MR Imaging Radiomics Signatures for Predicting the Risk of Breast Cancer Recurrence as Given by Research Versions of MammaPrint, Oncotype DX, and PAM50 Gene Assays

Global contextual representation via graph-transformer fusion for hepatocellular carcinoma prognosis in whole-slide images

Current methods of digital pathological images typically employ small image patches to learn local representative features to overcome the issues of computationally heavy and memory limitations. However, the global contextual features are not fully considered in whole-slide images (WSIs). Here, we designed a hybrid model that utilizes Graph Neural Network (GNN) module and Transformer module for the representation of global contextual features, called TransGNN. GNN module built a WSI-Graph for the foreground area of a WSI for explicitly capturing structural features, and the Transformer module through the self-attention mechanism implicitly learned the global context information. The prognostic markers of hepatocellular carcinoma (HCC) prognostic biomarkers were used to illustrate the importance of global contextual information in cancer histopathological analysis. Our model was validated using 362 WSIs from 355 HCC patients diagnosed from The Cancer Genome Atlas (TCGA). It showed impressive performance with a Concordance Index (C-Index) of 0.7308 (95% Confidence Interval (CI): (0.6283-0.8333)) for overall survival prediction and achieved the best performance among all models. Additionally, our model achieved an area under curve of 0.7904, 0.8087, and 0.8004 for 1-year, 3-year, and 5-year survival predictions, respectively. We further verified the superior performance of our model in HCC risk stratification and its clinical value through Kaplan-Meier curve and univariate and multivariate COX regression analysis. Our research demonstrated that TransGNN effectively utilized the context information of WSIs and contributed to the clinical prognostic evaluation of HCC.

Imaging at the nexus: how state of the art imaging techniques can enhance our understanding of cancer and fibrosis

Both cancer and fibrosis are diseases involving dysregulation of cell signaling pathways resulting in an altered cellular microenvironment which ultimately leads to progression of the condition. The two disease entities share common molecular pathophysiology and recent research has illuminated the how each promotes the other. Multiple imaging techniques have been developed to aid in the early and accurate diagnosis of each disease, and given the commonalities between the pathophysiology of the conditions, advances in imaging one disease have opened new avenues to study the other. Here, we detail the most up-to-date advances in imaging techniques for each disease and how they have crossed over to improve detection and monitoring of the other. We explore techniques in positron emission tomography (PET), magnetic resonance imaging (MRI), second generation harmonic Imaging (SGHI), ultrasound (US), radiomics, and artificial intelligence (AI). A new diagnostic imaging tool in PET/computed tomography (CT) is the use of radiolabeled fibroblast activation protein inhibitor (FAPI). SGHI uses high-frequency sound waves to penetrate deeper into the tissue, providing a more detailed view of the tumor microenvironment. Artificial intelligence with the aid of advanced deep learning (DL) algorithms has been highly effective in training computer systems to diagnose and classify neoplastic lesions in multiple organs. Ultimately, advancing imaging techniques in cancer and fibrosis can lead to significantly more timely and accurate diagnoses of both diseases resulting in better patient outcomes.

AI Applications to Breast MRI: Today and Tomorrow

In breast imaging, there is an unrelenting increase in the demand for breast imaging services, partly explained by continuous expanding imaging indications in breast diagnosis and treatment. As the human workforce providing these services is not growing at the same rate, the implementation of artificial intelligence (AI) in breast imaging has gained significant momentum to maximize workflow efficiency and increase productivity while concurrently improving diagnostic accuracy and patient outcomes. Thus far, the implementation of AI in breast imaging is at the most advanced stage with mammography and digital breast tomosynthesis techniques, followed by ultrasound, whereas the implementation of AI in breast magnetic resonance imaging (MRI) is not moving along as rapidly due to the complexity of MRI examinations and fewer available dataset. Nevertheless, there is persisting interest in AI-enhanced breast MRI applications, even as the use of and indications of breast MRI continue to expand. This review presents an overview of the basic concepts of AI imaging analysis and subsequently reviews the use cases for AI-enhanced MRI interpretation, that is, breast MRI triaging and lesion detection, lesion classification, prediction of treatment response, risk assessment, and image quality. Finally, it provides an outlook on the barriers and facilitators for the adoption of AI in breast MRI. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 6.

Comparative Effectiveness of Mammography, Ultrasound, and MRI in the Detection of Breast Carcinoma in Dense Breast Tissue: A Systematic Review

This systematic review aimed to critically assess the effectiveness of mammography, ultrasound, and magnetic resonance imaging (MRI) in the detection of breast carcinoma within dense breast tissue. An exhaustive search of contemporary literature was undertaken, focusing on the diagnostic accuracy, false positive and negative rates, and clinical implications of the aforementioned imaging modalities. Each modality was assessed in isolation and side by side against the others to draw comparative inferences. While mammography remains a foundational imaging modality, its effectiveness waned within the context of dense breast tissue. Ultrasound demonstrated a strong differentiation prowess, especially among specific demographic cohorts. MRI, despite its exceptional precision and differentiation capabilities, exhibited a tendency for slightly elevated false positive rates. No single modality emerged as singularly superior for all cases. Instead, an integrated approach, combining the strengths of each modality based on individual patient profiles and clinical scenarios, is recommended. This tailored approach ensures optimized detection rates and minimizes diagnostic ambiguities, underscoring the significance of individualized patient care in the field of diagnostic radiology.

Radiomics Model for Predicting FOXP3 Expression Level and Survival in Clear Cell Renal Carcinoma

RATIONALE AND OBJECTIVES

We aimed to evaluate the predictive significance of forkhead box protein 3 (FOXP3) expression levels among individuals with clear cell renal carcinoma (ccRCC) and establish a radiomics model for predicting FOXP3 expression.

MATERIALS AND METHODS

430 patients with ccRCC were included in the gene-based prognostic analyses; 100 samples were used for radiomics feature generation, model development, and evaluation. A gradient boosting machine was employed to model the selected radiomics features. The developed model generated radiomics scores (RS) that predicted FOXP3 expression. The FOXP3 prognostic model combining imaging features was applied for survival and clinical indicator correlation analyses.

RESULTS

FOXP3 was highly expressed in patients with ccRCC and served as an independent predictive marker (hazard ratio [HR]=2.357, 95% CI [confidence interval]: 1.582-3.511, p < 0.001). The radiomics model formed by three radiomics characteristics was identified as a strong prognostic indicator of overall survival (OS). The predictive power of the model was commendable (areas under the curve: 0.835 and 0.809 for training and validation sets, respectively). Significant between-group variations in RS distribution were identified, as indicated by gene expression levels (p < 0.05). Disparities were observed in pathological stage, pharmaceutical therapy, and neoplasm status between low and high RS cohorts (p < 0.001). Kaplan-Meier curves revealed a significant correlation between increased RS and decreased OS (p = 0.001), which was also observed in the multivariate analyses (HR=3.411, 95% CI: 1.039-11.196, p = 0.043).

CONCLUSION

Prognostic outcome of ccRCC is closely linked to FOXP3 expression level. Computed tomography-based radiomics shows promise for prognostic prediction in ccRCC.

Exploring a multiparameter MRI-based radiomics approach to predict tumor proliferation status of serous ovarian carcinoma

OBJECTIVES

To develop a multiparameter magnetic resonance imaging (MRI)-based radiomics approach that can accurately predict the tumor cell proliferation status of serous ovarian carcinoma (SOC).

MATERIALS AND METHODS

A total of 134 patients with SOC who met the inclusion and exclusion criteria were retrospectively screened from institution A, spanning from January 2016 to March 2022. Additionally, an external validation set comprising 42 SOC patients from institution B was also included. The region of interest was determined by drawing each ovarian mass boundaries manually slice-by-slice on T2-weighted imaging fat-suppressed fast spin-echo (T2FSE) and T1 with contrast enhancement (T1CE) images using ITK-SNAP software. The handcrafted radiomic features were extracted, and then were selected using variance threshold algorithm, SelectKBest algorithm, and least absolute shrinkage and selection operator. The optimal radiomic scores and the clinical/radiological independent predictors were integrated as a combined model.

RESULTS

Compared with the area under the curve (AUC) values of each radiomic signature of T2FSE and T1CE, respectively, the AUC value of the radiomic signature (T1CE-T2FSE) was the highest in the training set (0.999 vs. 0.965 and 0.860). The homogeneous solid component of the ovarian mass was considered the only independent predictor of tumor cell proliferation status among the clinical/radiological variables. The AUC of the radiomic-radiological model was 0.999.

CONCLUSIONS

The radiomic-radiological model combining radiomic scores and the homogeneous solid component of the ovarian mass can accurately predict tumor cell proliferation status of SOC which has high repeatability and may enable more targeted and effective treatment strategies.

CRITICAL RELEVANCE STATEMENT

The proposed radiomic-radiological model combining radiomic scores and the homogeneous solid component of the ovarian mass can predict tumor cell proliferation status of SOC which has high repeatability and may guide individualized treatment programs.

KEY POINTS

• The radiomic-radiological nomogram may guide individualized treatment programs of SOC. • This radiomic-radiological nomogram showed a favorable prediction ability. • Homogeneous slightly higher signal intensity on T2FSE is vital for Ki-67.

