five-fold cross-validation was utilized for feature selection and algorithm optimization to construct the top-performing radiomic model. The entire procedure was conducted on the training set, while the external test sets were used exclusively for model validation. Standardization parameters obtained from the training set were applied to preprocess the independent test set data to evaluate the model's generalization performance.

The potential association of radiomic signature (ie, Radscore) and DFS was assessed in the training set and then validated in the external test sets 1 and 2 using Kaplan-Meier survival analysis. The patients were divided into high- and low-risk subgroups according to the predictive Radscores; the threshold of which was identified by using X-tile software. The difference in the survival curves of the high- and low-risk subgroups was evaluated by using a weighted log-rank test.

Clinical models were established by employing univariate and multivariate Cox proportional hazards regression analysis to screen clinical characteristics. Likewise, utilizing the multivariate Cox regression method, the combined model was constructed by integrating significant clinical features and Radscores. The predictive performance of the radiomic model was compared with that of the clinical model, tumor volume, and combined model.

Radiology-pathology correlation analysis: Interpretability of the radiomic model

H&E WSI-derived morphological nuclear features

We collected 41 patients from the external test set 1, each with available H&E-stained slides of the osteosarcoma. The slides were digitalized into SVS format using the Digital Pathology Slide Scanner KF-PRO-400 (Ningbo, Zhejiang, China) at a 40x magnification. All the WSIs were manually checked by a senior pathologist for artifacts, and the images free of all types of artifacts were chosen. We pre-processed each WSI with a pixel threshold to filter out white background, where pixels with a mean value of the RGB channels greater than 210 were considered as background. We then divided all the non-background areas into non-overlapping 2048*2048 patches and leveraged an unsupervised segmentation algorithm [23] to segment cell nuclei on the patches. For each segmented nucleus, we extracted 10 types of morphological nuclear features, including nuclear area, lengths of the major and minor axes of cell nucleus, the ratio of major axis length to minor axis length (major, minor, and ratio), mean pixel values of nucleus in RGB three channels, and mean, maximum, and minimum distances (distant, distMax, and distMin) to its neighboring nuclei in the Delaunay triangulation graph [24] (see Supplementary Table S3). Finally, for each type of morphological nuclear feature, a 10-bin histogram and five statistic measurements (i.e., mean, standard deviation, skewness, kurtosis, and entropy) were used to aggregate the cell-level features and generate a 150-dimensional imaging feature for each patient (see Supplementary Table S4). All feature extraction was implemented using MATLAB R2021a.

IHC-derived hypoxia-immune-related biomarkers

Forty patients from the external test sets with available formalin-fixed paraffin-embedded (FFPE) tissue samples were included. FFPE samples were cut into 3-mm thick sections, which were then processed for IHC staining (see Appendix E1 of Supplementary Materials for details). The detected immune-related and hypoxia-related IHC biomarkers included CD3 (pan T cells), CD8 (cytotoxic T cells), CD68 (macrophages), FOXP3 (Treg cells), and Carbonic Anhydrase IX (CAIX) (Supplementary Figure S1). IHC evaluation was performed independently by two pathologists (with 5 and 10 years of experience, respectively) who were blinded to the outcome data. In cases of disagreement between the primary pathologists, a third pathologist was consulted to reach consensus. In detail, the tissue sections were initially screened at low power (100x) using an upright microscope (BX53; Olympus, Japan), and five most representative fields were selected. The density of immune and hypoxic cells was then assessed at 200× magnification within tumor regions. The cells stained for nuclear (FOXP3) and membrane (CD3, CD8, CD68, and CAIX) biomarkers were quantified and expressed as the number of cells per unit area (cells/mm²).

Correlation analysis

The 150 patient-level cytomorphological features derived from H&E WSIs, along with five hypoxia-immune-related biomarkers derived from IHC, were directly analyzed to assess Spearman’s rank correlation with the selected radiomic features. Correlation coefficients below 0.3 were considered weak correlations, 0.3 to 0.7 as moderate correlations, and > 0.7 as strong correlations [25].

Statistical Analysis

We compared two groups using the t-test for continuous variables and the chi-square test or Fisher's exact test for categorical variables. The predictive performance of the radiomic model was evaluated using time-dependent receiver operating characteristic (ROC) curves, calibration curve, decision curve analysis (DCA), and clinical impact curve (CIC). Time-dependent ROC analysis was conducted using the "timeROC" package. The DeLong test was employed to compare the area under the curve (AUC) between models. Calibration curve was generated using the "rms" package. Hosmer-Lemeshow test was carried out using the "ResourceSelection" package to examine the goodness of fit between model-predicted and observed probability. The DCA and CIC were plotted using the "rmda" package in R. The DCA illustrates the net benefit of using the predictive models compared to other strategies (such as always treating or never treating) at various probability thresholds [26]. The CIC shows the potential impact of adopting the predictive model on clinical decision outcomes across various probability thresholds [27]. The Shapley Additive Explanations (SHAP) algorithm was used to visually quantify the contribution of each feature within the radiomic model using “shap” package [28]. Statistical significance was set at a P value < 0.05. Statistical analyses were conducted using R version 4.4.0 and Python 3.7.6.