Radiomics for the prediction of a postoperative pancreatic fistula following a pancreatoduodenectomy: A systematic review and radiomic score quality assessment

BACKGROUND

Postoperative pancreatic fistula (POPF) is a severe complication following a pancreatoduodenectomy. An accurate prediction of POPF could assist the surgeon in offering tailor-made treatment decisions. The use of radiomic features has been introduced to predict POPF. A systematic review was conducted to evaluate the performance of models predicting POPF using radiomic features and to systematically evaluate the methodological quality.

METHODS

Studies with patients undergoing a pancreatoduodenectomy and radiomics analysis on computed tomography or magnetic resonance imaging were included. Methodological quality was assessed using the Radiomics Quality Score (RQS) and Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement.

RESULTS

Seven studies were included in this systematic review, comprising 1300 patients, of whom 364 patients (28 %) developed POPF. The area under the curve (AUC) of the included studies ranged from 0.76 to 0.95. Only one study externally validated the model, showing an AUC of 0.89 on this dataset. Overall adherence to the RQS (31 %) and TRIPOD guidelines (54 %) was poor.

CONCLUSION

This systematic review showed that high predictive power was reported of studies using radiomic features to predict POPF. However, the quality of most studies was poor. Future studies need to standardize the methodology.

REGISTRATION

not registered.

Prediction of a Multi-Gene Assay (Oncotype DX and Mammaprint) Recurrence Risk Group Using Machine Learning in Estrogen Receptor-Positive, HER2-Negative Breast Cancer-The BRAIN Study

This study aimed to develop a machine learning-based prediction model for predicting multi-gene assay (MGA) risk categories. Patients with estrogen receptor-positive (ER+)/HER2- breast cancer who had undergone Oncotype DX (ODX) or MammaPrint (MMP) were used to develop the prediction model. The development cohort consisted of a total of 2565 patients including 2039 patients tested with ODX and 526 patients tested with MMP. The MMP risk prediction model utilized a single XGBoost model, and the ODX risk prediction model utilized combined LightGBM, CatBoost, and XGBoost models through soft voting. Additionally, the ensemble (MMP + ODX) model combining MMP and ODX utilized CatBoost and XGBoost through soft voting. Ten random samples, corresponding to 10% of the modeling dataset, were extracted, and cross-validation was performed to evaluate the accuracy on each validation set. The accuracy of our predictive models was 84.8% for MMP, 87.9% for ODX, and 86.8% for the ensemble model. In the ensemble cohort, the sensitivity, specificity, and precision for predicting the low-risk category were 0.91, 0.66, and 0.92, respectively. The prediction accuracy exceeded 90% in several subgroups, with the highest prediction accuracy of 95.7% in the subgroup that met Ki-67 <20 and HG 1~2 and premenopausal status. Our machine learning-based predictive model has the potential to complement existing MGAs in ER+/HER2- breast cancer.

Magnetic resonance imaging-based radiomics was used to evaluate the level of prognosis-related immune cell infiltration in breast cancer tumor microenvironment

PURPOSE

The tumor immune microenvironment is a valuable source of information for predicting prognosis in breast cancer (BRCA) patients. To identify immune cells associated with BRCA patient prognosis from the Cancer Genetic Atlas (TCGA), we established an MRI-based radiomics model for evaluating the degree of immune cell infiltration in breast cancer patients.

METHODS

CIBERSORT was utilized to evaluate the degree of infiltration of 22 immune cell types in breast cancer patients from the TCGA database, and both univariate and multivariate Cox regressions were employed to determine the prognostic significance of immune cell infiltration levels in BRCA patients. We identified independent prognostic factors for BRCA patients. Additionally, we obtained imaging features from the Cancer Imaging Archive (TCIA) database for 73 patients who underwent preoperative MRI procedures, and used the Least Absolute Shrinkage and Selection Operator (LASSO) to select the best imaging features for constructing an MRI-based radiomics model for evaluating immune cell infiltration levels in breast cancer patients.

RESULTS

According to the results of Cox regression analysis, M2 macrophages were identified as an independent prognostic factor for BRCA patients (HR = 32.288, 95% CI: 3.100-357.478). A total of nine significant features were selected to calculate the radiomics-based score. We established an intratumoral model with AUCs (95% CI) of 0.662 (0.495-0.802) and 0.678 (0.438-0.901) in the training and testing cohorts, respectively. Additionally, a peritumoral model was created with AUCs (95% CI) of 0.826 (0.710-0.924) and 0.752 (0.525-0.957), and a combined model was established with AUCs (95% CI) of 0.843 (0.723-0.938) and 0.744 (0.491-0.965). The peritumoral model demonstrated the highest diagnostic efficacy, with an accuracy, sensitivity, and specificity of 0.773, 0.727, and 0.818, respectively, in its testing cohort.

CONCLUSION

The MRI-based radiomics model has the potential to evaluate the degree of immune cell infiltration in breast cancer patients, offering a non-invasive imaging biomarker for assessing the tumor microenvironment in this disease.

Prediction of the MammaPrint Risk Group Using MRI Features in Women With Estrogen Receptor-Positive, HER2-Negative, and 1 to 3 Node-Positive Invasive Breast Cancer

BACKGROUND

MammaPrint assigns chemotherapeutic benefits to patients with estrogen receptor (ER)-positive, human epidermal growth factor receptor 2 (HER2)-negative, and 1 to 3 node-positive invasive breast cancer. However, its cost and time-consuming nature limit its use in certain clinical settings. We aimed to develop and validate the prediction models for the low MammaPrint risk group using clinicopathologic and MRI features.

PATIENTS AND METHODS

Overall, 352 women with ER-positive, HER2-negative, and 1 to 3 node-positive invasive breast cancer were retrospectively reviewed and assigned to development (n = 235) and validation sets (n = 117). Univariate and multivariate analyses identified features associated with the low MammaPrint risk group. The area under the receiver operating characteristic curves (AUROCs) of models based on clinicopathologic, MRI, and combined features were evaluated.

RESULTS

Development set multivariate analysis showed that clinicopathologic features including low histologic grade (odds ratio [OR], 5.29; P = .02), progesterone receptor-positivity (OR, 3.23; P = .01), and low Ki-67 (OR, 6.05; P < .001) and MRI features, including peritumoral edema absence (OR, 2.24; P = .04) and a high proportion of persistent components (OR, 1.15; P = .004) were significantly associated with the low MammaPrint risk group. The AUROCs of models based on clinicopathologic, MRI, and combined features were 0.77, 0.64, and 0.80 in the development and 0.66, 0.60, and 0.70 in the validation sets, respectively.

CONCLUSION

The combined model incorporating clinicopathologic and MRI features showed potential in predicting the low MammaPrint risk group, and may support decision-making in clinical settings with limited access to MammaPrint.

Peri-lesion regions in differentiating suspicious breast calcification-only lesions specifically on contrast enhanced mammography

PURPOSE

The explore the added value of peri-calcification regions on contrast-enhanced mammography (CEM) in the differential diagnosis of breast lesions presenting as only calcification on routine mammogram.

METHODS

Patients who underwent CEM because of suspicious calcification-only lesions were included. The test set included patients between March 2017 and March 2019, while the validation set was collected between April 2019 and October 2019. The calcifications were automatically detected and grouped by a machine learning-based computer-aided system. In addition to extracting radiomic features on both low-energy (LE) and recombined (RC) images from the calcification areas, the peri-calcification regions, which is generated by extending the annotation margin radially with gradients from 1 mm to 9 mm, were attempted. Machine learning (ML) models were built to classify calcifications into malignant and benign groups. The diagnostic matrices were also evaluated by combing ML models with subjective reading.

RESULTS

Models for LE (significant features: wavelet-LLL_glcm_Imc2_MLO; wavelet-HLL_firstorder_Entropy_MLO; wavelet-LHH_glcm_DifferenceVariance_CC; wavelet-HLL_glcm_SumEntropy_MLO;wavelet-HLH_glrlm_ShortRunLowGray LevelEmphasis_MLO; original_firstorder_Entropy_MLO; original_shape_Elongation_MLO) and RC (significant features: wavelet-HLH_glszm_GrayLevelNonUniformityNormalized_MLO; wavelet-LLH_firstorder_10Percentile_CC; original_firstorder_Maximum_MLO; wavelet-HHH_glcm_Autocorrelation_MLO; original_shape_Elongation_MLO; wavelet-LHL_glszm_GrayLevelNonUniformityNormalized_MLO; wavelet-LLH_firstorder_RootMeanSquared_MLO) images were set up with 7 features. Areas under the curve (AUCs) of RC models are significantly better than those of LE models with compact and expanded boundary (RC v.s. LE, compact: 0.81 v.s. 0.73, p < 0.05; expanded: 0.89 v.s. 0.81, p < 0.05) and RC models with 3 mm boundary extension yielded the best performance compared to those with other sizes (AUC = 0.89). Combining with radiologists' reading, the 3mm-boundary RC model achieved a sensitivity of 0.871 and negative predictive value of 0.937 with similar accuracy of 0.843 in predicting malignancy.

CONCLUSIONS

The machine learning model integrating intra- and peri-calcification regions on CEM has the potential to aid radiologists' performance in predicting malignancy of suspicious breast calcifications.