Role of funding source

The funders played no role in the study design, data collection, data analysis, data interpretation, and writing of the report.

Patient characteristics

Baseline characteristics of the included patients are summarized in Table 1. A total of 270 patients were included in this retrospective study. The training set included 166 patients (median age of 17 years, 60.8% male), the external test set included 56 patients (median age of 18 years, 69.6% male), and the external test set included 48 patients (median age of 17 years, 56.3% male).

Table 1

Clinical characteristics of patients in the training and testing sets
Characteristics	Training set (n = 166)	External test set 1 (n = 56)	External test set 2 (n = 48)
Age (median [IQR], years)	17.0 [13.0, 24.0]	18.0 [14.0, 23.0]	17.0 [13.0, 23.0]
Gender
Female	65 (39.2)	17 (30.4)	21 (43.8)
Male	101 (60.8)	39 (69.6)	27 (56.3)
Tumor location
Long bone	122 (73.49)	47 (83.93)	40 (83.3)
Flat bone	44 (26.51)	9 (16.07)	8 (16.7)
Pathologic fracture
No	151 (91.0)	50 (89.3)	37 (77.1)
Yes	15 (9.0)	6 (10.7)	11 (22.9)
Pathological subtype#
Conventional	129 (77.7)	53 (94.6)	41 (85.4)
Others	37 (22.3)	3 (5.4)	7 (14.6)
Tumor size (median [IQR], cm)	8.3 [6.8, 11.0]	7.8 [6.6, 9.6]	6.8 [5.7, 8.4]
Tumor volume (median [IQR], cm³)	203.8 [95.3, 377.1]	256.5 [156.0, 421.3]	210.9 [95.4, 367.2]
T stage
T1	76 (45.8)	31 (55.4)	35 (72.9)
T2	87 (52.4)	25 (44.6)	12 (25.0)
T3	3 (1.8)	0	1 (2.1)
N stage
N0	165 (99.4)	56 (100)	48 (100)
N1	1 (0.6)	0	0
M stage
M0	155 (93.4)	49 (87.5)	46 (95.8)
M1*	11 (6.6)	7 (12.5)	2 (4.2)
Neoadjuvant chemotherapy†
Yes	156 (94.0)	8 (14.3)	28 (58.3)
No	6 (3.6)	47 (83.9)	13 (27.1)
Unknown	4 (2.4)	1 (1.8)	7 (14.6)
Note: Unless otherwise specified, data are numbers of patients, with percentages in parentheses. IQR, interquartile range. # Conventional osteosarcoma is the most common type, comprising various subtypes such as osteoblastic, chondroblastic, fibroblastic, and fibrohistiocytic osteosarcoma. *All distant metastases were lung metastases. †Some patients received a minimum of 4–6 cycles (ie, 2–3 months) of NAC before undergoing surgical resection, adhering to the National Comprehensive Cancer Network guidelines.

Predictive performance of the radiomic model

After performing dimensionality reduction, we identified 12 radiomic features significantly associated with clinical outcome, comprising four first-order features and eight textural features (Supplementary Table S5). Comparision of six machine learning algorithms using cross-validation revealed that the AdaBoost-based radiomic model achieved the highest performance (Fig. 3a-c), with AUC values of 0.916 (95% CI: 0.893–0.939), 0.802 (95% CI: 0.763–0.840), and 0.895 (95% CI: 0.869–0.920) in the training set, external test set 1, and external test set 2, respectively.