Assessment of Background Parenchymal Enhancement at Dynamic Contrast-enhanced MRI in Predicting Breast Cancer Recurrence Risk

Background Background parenchymal enhancement (BPE) at dynamic contrast-enhanced (DCE) MRI of cancer-free breasts increases the risk of developing breast cancer; implications of quantitative BPE in ipsilateral breasts with breast cancer are largely unexplored. Purpose To determine whether quantitative BPE measurements in one or both breasts could be used to predict recurrence risk in women with breast cancer, using the Oncotype DX recurrence score as the reference standard. Materials and Methods This HIPAA-compliant retrospective single-institution study included women diagnosed with breast cancer between January 2007 and January 2012 (development set) and between January 2012 and January 2017 (internal test set). Quantitative BPE was automatically computed using an in-house-developed computer algorithm in both breasts. Univariable logistic regression was used to examine the association of BPE with Oncotype DX recurrence score binarized into high-risk (recurrence score >25) and low- or intermediate-risk (recurrence score ≤25) categories. Models including BPE measures were assessed for their ability to distinguish patients with high risk versus those with low or intermediate risk and the actual recurrence outcome. Results The development set included 127 women (mean age, 58 years ± 10.2 [SD]; 33 with high risk and 94 with low or intermediate risk) with an actual local or distant recurrence rate of 15.7% (20 of 127) at a minimum 10 years of follow-up. The test set included 60 women (mean age, 57.8 years ± 11.6; 16 with high risk and 44 with low or intermediate risk). BPE measurements quantified in both breasts were associated with increased odds of a high-risk Oncotype DX recurrence score (odds ratio range, 1.27-1.66 [95% CI: 1.02, 2.56]; P < .001 to P = .04). Measures of BPE combined with tumor radiomics helped distinguish patients with a high-risk Oncotype DX recurrence score from those with a low- or intermediate-risk score, with an area under the receiver operating characteristic curve of 0.94 in the development set and 0.79 in the test set. For the combined models, the negative predictive values were 0.97 and 0.93 in predicting actual distant recurrence and local recurrence, respectively. Conclusion Ipsilateral and contralateral DCE MRI measures of BPE quantified in patients with breast cancer can help distinguish patients with high recurrence risk from those with low or intermediate recurrence risk, similar to Oncotype DX recurrence score. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Zhou and Rahbar in this issue.

AI-powered interpretable imaging phenotypes noninvasively characterize tumor microenvironment associated with diverse molecular signatures and survival in breast cancer

BACKGROUND AND OBJECTIVES

Tumor microenvironment (TME) is a determining factor in decision-making and personalized treatment for breast cancer, which is highly intra-tumor heterogeneous (ITH). However, the noninvasive imaging phenotypes of TME are poorly understood, even invasive genotypes have been largely known in breast cancer.

METHODS

Here, we develop an artificial intelligence (AI)-driven approach for noninvasively characterizing TME by integrating the predictive power of deep learning with the explainability of human-interpretable imaging phenotypes (IMPs) derived from 4D dynamic imaging (DCE-MRI) of 342 breast tumors linked to genomic and clinical data, which connect cancer phenotypes to genotypes. An unsupervised dual-attention deep graph clustering model (DGCLM) is developed to divide bulk tumor into multiple spatially segregated and phenotypically consistent subclusters. The IMPs ranging from spatial heterogeneity to kinetic heterogeneity are leveraged to capture architecture, interaction, and proximity between intratumoral subclusters.

RESULTS

We demonstrate that our IMPs correlate with well-known markers of TME and also can predict distinct molecular signatures, including expression of hormone receptor, epithelial growth factor receptor and immune checkpoint proteins, with the performance of accuracy, reliability and transparency superior to recent state-of-the-art radiomics and 'black-box' deep learning methods. Moreover, prognostic value is confirmed by survival analysis accounting for IMPs.

CONCLUSIONS

Our approach provides an interpretable, quantitative, and comprehensive perspective to characterize TME in a noninvasive and clinically relevant manner.

Molecular-subtype guided automatic invasive breast cancer grading using dynamic contrast-enhanced MRI

BACKGROUND AND OBJECTIVES

Histological grade and molecular subtype have presented valuable references in assigning personalized or precision medicine as the significant prognostic indicators representing biological behaviors of invasive breast cancer (IBC). To evaluate a two-stage deep learning framework for IBC grading that incorporates with molecular-subtype (MS) information using DCE-MRI.

METHODS

In Stage I, an innovative neural network called IOS2-DA is developed, which includes a dense atrous-spatial pyramid pooling block with a pooling layer (DA) and inception-octconved blocks with double kernel squeeze-and-excitations (IOS2). This method focuses on the imaging manifestation of IBC grades and performs preliminary prediction using a novel class F1-score loss function. In Stage II, a MS attention branch is introduced to fine-tune the integrated deep vectors from IOS2-DA via Kullback-Leibler divergence. The MS-guided information is weighted with preliminary results to obtain classification values, which are analyzed by ensemble learning for tumor grade prediction on three MRI post-contrast series. Objective assessment is quantitatively evaluated by receiver operating characteristic curve analysis. DeLong test is applied to measure statistical significance (P < 0.05).

RESULTS

The molecular-subtype guided IOS2-DA performs significantly better than the single IOS2-DA in terms of accuracy (0.927), precision (0.942), AUC (0.927, 95% CI: [0.908, 0.946]), and F1-score (0.930). The gradient-weighted class activation maps show that the feature representations extracted from IOS2-DA are consistent with tumor areas.

CONCLUSIONS

IOS2-DA elucidates its potential in non-invasive tumor grade prediction. With respect to the correlation between MS and histological grade, it exhibits remarkable clinical prospects in the application of relevant clinical biomarkers to enhance the diagnostic effectiveness of IBC grading. Therefore, DCE-MRI tends to be a feasible imaging modality for the thorough preoperative assessment of breast biological behavior and carcinoma prognosis.

An integrated model combined intra- and peritumoral regions for predicting chemoradiation response of non small cell lung cancers based on radiomics and deep learning

PURPOSE

The purpose of this study was to develop a model for predicting chemoradiation response in non-small cell lung cancer (NSCLC) patients by integrating radiomics and deep-learning features and combined intra- and peritumoral regions with pre-treated CT images.

MATERIALS AND METHODS

This study enrolled 462 patients with NSCLC who received chemoradiation. On the basis of pretreated CT images, we developed three models to compare the prediction of chemoradiation: intratumoral, peritumoral and combined regions. To further illustrate each model, we established different feature integration methods: a) radiomics model with 1500 features; b) deep learning model with a multiple instance learning algorithm; c) integrated model by integrating radiomic and deep learning features. For radiomics and integrated models, support vector machine and the least absolute shrinkage and selection operator were used to extract and select features. Transfer learning and max pooling algorithms were used to identify high informative features in deep learning models. We applied ten-fold cross validation in model training and testing.

RESULTS

The best area under the curve (AUC) of intratumoral, peritumoral and combined models were 0.89 (95% CI, 0.74-0.93), 0.86 (95% CI, 0.75-0.92) and 0.92 (95% CI, 0.81-0.95), respectively. It indicated the importance of the peritumoral region for treatment response prediction and should be used in combination with the intratumoral region. Integrated models gave better results than models with radiomics and deep learning features alone in all regions of interest and radiomics models outperformed deep learning models in any comparative models.

CONCLUSIONS

The model that integrate radiomic and deep learning features and combined intra- and peritumoral regions provide valuable information in predicting treatment response of chemoradiation. It can help oncologists customize personalized clinical treatment plans for NSCLC patients.

Artificial intelligence in breast imaging: potentials and challenges

Breast cancer, which is the most common type of malignant tumor among humans, is a leading cause of death in females. Standard treatment strategies, including neoadjuvant chemotherapy, surgery, postoperative chemotherapy, targeted therapy, endocrine therapy, and radiotherapy, are tailored for individual patients. Such personalized therapies have tremendously reduced the threat of breast cancer in females. Furthermore, early imaging screening plays an important role in reducing the treatment cycle and improving breast cancer prognosis. The recent innovative revolution in artificial intelligence (AI) has aided radiologists in the early and accurate diagnosis of breast cancer. In this review, we introduce the necessity of incorporating AI into breast imaging and the applications of AI in mammography, ultrasonography, magnetic resonance imaging, and positron emission tomography/computed tomography based on published articles since 1994. Moreover, the challenges of AI in breast imaging are discussed.

Radiogenomic analysis of cellular tumor-stroma heterogeneity as a prognostic predictor in breast cancer

BACKGROUND

The tumor microenvironment and intercellular communication between solid tumors and the surrounding stroma play crucial roles in cancer initiation, progression, and prognosis. Radiomics provides clinically relevant information from radiological images; however, its biological implications in uncovering tumor pathophysiology driven by cellular heterogeneity between the tumor and stroma are largely unknown. We aimed to identify radiogenomic signatures of cellular tumor-stroma heterogeneity (TSH) to improve breast cancer management and prognosis analysis.

METHODS

This retrospective multicohort study included five datasets. Cell subpopulations were estimated using bulk gene expression data, and the relative difference in cell subpopulations between the tumor and stroma was used as a biomarker to categorize patients into good- and poor-survival groups. A radiogenomic signature-based model utilizing dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) was developed to target TSH, and its clinical significance in relation to survival outcomes was independently validated.

RESULTS

The final cohorts of 1330 women were included for cellular TSH biomarker identification (n = 112, mean age, 57.3 years ± 14.6) and validation (n = 886, mean age, 58.9 years ± 13.1), radiogenomic signature of TSH identification (n = 91, mean age, 55.5 years ± 11.4), and prognostic (n = 241) assessments. The cytotoxic lymphocyte biomarker differentiated patients into good- and poor-survival groups (p < 0.0001) and was independently validated (p = 0.014). The good survival group exhibited denser cell interconnections. The radiogenomic signature of TSH was identified and showed a positive association with overall survival (p = 0.038) and recurrence-free survival (p = 3 × 10-4).

CONCLUSION

Radiogenomic signatures provide insights into prognostic factors that reflect the imbalanced tumor-stroma environment, thereby presenting breast cancer-specific biological implications and prognostic significance.

Five machine learning-based radiomics models for preoperative prediction of histological grade in hepatocellular carcinoma

PURPOSE

To compare the efficacy of radiomics models via five machine learning algorithms in predicting the histological grade of hepatocellular carcinoma (HCC) before surgery and to develop the most stable model to classify high-risk HCC patients.

METHODS

Contrast-enhanced computed tomography (CECT) images of 175 HCC patients before surgery were analysed, and radiomics features were extracted from CECT images (including arterial and portal phases). Five machine learning models, including Bayes, random forest (RF), k-nearest neighbors (KNN), logistic regression (LR), and support vector machine (SVM), were applied to establish the model. The stability of the five models was weighed by the relative standard deviation (RSD), and the lowest RSD value was chosen as the most stable model to predict the histological grade of HCC. The area under the curve (AUC) and Delong tests were devoted to assessing the predictive efficacy of the models.

RESULTS

High-grade HCC accounted for 28.57% (50/175) of the 175 patients. The RSD value of AUC via the RF machine learning model was the lowest (2.3%), followed by Bayes (3.2%), KNN (6.4%), SVM (8.7%) and LR (31.3%). In addition, the RF model (AUC = 0.995) was better than the other four models in the training set (p < 0.05), as well as obtained good predictive performance in the test set (AUC = 0.837).

CONCLUSION

Among the five machine learning models, the RF-based radiomics model was the most stable and performed excellently in identifying high histological grade of HCC.

Recent advancements in artificial intelligence for breast cancer: Image augmentation, segmentation, diagnosis, and prognosis approaches

Breast cancer is a significant global health burden, with increasing morbidity and mortality worldwide. Early screening and accurate diagnosis are crucial for improving prognosis. Radiographic imaging modalities such as digital mammography (DM), digital breast tomosynthesis (DBT), magnetic resonance imaging (MRI), ultrasound (US), and nuclear medicine techniques, are commonly used for breast cancer assessment. And histopathology (HP) serves as the gold standard for confirming malignancy. Artificial intelligence (AI) technologies show great potential for quantitative representation of medical images to effectively assist in segmentation, diagnosis, and prognosis of breast cancer. In this review, we overview the recent advancements of AI technologies for breast cancer, including 1) improving image quality by data augmentation, 2) fast detection and segmentation of breast lesions and diagnosis of malignancy, 3) biological characterization of the cancer such as staging and subtyping by AI-based classification technologies, 4) prediction of clinical outcomes such as metastasis, treatment response, and survival by integrating multi-omics data. Then, we then summarize large-scale databases available to help train robust, generalizable, and reproducible deep learning models. Furthermore, we conclude the challenges faced by AI in real-world applications, including data curating, model interpretability, and practice regulations. Besides, we expect that clinical implementation of AI will provide important guidance for the patient-tailored management.

Artificial intelligence and allied subsets in early detection and preclusion of gynecological cancers

Gynecological cancers including breast, cervical, ovarian, uterine, and vaginal, pose the greatest threat to world health, with early identification being crucial to patient outcomes and survival rates. The application of machine learning (ML) and artificial intelligence (AI) approaches to the study of gynecological cancer has shown potential to revolutionize cancer detection and diagnosis. The current review outlines the significant advancements, obstacles, and prospects brought about by AI and ML technologies in the timely identification and accurate diagnosis of different types of gynecological cancers. The AI-powered technologies can use genomic data to discover genetic alterations and biomarkers linked to a particular form of gynecologic cancer, assisting in the creation of targeted treatments. Furthermore, it has been shown that the potential benefits of AI and ML technologies in gynecologic tumors can greatly increase the accuracy and efficacy of cancer diagnosis, reduce diagnostic delays, and possibly eliminate the need for needless invasive operations. In conclusion, the review focused on the integrative part of AI and ML based tools and techniques in the early detection and exclusion of various cancer types; together with a collaborative coordination between research clinicians, data scientists, and regulatory authorities, which is suggested to realize the full potential of AI and ML in gynecologic cancer care.

A Comprehensive Model Based on Dynamic Contrast-Enhanced Magnetic Resonance Imaging Can Better Predict the Preoperative Histological Grade of Breast Cancer Than a Radiomics Model

Background

Histological grade is an important prognostic factor for patients with breast cancer and can affect clinical decision-making. From a clinical perspective, developing an efficient and non-invasive method for evaluating histological grading is desirable, facilitating improved clinical decision-making by physicians. This study aimed to develop an integrated model based on radiomics and clinical imaging features for preoperative prediction of histological grade invasive breast cancer.

Methods

In this retrospective study, we recruited 211 patients with invasive breast cancer and randomly assigned them to either a training group (n=147) or a validation group (n=64) with a 7:3 ratio. Patients were classified as having low-grade tumors, which included grade I and II tumors, or high-grade tumors, which included grade III tumors. Three models were constructed based on basic clinical features, radiomics features, and the sum of the two. To assess diagnostic performance of the radiomics models, we employed measures such as receiver operating characteristic (ROC) curve, decision curve analysis (DCA), accuracy, sensitivity, and specificity, and the predictive performance of the three models was compared using the DeLong test and net reclassification improvement (NRI).

Results

The area under the curve (AUC) of the clinical model, radiomics model, and comprehensive model was 0.682, 0.833, and 0.882 in the training set and 0.741, 0.751, and 0.836 in the validation set, respectively. NRI analysis confirmed that the combined model was better than the other two models in predicting the histological grade of breast cancer (NRI=21.4% in the testing cohort).

Conclusion

Compared with the other models, the comprehensive model based on the combination of basic clinical features and radiomics features exhibits more significant potential for predicting histological grade and can better assist clinicians in optimal decision-making.

Cancer care at the time of the fourth industrial revolution: an insight to healthcare professionals' perspectives on cancer care and artificial intelligence

BACKGROUND

The integration of Artificial Intelligence (AI) technology in cancer care has gained unprecedented global attention over the past few decades. This has impacted the way that cancer care is practiced and delivered across settings. The purpose of this study was to explore the perspectives and experiences of healthcare professionals (HCPs) on cancer treatment and the need for AI. This study is a part of the INCISIVE European Union H2020 project's development of user requirements, which aims to fully explore the potential of AI-based cancer imaging technologies.

METHODS

A mixed-methods research design was employed. HCPs participating in cancer care in the UK, Greece, Italy, Spain, Cyprus, and Serbia were first surveyed anonymously online. Twenty-seven HCPs then participated in semi-structured interviews. Appropriate statistical method was adopted to report the survey results by using SPSS. The interviews were audio recorded, verbatim transcribed, and then thematically analysed supported by NVIVO.

RESULTS

The survey drew responses from 95 HCPs. The occurrence of diagnostic delay was reported by 56% (n = 28/50) for breast cancer, 64% (n = 27/42) for lung cancer, 76% (n = 34/45) for colorectal cancer and 42% (n = 16/38) for prostate cancer. A proportion of participants reported the occurrence of false positives in the accuracy of the current imaging techniques used: 64% (n = 32/50) reported this for breast cancer, 60% (n = 25/42) for lung cancer, 51% (n = 23/45) for colorectal cancer and 45% (n = 17/38) for prostate cancer. All participants agreed that the use of technology would enhance the care pathway for cancer patients. Despite the positive perspectives toward AI, certain limitations were also recorded. The majority (73%) of respondents (n = 69/95) reported they had never utilised technology in the care pathway which necessitates the need for education and training in the qualitative finding; compared to 27% (n = 26/95) who had and were still using it. Most, 89% of respondents (n = 85/95) said they would be opened to providing AI-based services in the future to improve medical imaging for cancer care. Interviews with HCPs revealed lack of widespread preparedness for AI in oncology, several barriers to introducing AI, and a need for education and training. Provision of AI training, increasing public awareness of AI, using evidence-based technology, and developing AI based interventions that will not replace HCPs were some of the recommendations.

CONCLUSION

HCPs reported favourable opinions of AI-based cancer imaging technologies and noted a number of care pathway concerns where AI can be useful. For the future design and execution of the INCISIVE project and other comparable AI-based projects, the characteristics and recommendations offered in the current research can serve as a reference.