Tumor volume is well recognized as a critical prognostic factor in solid tumors. We evaluated its predictive accuracy for DFS in osteosarcoma patients, with time-dependent AUC values of 0.632 (95% CI: 0.582–0.681), 0.623 (95% CI: 0.575–0.671), and 0.527 (95% CI: 0.480–0.574) in the training set, external test set 1, and external test set 2, respectively (Table 2 and Fig. 3d-f). Following Cox analyses of clinical variables, only tumor T stage emerged as an independent risk factor (Supplementary Table S6), yielding AUC values of 0.633 (95% CI: 0.591–0.676), 0.567 (95% CI: 0.525–0.609), and 0.511 (95% CI: 0.474–0.548) in the same sets (Table 2 and Fig. 3d-f). Hence, a combined model integrating tumor T stage and Radscore achieved AUC values of 0.841 (95% CI: 0.808–0.875), 0.756 (95% CI: 0.717–0.794), and 0.789 (95% CI: 0.751–0.828) in these respective sets (Table 2 and Fig. 3d-f). In comparison, Delong tests indicated that the radiomic model significantly outperformed tumor T stage, tumor volume, and the combined model (all P < 0.05). Thus, tumor T stage did not contribute additional predictive value beyond the radiomic model. The dynamic change of model performance over time further validated these findings (Fig. 3g-i). Figure 3j-k illustrate the SHAP values, reflecting the contributions of different radiomic features to the model's prediction.

Table 2

Predictive performance of the models
Dataset	Model	AUC (95%CI)	P-value	Sensitivity (%)	Specificity (%)	Accuracy (%)
Training set	Tumor T stage	0.633 (0.591–0.676)	< 0.001	66.7	58.5	62.7
	Tumor volume	0.632 (0.582–0.681)	< 0.001	58.3	67.1	62.7
	Radiomic model	0.916 (0.893–0.939)	Ref.	91.7	70.7	81.3
	Combined model	0.841 (0.808–0.875)	< 0.001	75.0	59.8	67.5
External test set 1	Tumor T stage	0.567 (0.525–0.609)	< 0.001	48.7	64.7	53.6
	Tumor volume	0.623 (0.575–0.671)	< 0.001	53.8	70.6	58.9
	Radiomic model	0.802 (0.763–0.840)	Ref.	79.5	64.7	75.0
	Combined model	0.756 (0.717–0.794)	0.031	71.8	52.9	66.1
External test set 2	Tumor T stage	0.511 (0.474–0.548)	< 0.001	28.6	75.0	47.9
	Tumor volume	0.527 (0.480–0.574)	< 0.001	50.0	55.0	52.1
	Radiomic model	0.895 (0.869–0.920)	Ref.	96.4	60.0	81.3
	Combined model	0.789 (0.751–0.828)	< 0.001	67.9	65.0	66.7

The radiomic model could classify patients into high- and low-risk subgroups with distinct survival probabilities based on a threshold of 0.49 (Fig. 4a-c). The calibration curves (Fig. 4d-f) show good agreement between radiomic model-predicted DFS probabilities and actual DFS probabilities (Hosmer-Lemeshow P values = 0.596, 0.581, 0.317, respectively). The decision curves show that the radiomic model achieves a superior net benefit compared to the T stage and tumor volume (Fig. 4g-i). The CICs show that the radiomic model provided a superior overall net beneft within the wide ranges (65%-100%) of threshold probabilities and impacted patient outcomes in both the training and test sets (Fig. 4j-l).

The reporting and methodological quality of this present study was evaluated by the METRICS tool (https://metricsscore.github.io/metrics/METRICS.html), which includes 30 items across 9 categories. The total METRICS score of our study was 93.7%, suggesting excellent quality (see Supplementary Figure S2).

Radiopathomic correlation analysis: Interpretability of the radiomic model

To elucidate the underlying pathobiology of radiomic features, we conducted Spearman correlation analysis involving 150 morphological nuclear features and 12 top-predictive radiomic features (Fig. 5a). Overall, all 12 radiomic features showed significant correlations with 109–133 (72.7%-88.7%) nuclear morphological features, yielding 1460 (81.1%) correlation pairs. In detail, there were 574 (39.3%) correlation pairs with absolute coefficients |r| between 0 and 0.1, 516 (35.3%) pairs with |r| between 0.1 and 0.2, 241 (16.5%) pairs with |r| between 0.2 and 0.3, 99 (6.8%) pairs with |r| between 0.3 and 0.4, and 30 (2.0%) pairs with |r| exceeding 0.4 (up to 0.55) (Fig. 5b). Figure 5c-d depicts the count and proportion of nuclear feature types within each radiomic feature. The chord plot shows significant radiology-pathology correlations (all P < 0.05). Specially, the correlogram (Fig. 5e) shows 129 (8.8%) correlation pairs with |r| > 0.3 (moderate to strong correlation).