Radiogenomic-based multiomic analysis reveals imaging intratumor heterogeneity phenotypes and therapeutic targets

Intratumor heterogeneity (ITH) profoundly affects therapeutic responses and clinical outcomes. However, the widespread methods for assessing ITH based on genomic sequencing or pathological slides, which rely on limited tissue samples, may lead to inaccuracies due to potential sampling biases. Using a newly established multicenter breast cancer radio-multiomic dataset (n = 1474) encompassing radiomic features extracted from dynamic contrast-enhanced magnetic resonance images, we formulated a noninvasive radiomics methodology to effectively investigate ITH. Imaging ITH (IITH) was associated with genomic and pathological ITH, predicting poor prognosis independently in breast cancer. Through multiomic analysis, we identified activated oncogenic pathways and metabolic dysregulation in high-IITH tumors. Integrated metabolomic and transcriptomic analyses highlighted ferroptosis as a vulnerability and potential therapeutic target of high-IITH tumors. Collectively, this work emphasizes the superiority of radiomics in capturing ITH. Furthermore, we provide insights into the biological basis of IITH and propose therapeutic targets for breast cancers with elevated IITH.

Contrast-enhanced magnetic resonance image segmentation based on improved U-Net and Inception-ResNet in the diagnosis of spinal metastases

Objective

The objective of this study was to investigate the use of contrast-enhanced magnetic resonance imaging (CE-MRI) combined with radiomics and deep learning technology for the identification of spinal metastases and primary malignant spinal bone tumor.

Methods

The region growing algorithm was utilized to segment the lesions, and two parameters were defined based on the region of interest (ROI). Deep learning algorithms were employed: improved U-Net, which utilized CE-MRI parameter maps as input, and used 10 layers of CE images as input. Inception-ResNet model was used to extract relevant features for disease identification and construct a diagnosis classifier.

Results

The diagnostic accuracy of radiomics was 0.74, while the average diagnostic accuracy of improved U-Net was 0.98, respectively. the PA of our model is as high as 98.001%. The findings indicate that CE-MRI based radiomics and deep learning have the potential to assist in the differential diagnosis of spinal metastases and primary malignant spinal bone tumor.

Conclusion

CE-MRI combined with radiomics and deep learning technology can potentially assist in the differential diagnosis of spinal metastases and primary malignant spinal bone tumor, providing a promising approach for clinical diagnosis.

Feasibility analysis of magnetic resonance imaging-based radiomics features for preoperative prediction of nuclear grading of ductal carcinoma in situ

Background

The nuclear grading of ductal carcinoma in situ (DCIS) affects its clinical risk. The aim of this study was to investigate the possibility of predicting the nuclear grading of DCIS, by magnetic resonance imaging (MRI)-based radiomics features. And to develop a nomogram combining radiomics features and MRI semantic features to explore the potential role of MRI radiomic features in the assessment of DCIS nuclear grading.

Methods

A total of 156 patients (159 lesions) with DCIS and DCIS with microinvasive (DCIS-MI) were enrolled in this retrospective study, with 112 lesions included in the training cohort and 47 lesions included in the validation cohort. Radiomics features were extracted from Dynamic contrast-enhanced MRI (DCE-MRI) phases 1st and 5th. After feature selection, radiomics signature was constructed and radiomics score (Rad-score) was calculated. Multivariate analysis was used to identify MRI semantic features that were significantly associated with DCIS nuclear grading and combined with Rad-score to construct a Nomogram. Receiver operating characteristic curves were used to evaluate the predictive performance of Rad-score and Nomogram, and decision curve analysis (DCA) was used to evaluate the clinical utility.

Results

In multivariate analyses of MRI semantic features, larger tumor size and heterogeneous enhancement pattern were significantly associated with high-nuclear grade DCIS (HNG DCIS). In the training cohort, Nomogram had an area under curve (AUC) of 0.879 and Rad-score had an AUC of 0.828. Similarly, in the independent validation cohort, Nomogram had an AUC value of 0.828 and Rad-score had an AUC of 0.772. In both the training and validation cohorts, Nomogram had a significantly higher AUC value than Rad-score (P<0.05). DCA confirmed that Nomogram had a higher net clinical benefit.

Conclusions

MRI-based radiomic features can be used as potential biomarkers for assessing nuclear grading of DCIS. The nomogram constructed by radiomic features combined with semantic features is feasible in discriminating non-HNG and HNG DCIS.

Radiogenomics: a key component of precision cancer medicine

Radiogenomics, focusing on the relationship between genomics and imaging phenotypes, has been widely applied to address tumour heterogeneity and predict immune responsiveness and progression. It is an inevitable consequence of current trends in precision medicine, as radiogenomics costs less than traditional genetic sequencing and provides access to whole-tumour information rather than limited biopsy specimens. By providing voxel-by-voxel genetic information, radiogenomics can allow tailored therapy targeting a complete, heterogeneous tumour or set of tumours. In addition to quantifying lesion characteristics, radiogenomics can also be used to distinguish benign from malignant entities, as well as patient characteristics, to better stratify patients according to disease risk, thereby enabling more precise imaging and screening. Here, we have characterised the radiogenomic application in precision medicine using a multi-omic approach. we outline the main applications of radiogenomics in diagnosis, treatment planning and evaluations in the field of oncology with the aim of developing quantitative and personalised medicine. Finally, we discuss the challenges in the field of radiogenomics and the scope and clinical applicability of these methods.

Adversarial Attack and Defense in Breast Cancer Deep Learning Systems

Deep-learning-assisted medical diagnosis has brought revolutionary innovations to medicine. Breast cancer is a great threat to women's health, and deep-learning-assisted diagnosis of breast cancer pathology images can save manpower and improve diagnostic accuracy. However, researchers have found that deep learning systems based on natural images are vulnerable to attacks that can lead to errors in recognition and classification, raising security concerns about deep systems based on medical images. We used the adversarial attack algorithm FGSM to reveal that breast cancer deep learning systems are vulnerable to attacks and thus misclassify breast cancer pathology images. To address this problem, we built a deep learning system for breast cancer pathology image recognition with better defense performance. Accurate diagnosis of medical images is related to the health status of patients. Therefore, it is very important and meaningful to improve the security and reliability of medical deep learning systems before they are actually deployed.

Weakly Supervised Breast Lesion Detection in Dynamic Contrast-Enhanced MRI

Currently, obtaining accurate medical annotations requires high labor and time effort, which largely limits the development of supervised learning-based tumor detection tasks. In this work, we investigated a weakly supervised learning model for detecting breast lesions in dynamic contrast-enhanced MRI (DCE-MRI) with only image-level labels. Two hundred fifty-four normal and 398 abnormal cases with pathologically confirmed lesions were retrospectively enrolled into the breast dataset, which was divided into the training set (80%), validation set (10%), and testing set (10%) at the patient level. First, the second image series S2 after the injection of a contrast agent was acquired from the 3.0-T, T1-weighted dynamic enhanced MR imaging sequences. Second, a feature pyramid network (FPN) with convolutional block attention module (CBAM) was proposed to extract multi-scale feature maps of the modified classification network VGG16. Then, initial location information was obtained from the heatmaps generated using the layer class activation mapping algorithm (Layer-CAM). Finally, the detection results of breast lesion were refined by the conditional random field (CRF). Accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were utilized for evaluation of image-level classification. Average precision (AP) was estimated for breast lesion localization. Delong's test was used to compare the AUCs of different models for significance. The proposed model was effective with accuracy of 95.2%, sensitivity of 91.6%, specificity of 99.2%, and AUC of 0.986. The AP for breast lesion detection was 84.1% using weakly supervised learning. Weakly supervised learning based on FPN combined with Layer-CAM facilitated automatic detection of breast lesion.

Comparison of Magnetic Resonance Imaging-Based Radiomics Features with Nomogram for Prediction of Prostate Cancer Invasion

Objective

To explore the value of the magnetic resonance imaging (MRI) radiomics model in predicting prostate cancer (PCa) invasion.

Methods

Clinical data of 86 pathologically confirmed PCa patients in our hospital were collected, including 44 cases in the invasive group and 42 cases in the non-invasive group. All patients underwent MRI examinations, and the same parameters were used. The lesion area was manually delineated and the radiomics features were extracted from T2WI. The radiomics signature based on LASSO regression was established. Besides, logistic regression was used to identify independent clinical predictors, and a combined model incorporating the radiomics signature and independent clinical risk factor was constructed. Finally, the receiver operating characteristic curve (ROC) analysis and decision curve analysis (DCA) was performed to compare the prediction efficiency and clinical benefit of each model.

Results

A total of 867 radiomics features were obtained, and six of them were incorporated into the radiomics model. Multivariate logistic regression analysis exhibited the Gleason score as an independent clinical risk factor for PCa invasion. ROC results showed that the performance of the radiomics model was comparable to that of the clinical-radiomics model in predicting PCa invasion, and it was better than that of the single Gleason score. DCA also confirmed the considerable clinical application value of the radiomics and the clinical-radiomics models.

Conclusion

As a simple, non-invasive, and efficient method, the radiomics model has important predictive value for PCa invasion.

Noninvasive evaluation of neutrophil extracellular traps signature predicts clinical outcomes and immunotherapy response in hepatocellular carcinoma

Background

Neutrophil extracellular traps (NETs) have been shown to play a pivotal role in promoting metastasis and immune escape in hepatocellular carcinoma (HCC). Therefore, noninvasive tests to detect the formation of NETs in tumors can have significant implications for the treatment and prognoses of patients. Here, we sought to develop and validate a computed tomography (CT)-based radiomics model to predict the gene expression profiles that regulate the formation of NETs in HCC.

Methods

This study included 1133 HCC patients from five retrospective cohorts. Based on the mRNA expression levels of 69 biomarkers correlated with NET formation, a 6-gene score (NETs score, NETS) was constructed in cohort 1 from TCIA database (n=52) and validated in cohort 2 (n=232) from ICGC database and cohort 3 (n=365) from TCGA database. And then based on the radiomics features of CT images, a radiomics signature (RNETS) was developed in cohort 1 to predict NETS status (high- or low-NETS). We further employed two cohorts from Nanfang Hospital (Guangzhou, China) to evaluate the predictive power of RNETS in predicting prognosis in cohort 4 (n=347) and the responses to PD-1 inhibitor of HCC patients in cohort 5 (n=137).

Results

For NETS, in cohort 1, the area under the curve (AUC) values predicting 1, 2, and 3-year overall survival (OS) were 0.836, 0.879, and 0.902, respectively. The low-NETS was associated with better survival and higher levels of immune cell infiltration. The RNETS yielded an AUC value of 0.853 in distinguishing between high-NETS or low-NETS and patients with low-RNETS were associated with significantly longer survival time in cohort 1 (P<0.001). Notably, the RNETS was competent in predicting disease-free survival (DFS) and OS in cohort 4 (P<0.001). In cohort 5, the RNETS was found to be an independent risk factor for progression-free survival (PFS) (P<0.001). In addition, the objective response rate of HCC patients treated with PD-1 inhibitor was significantly higher in the low-RNETS group (27.8%) than in the high-RNETS group (10.8%).

Conclusions

This study revealed that RNETS as a radiomics biomarker could effectively predict prognosis and immunotherapy response in HCC patients.

Machine learning with multimodal data for COVID-19

In response to the unprecedented global healthcare crisis of the COVID-19 pandemic, the scientific community has joined forces to tackle the challenges and prepare for future pandemics. Multiple modalities of data have been investigated to understand the nature of COVID-19. In this paper, MIDRC investigators present an overview of the state-of-the-art development of multimodal machine learning for COVID-19 and model assessment considerations for future studies. We begin with a discussion of the lessons learned from radiogenomic studies for cancer diagnosis. We then summarize the multi-modality COVID-19 data investigated in the literature including symptoms and other clinical data, laboratory tests, imaging, pathology, physiology, and other omics data. Publicly available multimodal COVID-19 data provided by MIDRC and other sources are summarized. After an overview of machine learning developments using multimodal data for COVID-19, we present our perspectives on the future development of multimodal machine learning models for COVID-19.

A machine learning approach for prediction of auditory brain stem response in patients after head-and-neck radiation therapy

Objective

The present study aimed to assess machine learning (ML) models according to radiomic features to predict ototoxicity using auditory brain stem responses (ABRs) in patients with radiation therapy (RT) for head-and-neck cancers.

Materials and Methods

The ABR test was performed on 50 patients having head-and-neck RT. Radiomic features were extracted from the brain stem in computed tomography images to generate a radiomic signature. Moreover, accuracy, sensitivity, specificity, the area under the curve, and mean cross-validation were used to evaluate six different ML models.

Results

Out of 50 patients, 21 participants experienced ototoxicity. Furthermore, 140 radiomic features were extracted from the segmented area. Among the six ML models, the Random Forest method with 77% accuracy provided the best result.

Conclusion

According to the ML approach, we showed the relatively high prediction power of the radiomic features in radiation-induced ototoxicity. To better predict the outcomes, future studies on a larger number of participants are recommended.

MRI-based breast cancer radiogenomics using RNA profiling: association with subtypes in a single-center prospective study

BACKGROUND

There are few prospective studies on the correlations between MRI features and whole RNA-sequencing data in breast cancer according to molecular subtypes. The purpose of our study was to explore the association between genetic profiles and MRI phenotypes of breast cancer and to identify imaging markers that influences the prognosis and treatment according to subtypes.

METHODS

From June 2017 to August 2018, MRIs of 95 women with invasive breast cancer were prospectively analyzed, using the breast imaging-reporting and data system and texture analysis. Whole RNA obtained from surgical specimens was analyzed using next-generation sequencing. The association between MRI features and gene expression profiles was analyzed in the entire tumor and subtypes. Gene networks, enriched functions, and canonical pathways were analyzed using Ingenuity Pathway Analysis. The P value for differential expression was obtained using a parametric F test comparing nested linear models and adjusted for multiple testing by reporting Q value.

RESULTS

In 95 participants (mean age, 53 years ± 11 [standard deviation]), mass lesion type was associated with upregulation of CCL3L1 (sevenfold) and irregular mass shape was associated with downregulation of MIR421 (sixfold). In estrogen receptor-positive cancer with mass lesion type, CCL3L1 (21-fold), SNHG12 (11-fold), and MIR206 (sevenfold) were upregulated, and MIR597 (265-fold), MIR126 (12-fold), and SOX17 (fivefold) were downregulated. In triple-negative breast cancer with increased standard deviation of texture analysis on precontrast T1-weighted imaging, CLEC3A (23-fold), SRGN (13-fold), HSPG2 (sevenfold), KMT2D (fivefold), and VMP1 (fivefold) were upregulated, and IGLC2 (73-fold) and PRDX4 (sevenfold) were downregulated (all, P < 0.05 and Q < 0.1). Gene network and functional analysis showed that mass type estrogen receptor-positive cancers were associated with cell growth, anti-estrogen resistance, and poor survival.

CONCLUSION

MRI characteristics are associated with the different expressions of genes related to metastasis, anti-drug resistance, and prognosis, depending on the molecular subtypes of breast cancer.

MRI-Based Radiomics Approach Predicts Tumor Recurrence in ER + /HER2 - Early Breast Cancer Patients

Oncotype Dx Recurrence Score (RS) has been validated in patients with ER + /HER2 - invasive breast carcinoma to estimate patient risk of recurrence and guide the use of adjuvant chemotherapy. We investigated the role of MRI-based radiomics features extracted from the tumor and the peritumoral tissues to predict the risk of tumor recurrence. A total of 62 patients with biopsy-proved ER + /HER2 - breast cancer who underwent pre-treatment MRI and Oncotype Dx were included. An RS > 25 was considered discriminant between low-intermediate and high risk of tumor recurrence. Two readers segmented each tumor. Radiomics features were extracted from the tumor and the peritumoral tissues. Partial least square (PLS) regression was used as the multivariate machine learning algorithm. PLS β-weights of radiomics features included the 5% features with the largest β-weights in magnitude (top 5%). Leave-one-out nested cross-validation (nCV) was used to achieve hyperparameter optimization and evaluate the generalizable performance of the procedure. The diagnostic performance of the radiomics model was assessed through receiver operating characteristic (ROC) analysis. A null hypothesis probability threshold of 5% was chosen (p < 0.05). The exploratory analysis for the complete dataset revealed an average absolute correlation among features of 0.51. The nCV framework delivered an AUC of 0.76 (p = 1.1∙10-3). When combining "early" and "peak" DCE images of only T or TST, a tendency toward statistical significance was obtained for TST with an AUC of 0.61 (p = 0.05). The 47 features included in the top 5% were balanced between T and TST (23 and 24, respectively). Moreover, 33/47 (70%) were texture-related, and 25/47 (53%) were derived from high-resolution images (1 mm). A radiomics-based machine learning approach shows the potential to accurately predict the recurrence risk in early ER + /HER2 - breast cancer patients.

Radiomics model to classify mammary masses using breast DCE-MRI compared to the BI-RADS classification performance

BACKGROUND

Recent advanced in radiomics analysis could help to identify breast cancer among benign mammary masses. The aim was to create a radiomics signature using breast DCE-MRI extracted features to classify tumors and to compare the performances with the BI-RADS classification.

MATERIAL AND METHODS

From September 2017 to December 2019 images, exams and records from consecutive patients with mammary masses on breast DCE-MRI and available histology from one center were retrospectively reviewed (79 patients, 97 masses). Exclusion criterion was malignant uncertainty. The tumors were split in a train-set (70%) and a test-set (30%). From 14 kinetics maps, 89 radiomics features were extracted, for a total of 1246 features per tumor. Feature selection was made using Boruta algorithm, to train a random forest algorithm on the train-set. BI-RADS classification was recorded from two radiologists.

RESULTS

Seventy-seven patients were analyzed with 94 tumors, (71 malignant, 23 benign). Over 1246 features, 17 were selected from eight kinetic maps. On the test-set, the model reaches an AUC = 0.94 95 CI [0.85-1.00] and a specificity of 33% 95 CI [10-70]. There were 43/94 (46%) lesions BI-RADS4 (4a = 12/94 (13%); 4b = 9/94 (10%); and 4c = 22/94 (23%)). The BI-RADS score reached an AUC = 0.84 95 CI [0.73-0.95] and a specificity of 17% 95 CI [3-56]. There was no significant difference between the ROC curves for the model or the BI-RADS score (p = 0.19).