To deepen our understanding of how the radiomic features related to TME, we performed Spearman rank correlation analyses with IHC-derived hypoxia-immune-related biomarkers (Fig. 5f). The results demonstrate varying degrees of correlations between the radiomic features and these biomarkers. Overall, six radiomic features were correlated with CAIX (|r| = 0.03–0.17), 10 features with CD3 (|r| = 0.02–0.71), eight features with CD8 (|r| = 0.05–0.42), nine features with FOXP3 (|r| = 0.01–0.55), 11 features with CD8 / FOXP3 (|r| = 0.004–0.74), and 11 features with CD68 (|r| = 0.02–0.47). There were 23 pairs showing negative correlations (r ranging from − 0.004 to -0.26): two radiomic features correlated with CAIX (both r = -0.17), two features with CD3 (r = -0.02 and − 0.03), two features with CD8 (r = -0.05 and

-0.06), four features with FOXP3 (r ranging from − 0.01 to -0.08), eight features with CD8 / FOXP3 (r ranging from − 0.004 to -0.259), and five features with CD68 (r ranging from − 0.07 to -0.13). There were 32 pairs demonstrating positive correlations (r ranging from 0.02–0.74): four radiomic features correlated with CAIX (r ranging from 0.03 to 0.23), eight features with CD3 (r ranging from 0.02 and 0.71), six features with CD8 (r ranging from 0.12 and 0.42), five features with FOXP3 (r ranging from 0.21 to 0.55), three features with CD8 / FOXP3 (r ranging from 0.025 to 0.736), and six features with CD68 (r ranging from 0.02 to 0.47). Significantly, there were 19 correlation pairs (34.5%) with r values ranging from 0.3 to 0.7 (moderate correlation), and two pairs (3.6%) with r values exceeding 0.7 (strong correlation).

In this work, we developed and validated a MRI-based radiomic model that can accurately predict DFS in patients with osteosarcoma, which outperformed empirical clinical markers such as tumor T stage and tumor volume. Furthermore, we constructed a cross-scale correlation framework between macroscopic radiomic features and microscopic digital pathology-based TME features to interpret the underlying pathobiology of the former. We demonstrate that all identified radiomic features correlated with 72.7–88.7% of cellular morphological features extracted from H&E-stained WSIs, with 129 (8.8%) pairs exhibiting moderate or higher correlations. Moreover, almost all radiomic features showed significant correlations with at least three immune- or hypoxia-related biomarkers, particularly the CD68.

Radiological imaging allows for noninvasive, comprehensive evaluation of the entire tumor and can be repeated as needed throughout the treatment course. Radiomics refers to the process of extracting quantitative data from routine medical images, which can capture the inherent heterogeneity within tumors [29]. This approach involves advanced computational analysis techniques that transform standard medical images into mineable data [30]. By extracting a large number of quantitative features from images, radiomics allows for a more global characterization of tumor heterogeneity than traditional visual or qualitative assessments alone [31]. Understanding and quantifying tumor heterogeneity through radiomics can provide valuable insights into tumor biology, prognosis, and treatment response. Previous radiomic studies have employed mainly in treatment responses to NAC and prognosis prediction in patients with osteosarcoma [6–11]. However, most studies lacked external validation and had small sample sizes due to the rarity of osteosarcoma (around 100), which could lead to an overestimation of model predictive accuracy or overfitting. Improvements in study design and validation are necessary to show the generalizability of findings and to facilitate clinical applications. Although our study did not have a large sample size, to the best of our knowledge, it represents the largest sample size in previous research on MRI radiomics predicting treatment response and prognosis in osteosarcoma. Our study demonstrates that the imaging-based radiomic model containing four first-order and eight texture features significantly outperformed clinicopathologic factors such as T stage and tumor volume, demonstrating superior prognostic accuracy. Given its complementary value, this model has the potential to enhance the precision of the current staging system and improve risk stratification for osteosarcoma. One strength of our study is the validation of the prognostic model across diverse populations.

A significant obstacle to the clinical translation of current data-driven radiomic models is their lack of interpretability, often stemming from a disconnect from the underlying pathobiology [13]. The absence of biological interpretability in the model leads to our inability to intuitively understand the mechanisms behind the predictions. Although some previous studies [32–36] have utilized transcriptomic data to decipher the biological functions of radiomic features; yet, compared to other omics data such as transcriptomics and proteomics, pathological data is biologically closer to imaging data, may suggesting potentially stronger correlations between them. Moreover, pathological images contain rich information regarding biological behavior of tumors, with robust evidence showing that nuclear heterogeneity is closely linked to tumor prognosis [37]. In this current study, we extracted ten types of morphological features of tumor cell nuclei to quantitatively evaluate nuclear heterogeneity and then aggregated them to 150 patient-level features. We studied the correlation between the significant radiomic features and nuclear morphology, and the results showed that all 12 radiomic features (not only first-order but also texture ones) exhibited varying degrees of significant correlation with most nuclear morphological features, which reflects that nuclear heterogeneity to some extent determines the heterogeneity of imaging. Among the correlation coefficients exceeding 0.30, 10 radiomic features exhibited significant correlations with 51 (34%) nuclear features, notably including two first-order and four texture features (one GLRLM and three GLCM).

The prognostic significance of the TME has been well established across various solid tumors. The TME of osteosarcoma exhibits highly immunosuppressive properties and is often described as an immune "cold" tumor [38]. Prior studies have shown varying levels of exhaustion in CD8⁺ T cells in the osteosarcoma microenvironment [39]. The intratumoral accumulation of FOXP3⁺ Tregs has been identified as a major immune evasion mechanism in various tumors, including osteosarcoma [40]. Fritzsching et al. [41] showed that osteosarcoma patients with higher intratumoral CD8⁺/FOXP3⁺ ratios had better outcomes compared to those with lower ratios. Our study also showed that prognostic radiomic features were associated with the CD8/FOXP3 ratio, and these correlations were stronger compared to using a single biomarker alone (Fig. 5f). Previous reports suggest that tumor-associated macrophages (TAMs) and CD3 + T cells are the predominant tumor-infiltrating immune cells in osteosarcoma, with M2 macrophages being the most abundant immune cell type [42, 43]. TAMs are closely related to tumor growth, angiogenesis, invasion, and metastasis, rendering them key targets for therapy [44]. Apart from the immunologic microenvironment, hypoxia is also an inherent feature of osteosarcoma, which is related to tumor invasion, treatment resistance, and metastasis [45]. Tumors overexpressing hypoxic biomarkers such as HIF-1α and CAIX are associated with worse outcomes [46]. In the present study, we analyzed 40 FFPE tissue specimens using IHC and digitally quantified one hypoxia-related and four immune-related biomarkers. Our study also observed that the TME of osteosarcoma is predominantly characterized by macrophages and exhibits an immunosuppressive state. Correlation analysis revealed that all radiomic features were linked to the immune response, particularly tumor macrophage infiltration, consistent with findings from a previous study on glioma radiomics [36]. The weak correlation between the hypoxia marker CAIX and radiomic features may be due to the sampling specimens not accurately reflecting the extent of necrosis; however, further investigation is needed to understand this phenomenon. Overall, radiomic features show a stronger correlation with TME biomarkers compared to nuclear features.

Our study also has some limitations. First, due to the retrospective multicenter collection of MRI data, the radiomic model was inevitably affected by variations in MR image acquisition and scanning parameters. To mitigate this issue, we conducted external testing to evaluate how data heterogeneity affects model performance. Moreover, we implemented grayscale normalization to minimize variations in grayscale levels across MRI images. Second, we utilized only CE-FS-T1WI images to construct the radiomic model. While multi-parametric MRI combining CE-FS-T1WI, T2-weighted imaging, and diffusion weighted imaging may potentially enhance model discrimination in predicting survival outcomes. However, as per the recommendation of the METRICS scoring tool, using a single MRI sequence rather than multiple should be preferred, as multi-parametric imaging may unnecessarily increase data dimensionality and risk of overfitting [19]. Third, manual segmentation is crucial for the best model accuracy. This is mainly because with current deep learning it is challenging to reliably segment osteosarcoma due to the inherent nature of these tumors, such as irregular shapes and indistinct boundary with surrounding edema. Finally, despite the favorable performance of the radiomic model on two external test sets, further validation in prospective settings and diverse ethnicities should be thoroughly conducted.

In conclusion, our MRI-based radiomic model accurately predicts DFS in osteosarcoma patients with significant clinical benefits, offering crucial evidence for personalized treatment approaches. The varying degrees of correlations observed between H&E- and IHC-derived TME biomarkers and radiomic features validate the biological underpinnings of the radiomic features, affirming its interpretability at the cellular level and molecular level. This interdisciplinary approach facilitates a thorough evaluation of tumor characteristics spanning from microscopic to macroscopic levels, enhancing our understanding of tumor heterogeneity and guiding the development of customized treatment strategies.