CONCLUSION

A radiomics signature from features extracted using breast DCE-MRI can reach an AUC of 0.94 on a test-set and could provide as good results as BI-RADS to classify mammary masses.

MRI Radiomics and Machine Learning for the Prediction of Oncotype Dx Recurrence Score in Invasive Breast Cancer

AIM

To non-invasively predict Oncotype DX recurrence scores (ODXRS) in patients with ER+ HER2- invasive breast cancer (IBC) using dynamic contrast-enhanced (DCE) MRI-derived radiomics features extracted from primary tumor lesions and a ML algorithm.

MATERIALS AND METHODS

Pre-operative DCE-MRI of patients with IBC, no history of neoadjuvant therapy prior to MRI, and for which the ODXRS was available, were retrospectively selected from a public dataset. ODXRS was obtained on histological tumor samples and considered as positive if greater than 16 and 26 in patients aged under and over 50 years, respectively. Tumor lesions were manually annotated by three independent operators on DCE-MRI images through 3D ROIs positioning. Radiomic features were therefore extracted and selected using multistep feature selection process. A logistic regression ML classifier was then employed for the prediction of ODXRS.

RESULTS

248 patients were included, of which 87 with positive ODXRS. 166 (66%) patients were grouped in the training set, while 82 (33%) in the test set. A total of 1288 features was extracted. Of these, 1244 were excluded as 771, 82 and 391 were excluded as not stable (n = 771), not variant (n = 82), and highly intercorrelated (n = 391), respectively. After the use of recursive feature elimination with logistic regression estimator and polynomial transformation, 92 features were finally selected. In the training set, the logistic regression classifier obtained an overall mean accuracy of 60%. In the test set, the accuracy of the ML classifier was 63%, with a sensitivity of 80%, specificity of 43%, and AUC of 66%.

CONCLUSIONS

Radiomics and ML applied to pre-operative DCE-MRI in patients with IBC showed promises for the non-invasive prediction of ODXRS, aiding in selecting patients who will benefit from NAC.

Predicting Malignancy of Breast Imaging Findings Using Quantitative Analysis of Contrast-Enhanced Mammography (CEM)

We sought to develop new quantitative approaches to characterize the spatial distribution of mammographic density and contrast enhancement of suspicious contrast-enhanced mammography (CEM) findings to improve malignant vs. benign classifications of breast lesions. We retrospectively analyzed all breast lesions that underwent CEM imaging and tissue sampling at our institution from 2014–2020 in this IRB-approved study. A penalized linear discriminant analysis was used to classify lesions based on the averaged histograms of radial distributions of mammographic density and contrast enhancement. T-tests were used to compare the classification accuracies of density, contrast, and concatenated density and contrast histograms. Logistic regression and AUC-ROC analyses were used to assess if adding demographic and clinical data improved the model accuracy. A total of 159 suspicious findings were evaluated. Density histograms were more accurate in classifying lesions as malignant or benign than a random classifier (62.37% vs. 48%; p < 0.001), but the concatenated density and contrast histograms demonstrated a higher accuracy (71.25%; p < 0.001) than the density histograms alone. Including the demographic and clinical data in our models led to a higher AUC-ROC than concatenated density and contrast images (0.81 vs. 0.70; p < 0.001). In the classification of invasive vs. non-invasive malignancy, the concatenated density and contrast histograms demonstrated no significant improvement in accuracy over the density histograms alone (77.63% vs. 78.59%; p = 0.504). Our findings suggest that quantitative differences in the radial distribution of mammographic density could be used to discriminate malignant from benign breast findings; however, classification accuracy was significantly improved with the addition of contrast-enhanced imaging data from CEM. Adding patient demographic and clinical information further improved the classification accuracy.

Attention-based deep learning for breast lesions classification on contrast enhanced spectral mammography: a multicentre study

BACKGROUND

This study aims to develop an attention-based deep learning model for distinguishing benign from malignant breast lesions on CESM.

METHODS

Preoperative CESM images of 1239 patients, which were definitely diagnosed on pathology in a multicentre cohort, were divided into training and validation sets, internal and external test sets. The regions of interest of the breast lesions were outlined manually by a senior radiologist. We adopted three conventional convolutional neural networks (CNNs), namely, DenseNet 121, Xception, and ResNet 50, as the backbone architectures and incorporated the convolutional block attention module (CBAM) into them for classification. The performance of the models was analysed in terms of the receiver operating characteristic (ROC) curve, accuracy, the positive predictive value (PPV), the negative predictive value (NPV), the F1 score, the precision recall curve (PRC), and heat maps. The final models were compared with the diagnostic performance of conventional CNNs, radiomics models, and two radiologists with specialised breast imaging experience.

RESULTS

The best-performing deep learning model, that is, the CBAM-based Xception, achieved an area under the ROC curve (AUC) of 0.970, a sensitivity of 0.848, a specificity of 1.000, and an accuracy of 0.891 on the external test set, which was higher than those of other CNNs, radiomics models, and radiologists. The PRC and the heat maps also indicated the favourable predictive performance of the attention-based CNN model. The diagnostic performance of two radiologists improved with deep learning assistance.

CONCLUSIONS

Using an attention-based deep learning model based on CESM images can help to distinguishing benign from malignant breast lesions, and the diagnostic performance of radiologists improved with deep learning assistance.

PET/CT May Assist in Avoiding Pointless Thyroidectomy in Indeterminate Thyroid Nodules: A Narrative Review

Indeterminate thyroid nodules (ITN) are commonly encountered among the general population, with a malignancy rate of 10 to 40%. However, many patients may be overtreated with futile surgery for benign ITN. To avoid unnecessary surgery, PET/CT scan is a possible alternative to help differentiate between benign and malignant ITN. In this narrative review, the major results and limitations of the most recent studies on PET/CT efficacy (from PET/CT visual assessment to quantitative PET parameters and recent radiomic features analysis) and on cost-effectiveness (compared to other alternatives (such as surgery)) are presented. PET/CT can reduce futile surgery with visual assessment (around 40%; if ITN ≥ 10 mm). Moreover, PET/CT conventional parameters and radiomic features extracted from PET/CT imaging can be associated together in a predictive model to rule out malignancy in ITN, with a high NPV (96%) when certain criteria are met. Even though promising results were obtained in these recent PET/CT studies, further studies are needed to enable PET/CT to become the definitive diagnostic tool once a thyroid nodule is identified as indeterminate.

Surgical Planning after Neoadjuvant Treatment in Breast Cancer: A Multimodality Imaging-Based Approach Focused on MRI

Neoadjuvant chemotherapy (NACT) today represents a cornerstone in the treatment of locally advanced breast cancer and highly chemo-sensitive tumors at early stages, increasing the possibilities of performing more conservative treatments and improving long term outcomes. Imaging has a fundamental role in the staging and prediction of the response to NACT, thus aiding surgical planning and avoiding overtreatment. In this review, we first examine and compare the role of conventional and advanced imaging techniques in preoperative T Staging after NACT and in the evaluation of lymph node involvement. In the second part, we analyze the different surgical approaches, discussing the role of axillary surgery, as well as the possibility of non-operative management after-NACT, which has been the subject of recent trials. Finally, we focus on emerging techniques that will change the diagnostic assessment of breast cancer in the near future.

Predicting gene mutation status via artificial intelligence technologies based on multimodal integration (MMI) to advance precision oncology

Personalized treatment strategies for cancer frequently rely on the detection of genetic alterations which are determined by molecular biology assays. Historically, these processes typically required single-gene sequencing, next-generation sequencing, or visual inspection of histopathology slides by experienced pathologists in a clinical context. In the past decade, advances in artificial intelligence (AI) technologies have demonstrated remarkable potential in assisting physicians with accurate diagnosis of oncology image-recognition tasks. Meanwhile, AI techniques make it possible to integrate multimodal data such as radiology, histology, and genomics, providing critical guidance for the stratification of patients in the context of precision therapy. Given that the mutation detection is unaffordable and time-consuming for a considerable number of patients, predicting gene mutations based on routine clinical radiological scans or whole-slide images of tissue with AI-based methods has become a hot issue in actual clinical practice. In this review, we synthesized the general framework of multimodal integration (MMI) for molecular intelligent diagnostics beyond standard techniques. Then we summarized the emerging applications of AI in the prediction of mutational and molecular profiles of common cancers (lung, brain, breast, and other tumor types) pertaining to radiology and histology imaging. Furthermore, we concluded that there truly exist multiple challenges of AI techniques in the way of its real-world application in the medical field, including data curation, feature fusion, model interpretability, and practice regulations. Despite these challenges, we still prospect the clinical implementation of AI as a highly potential decision-support tool to aid oncologists in future cancer treatment management.