DFS: Disease-free Survival

H&E: Hematoxylin and eosin

IHC: Immunohistochemistry

WSIs: Whole slide images

MRI: Magnetic resonance imaging

NAC: Neoadjuvant chemotherapy

TME: Tumor microenvironment

METRICS: Methodological Radiomics Score

CE-FS-T1WI: Contrast-enhanced fat-suppressed T1-weighted images

ROI: Regions of interest

GLCM: Grey-level co-occurrence matrix

GLRLM: Grey-level run length matrix

GLSZM: Grey-level size zone matrix

GLDM: Grey-level dependence matrix

NGTDM: Neighbourhood grey-tone difference matrix

RF: Random Forest

SVM: Support Vector Machine

LR: Logistic Regression

AdaBoost: Adaptive Boosting

MLP: Multilayer Perceptron

NB: Naive Bayes

FFPE: Formalin-fixed paraffin-embedded

CD3: Pan T cells

CD8: Cytotoxic T cells

CD68: macrophages

FOXP3: Treg cells

CAIX: Carbonic Anhydrase IX

ROC: Receiver operating characteristic

DCA: Decision curve analysis

CIC: Clinical impact curve

AUC: Area under the curve

SHAP: Shapley Additive Explanations

Abbreviations

DFS, disease-free survival; DCA, decision curve analysis; CIC, clinical impact curve.

Author Contribution

Conceptualization: Q.P.R, G.W., B.Z., G. W., X.DZ. Data curation: Q.P.R., H.Z., Y.X. Z. , Y.B.Y., Y.P.L., Y.M.L., Y.M.H., J.E., R.F.H.Formal analysis: X.Z., X.W.X., B.Z., X.Y.Z., N.T. Funding acquisition: S.X.Z. Project administration: Q.P.R., B. Z. Resources: G.Z., X.D.Z., S.X.Z., and B.Z. Supervision: B.Z., S.X.Z. Visualization: B.Z., S.X.Z., X.D.Z., and G.W. Roles/Writing - original draft: Q.P.R., and B.Z. Verified the underlying data: X.Z., and B.Z. Writing - review & editing: B.Z., X.Z., and X.W.W. All authors read and approved the final version of the manuscript.

Acknowledgement

Not applicable.

Gianferante DM, Mirabello L, Savage SA. Germline and somatic genetics of osteosarcoma — connecting aetiology, biology and therapy. Nat Rev Endocrinol. 2017;13:480–91.
Meltzer PS, Helman LJ. New Horizons in the Treatment of Osteosarcoma. N Engl J Med. 2021;385:2066–76.
Strauss SJ, Frezza AM, Abecassis N, Bajpai J, Bauer S, Biagini R, et al. Bone sarcomas: ESMO–EURACAN–GENTURIS–ERN PaedCan Clinical Practice Guideline for diagnosis, treatment and follow-up. Annals of Oncology. 2021;32:1520–36.
Kempf-Bielack B, Bielack SS, Jürgens H, Branscheid D, Berdel WE, Exner GU, et al. Osteosarcoma Relapse After Combined Modality Therapy: An Analysis of Unselected Patients in the Cooperative Osteosarcoma Study Group (COSS). JCO. 2005;23:559–68.
Hao Y, An R, Xue Y, Li F, Wang H, Zheng J, et al. Prognostic value of tumoral and peritumoral magnetic resonance parameters in osteosarcoma patients for monitoring chemotherapy response. Eur Radiol. 2021;31:3518–29.
Bouhamama A, Leporq B, Khaled W, Nemeth A, Brahmi M, Dufau J, et al. Prediction of Histologic Neoadjuvant Chemotherapy Response in Osteosarcoma Using Pretherapeutic MRI Radiomics. Radiology: Imaging Cancer. 2022;4:e210107.
Chen H, Zhang X, Wang X, Quan X, Deng Y, Lu M, et al. MRI-based radiomics signature for pretreatment prediction of pathological response to neoadjuvant chemotherapy in osteosarcoma: a multicenter study. Eur Radiol. 2021;31:7913–24.
Zhong J, Zhang C, Hu Y, Zhang J, Liu Y, Si L, et al. Automated prediction of the neoadjuvant chemotherapy response in osteosarcoma with deep learning and an MRI-based radiomics nomogram. Eur Radiol. 2022;32:6196–206.
White LM, Atinga A, Naraghi AM, Lajkosz K, Wunder JS, Ferguson P, et al. T2-weighted MRI radiomics in high-grade intramedullary osteosarcoma: predictive accuracy in assessing histologic response to chemotherapy, overall survival, and disease-free survival. Skeletal Radiol. 2023;52:553–64.
Zhao S, Su Y, Duan J, Qiu Q, Ge X, Wang A, et al. Radiomics signature extracted from diffusion-weighted magnetic resonance imaging predicts outcomes in osteosarcoma. Journal of Bone Oncology. 2019;19:100263.
Luo Z, Li J, Liao Y, Liu R, Shen X, Chen W. Radiomics Analysis of Multiparametric MRI for Prediction of Synchronous Lung Metastases in Osteosarcoma. Front Oncol. 2022;12:802234.
Tomaszewski MR, Gillies RJ. The Biological Meaning of Radiomic Features. Radiology. 2021;299:E256–E256.
Jiang Y, Zhang Z, Wang W, Huang W, Chen C, Xi S, et al. Biology-guided deep learning predicts prognosis and cancer immunotherapy response. Nat Commun. 2023;14:5135.
Wu X, Yan H, Qiu M, Qu X, Wang J, Xu S, et al. Comprehensive characterization of tumor microenvironment in colorectal cancer via molecular analysis. Elife. 2023;12:e86032.
Yoo S-Y, Park HE, Kim JH, Wen X, Jeong S, Cho N-Y, et al. Whole-Slide Image Analysis Reveals Quantitative Landscape of Tumor–Immune Microenvironment in Colorectal Cancers. Clin Cancer Res. 2020;26:870–81.
Zhao K, Li Z, Yao S, Wang Y, Wu X, Xu Z, et al. Artificial intelligence quantified tumour-stroma ratio is an independent predictor for overall survival in resectable colorectal cancer. EBioMedicine. 2020;61:103054.
Lin H, Pan X, Feng Z, Yan L, Hua J, Liang Y, et al. Automated whole-slide images assessment of immune infiltration in resected non-small-cell lung cancer: towards better risk-stratification. J Transl Med. 2022;20:261.
Park S, Ock C-Y, Kim H, Pereira S, Park S, Ma M, et al. Artificial Intelligence–Powered Spatial Analysis of Tumor-Infiltrating Lymphocytes as Complementary Biomarker for Immune Checkpoint Inhibition in Non–Small-Cell Lung Cancer. JCO. 2022;40:1916–28.
Kocak B, Akinci D’Antonoli T, Mercaldo N, Alberich-Bayarri A, Baessler B, Ambrosini I, et al. METhodological RadiomICs Score (METRICS): a quality scoring tool for radiomics research endorsed by EuSoMII. Insights Imaging. 2024;15:8.
Pflanzer R, Hofmann M, Shelke A, Habib A, Derwich W, Schmitz-Rixen T, et al. Advanced 3D-Sonographic Imaging as a Precise Technique to Evaluate Tumor Volume. Translational Oncology. 2014;7:681–6.
Amin MB, Greene FL, Edge SB, Compton CC, Gershenwald JE, Brookland RK, 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. CA A Cancer J Clinicians. 2017;67:93–9.
Zwanenburg A, Vallières M, Abdalah MA, Aerts HJWL, Andrearczyk V, Apte A, et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology. 2020;295:328–38.
Ahmady Phoulady H, Goldgof DB, Hall LO, Mouton PR. Nucleus segmentation in histology images with hierarchical multilevel thresholding. In: Gurcan MN, Madabhushi A, editors. Medical Imaging 2016: Digital Pathology. San Diego, California, United States: SPIE; 2016. p. 979111.
Yang Y, Li F, Gao L, Wang Z, Thrall MJ, Shen SS, et al. Differential diagnosis of breast cancer using quantitative, label-free and molecular vibrational imaging. Biomed Opt Express. 2011;2:2160.
Ratner B. The correlation coefficient: Its values range between +1/−1, or do they? J Target Meas Anal Mark. 2009;17:139–42.
Vickers AJ, Elkin EB. Decision Curve Analysis: A Novel Method for Evaluating Prediction Models. Med Decis Making. 2006;26:565–74.
Kerr KF, Brown MD, Zhu K, Janes H. Assessing the Clinical Impact of Risk Prediction Models With Decision Curves: Guidance for Correct Interpretation and Appropriate Use. JCO. 2016;34:2534–40.
Luxburg U von, Guyon I, Bengio S, Wallach H, Fergus R, Vishwanathan SVN, et al., editors. A unified approach to interpreting model predictions. Red Hook, NY, USA: Curran Associates Inc.; 2017.
Lambin P, Leijenaar RTH, Deist TM, Peerlings J, De Jong EEC, Van Timmeren J, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749–62.
Shur JD, Doran SJ, Kumar S, Ap Dafydd D, Downey K, O’Connor JPB, et al. Radiomics in Oncology: A Practical Guide. Radiographics. 2021;41:1717–32.
Eloyan A, Yue MS, Khachatryan D. Tumor heterogeneity estimation for radiomics in cancer. Statistics in Medicine. 2020;39:4704–23.
Lin D-F, Li H-L, Liu T, Lv X-F, Xie C-M, Ou X-M, et al. Radiomic signatures associated with tumor immune heterogeneity predict survival in locally recurrent nasopharyngeal carcinoma. JNCI: J Natl Cancer Inst. 2024;116:1294–302.
Yan J, Zhang S, Li KK-W, Wang W, Li K, Duan W, et al. Incremental prognostic value and underlying biological pathways of radiomics patterns in medulloblastoma. EBioMedicine. 2020;61:103093.
Guan F, Wang Z, Qiu Y, Guo Y, Pei D, Wang M, et al. Biological underpinnings of radiomic magnetic resonance imaging phenotypes for risk stratification in IDH wild-type glioblastoma. J Transl Med. 2023;21:841.
Feng Z, Li H, Liu Q, Duan J, Zhou W, Yu X, et al. CT Radiomics to Predict Macrotrabecular-Massive Subtype and Immune Status in Hepatocellular Carcinoma. Radiology. 2023;307:e221291.
Li G, Li L, Li Y, Qian Z, Wu F, He Y, et al. An MRI radiomics approach to predict survival and tumour-infiltrating macrophages in gliomas. Brain. 2022;145:1151–61.
Cheng J, Zhang J, Han Y, Wang X, Ye X, Meng Y, et al. Integrative Analysis of Histopathological Images and Genomic Data Predicts Clear Cell Renal Cell Carcinoma Prognosis. Cancer Res. 2017;77:e91–100.
Zhou Y, Yang D, Yang Q, Lv X, Huang W, Zhou Z, et al. Single-cell RNA landscape of intratumoral heterogeneity and immunosuppressive microenvironment in advanced osteosarcoma. Nat Commun. 2020;11:6322.
Shi S, Zhang L, Guo X. Comprehensive investigation of tumor immune microenvironment and prognostic biomarkers in osteosarcoma through integrated bulk and single-cell transcriptomic analysis. Front Immunol. 2024;15:1424950.
Saleh R, Elkord E. FoxP3+ T regulatory cells in cancer: Prognostic biomarkers and therapeutic targets. Cancer Letters. 2020;490:174–85.
Fritzsching B, Fellenberg J, Moskovszky L, Sápi Z, Krenacs T, Machado I, et al. CD8 ⁺ /FOXP3 ⁺ -ratio in osteosarcoma microenvironment separates survivors from non-survivors: a multicenter validated retrospective study. OncoImmunology. 2015;4:e990800.
Yang Z, Liu L, Zhu Z, Hu Z, Liu B, Gong J, et al. Tumor‐Associated Monocytes Reprogram CD8 ⁺ T Cells into Central Memory‐Like Cells with Potent Antitumor Effects. Advanced Science. 2024;11:2304501.
Deng C, Xu Y, Fu J, Zhu X, Chen H, Xu H, et al. Reprograming the tumor immunologic microenvironment using neoadjuvant chemotherapy in osteosarcoma. Cancer Science. 2020;111:1899–909.
Yang Q, Guo N, Zhou Y, Chen J, Wei Q, Han M. The role of tumor-associated macrophages (TAMs) in tumor progression and relevant advance in targeted therapy. Acta Pharmaceutica Sinica B. 2020;10:2156–70.
Liao C, Liu X, Zhang C, Zhang Q. Tumor hypoxia: From basic knowledge to therapeutic implications. Seminars in Cancer Biology. 2023;88:172–86.
Forker L, Gaunt P, Sioletic S, Shenjere P, Potter R, Roberts D, et al. The hypoxia marker CAIX is prognostic in the UK phase III VorteX-Biobank cohort: an important resource for translational research in soft tissue sarcoma. Br J Cancer. 2018;118:698–704.

No competing interests reported.

SupplementaryMaterials.docx

Download PDF

Reviewers agreed at journal
25 Oct, 2024
Reviews received at journal
08 Oct, 2024
Reviewers agreed at journal
10 Sep, 2024
Reviewers invited by journal
04 Sep, 2024
Editor assigned by journal
20 Aug, 2024
Submission checks completed at journal
20 Aug, 2024
First submitted to journal
20 Aug, 2024

You are reading this latest preprint version

Pathology-interpretable radiomic model for predicting clinical outcome in patients with osteosarcoma: a retrospective, multicentre study

Status:

Version 1

Abstract

Figures

Background

Methods

Ethics approval

Study design

Patient cohort and data acquisition

MRI image normalization, tumor segmentation, and radiomic feature extraction

Radiomic model development

Radiology-pathology correlation analysis: Interpretability of the radiomic model

Statistical Analysis

Role of funding source

Results

Patient characteristics

Predictive performance of the radiomic model

Radiopathomic correlation analysis: Interpretability of the radiomic model

Discussion

Abbreviations

Declarations

Abbreviations

Author Contribution

Acknowledgement

References

Additional Declarations

Supplementary Files

Status:

Version 1