Synergies of Radiomics and Transcriptomics in Lung Cancer Diagnosis: A Pilot Study

Radiotranscriptomics is an emerging field that aims to investigate the relationships between the radiomic features extracted from medical images and gene expression profiles that contribute in the diagnosis, treatment planning, and prognosis of cancer. This study proposes a methodological framework for the investigation of these associations with application on non-small-cell lung cancer (NSCLC). Six publicly available NSCLC datasets with transcriptomics data were used to derive and validate a transcriptomic signature for its ability to differentiate between cancer and non-malignant lung tissue. A publicly available dataset of 24 NSCLC-diagnosed patients, with both transcriptomic and imaging data, was used for the joint radiotranscriptomic analysis. For each patient, 749 Computed Tomography (CT) radiomic features were extracted and the corresponding transcriptomics data were provided through DNA microarrays. The radiomic features were clustered using the iterative K-means algorithm resulting in 77 homogeneous clusters, represented by meta-radiomic features. The most significant differentially expressed genes (DEGs) were selected by performing Significance Analysis of Microarrays (SAM) and 2-fold change. The interactions among the CT imaging features and the selected DEGs were investigated using SAM and a Spearman rank correlation test with a False Discovery Rate (FDR) of 5%, leading to the extraction of 73 DEGs significantly correlated with radiomic features. These genes were used to produce predictive models of the meta-radiomics features, defined as p-metaomics features, by performing Lasso regression. Of the 77 meta-radiomic features, 51 can be modeled in terms of the transcriptomic signature. These significant radiotranscriptomics relationships form a reliable basis to biologically justify the radiomics features extracted from anatomic imaging modalities. Thus, the biological value of these radiomic features was justified via enrichment analysis on their transcriptomics-based regression models, revealing closely associated biological processes and pathways. Overall, the proposed methodological framework provides joint radiotranscriptomics markers and models to support the connection and complementarities between the transcriptome and the phenotype in cancer, as demonstrated in the case of NSCLC.

How Radiomics Can Improve Breast Cancer Diagnosis and Treatment

Recent technological advances in the field of artificial intelligence hold promise in addressing medical challenges in breast cancer care, such as early diagnosis, cancer subtype determination and molecular profiling, prediction of lymph node metastases, and prognostication of treatment response and probability of recurrence. Radiomics is a quantitative approach to medical imaging, which aims to enhance the existing data available to clinicians by means of advanced mathematical analysis using artificial intelligence. Various published studies from different fields in imaging have highlighted the potential of radiomics to enhance clinical decision making. In this review, we describe the evolution of AI in breast imaging and its frontiers, focusing on handcrafted and deep learning radiomics. We present a typical workflow of a radiomics analysis and a practical "how-to" guide. Finally, we summarize the methodology and implementation of radiomics in breast cancer, based on the most recent scientific literature to help researchers and clinicians gain fundamental knowledge of this emerging technology. Alongside this, we discuss the current limitations of radiomics and challenges of integration into clinical practice with conceptual consistency, data curation, technical reproducibility, adequate accuracy, and clinical translation. The incorporation of radiomics with clinical, histopathological, and genomic information will enable physicians to move forward to a higher level of personalized management of patients with breast cancer.

Criteria for the translation of radiomics into clinically useful tests

Computer-extracted tumour characteristics have been incorporated into medical imaging computer-aided diagnosis (CAD) algorithms for decades. With the advent of radiomics, an extension of CAD involving high-throughput computer-extracted quantitative characterization of healthy or pathological structures and processes as captured by medical imaging, interest in such computer-extracted measurements has increased substantially. However, despite the thousands of radiomic studies, the number of settings in which radiomics has been successfully translated into a clinically useful tool or has obtained FDA clearance is comparatively small. This relative dearth might be attributable to factors such as the varying imaging and radiomic feature extraction protocols used from study to study, the numerous potential pitfalls in the analysis of radiomic data, and the lack of studies showing that acting upon a radiomic-based tool leads to a favourable benefit-risk balance for the patient. Several guidelines on specific aspects of radiomic data acquisition and analysis are already available, although a similar roadmap for the overall process of translating radiomics into tools that can be used in clinical care is needed. Herein, we provide 16 criteria for the effective execution of this process in the hopes that they will guide the development of more clinically useful radiomic tests in the future.

Development and validation of an MRI-radiomics nomogram for the prognosis of pancreatic ductal adenocarcinoma

Objective

To develop and validate an MRI-radiomics nomogram for the prognosis of pancreatic ductal adenocarcinoma (PDAC).

Background

"Radiomics" enables the investigation of huge amounts of radiological features in parallel by extracting high-throughput imaging data. MRI provides better tissue contrast with no ionizing radiation for PDAC.

Methods

There were 78 PDAC patients enrolled in this study. In total, there were 386 radiomics features extracted from MRI scan, which were screened by the least absolute shrinkage and selection operator algorithm to develop a risk score. Cox multivariate regression analysis was applied to develop the radiomics-based nomogram. The performance was assessed by discrimination and calibration.

Results

The radiomics-based risk-score was significantly associated with PDAC overall survival (OS) (P < 0.05). With respect to survival prediction, integrating the risk score, clinical data and TNM information into the nomogram exhibited better performance than the TNM staging system, radiomics model and clinical model. In addition, the nomogram showed fine discrimination and calibration.

Conclusions

The radiomics nomogram incorporating the radiomics data, clinical data and TNM information exhibited precise survival prediction for PDAC, which may help accelerate personalized precision treatment.

Clinical trial registration

clinicaltrials.gov, identifier NCT05313854.

Shear-wave elastography-based nomograms predicting 21-gene recurrence score for adjuvant chemotherapy decisions in patients with breast cancer

PURPOSE

To develop and validate nomograms based on shear-wave elastography (SWE) combined with clinicopathologic features for predicting Oncotype DX recurrence score (RS) for use with adjuvant systemic therapy guidelines.

METHODS

In a retrospective study, patients with breast cancer who underwent definitive surgery of the breast between August 2011 and December 2019 were eligible for this study. Those with surgery between August 2011 and March 2019 were assigned to a development set and the rest were assigned to an independent validation set. Clinicopathologic features and SWE elasticity indices were assessed with logistic regression to develop nomograms for predicting RS ≥ 16 and ≥ 26. Analysis of the area under the receiver operating characteristic curve (AUROC) was used to assess the performance of the nomograms.

RESULTS

Of a total 381 women (mean age, 51 ± 9 years), 286 (mean age, 51 ± 9 years) were in the development set and 95 (mean age, 51 ± 9 years) in the validation set. All SWE elasticity indices were independently associated with each RS cutoff (odds ratio, 1.006-1.039 for RS ≥ 16; odds ratio, 1.008-1.076 for RS ≥ 26). Nomograms based on SWE combined with clinicopathologic features were developed and validated for RS ≥ 16 (mean elasticity [AUROC, 0.74; 95% CI: 0.68, 0.80] and maximum elasticity [AUROC, 0.74; 95% CI: 0.69, 0.80]) and for RS ≥ 26 (mean elasticity [AUROC, 0.81; 95% CI: 0.73, 0.89], maximum elasticity [AUROC, 0.82; 95% CI: 0.74, 0.89], and elasticity ratio [AUROC, 0.86; 95% CI: 0.80, 0.93]).

CONCLUSION

Nomograms based on SWE can predict Oncotype DX RS for use in adjuvant systemic therapy decisions.

Validation study of a nomogram for predicting probability of low risk of MammaPrint results in women with clinically high-risk breast cancer

BACKGROUND

MammaPrint (MMP) helps clinicians identify the ideal time for adjuvant treatment for patients with early HR+/HER2- breast cancer. We aimed to validate a nomogram designed to predict probability of low risk of MMP results and to evaluate the difference in survival outcome between two groups divided by nomogram score.

METHODS

In this retrospective cohort study, we evaluated 172 patients from Asan Medical Center, Seoul, Korea, who underwent breast cancer surgery and MMP during 2020-2021. First, we validated the nomogram by calculating the area under the curve (AUC) and using calibration. Additionally, with the data of 1,835 T1-3N0-1M0 HR+/HER2- patients from Asan Medical Center during 2010-2013, we compared the disease-free survival (DFS), overall survival (OS), and breast cancer-specific survival (BCSS) rates by Kaplan-Meier analysis between the two groups divided by nomogram total point (TP) of 183.

RESULTS

The AUC calculated by validation of 172 patients was 0.73 (95% confidence interval [CI], 0.66-0.81). The discrimination and calibration of the prediction model were satisfactory following additional validation of 1835 patients. The high-risk and low-risk groups had different 5-year OS (97.9% vs. 98.1%, p = 0.056), DFS (98.6% vs. 99.4%, p = 0.008), and BCSS rates (98.6% vs. 99.4%, p = 0.002).

CONCLUSION

For treatment decision-making among clinically high-risk patients with HR+/HER2- and node-positive disease, the nomogram showed satisfactory performance in predicting patients with low genomic risk. Survival outcome significantly differed between two groups divided by nomogram TP. More studies are needed to validate this model in international cohorts and large prospective cohorts from other institutions.

Precision Medicine in Breast Cancer: Do MRI Biomarkers Identify Patients Who Truly Benefit from the Oncotype DX Recurrence Score® Test?

The aim of this study was to combine breast MRI-derived biomarkers with clinical-pathological parameters to identify patients who truly need an Oncotype DX Breast Recurrence Score® (ODXRS) genomic assay, currently used to predict the benefit of adjuvant chemotherapy in ER-positive/HER2-negative early breast cancer, with the ultimate goal of customizing therapeutic decisions while reducing healthcare costs. Patients who underwent a preoperative multiparametric MRI of the breast and ODXRS tumor profiling were retrospectively included in this study. Imaging sets were evaluated independently by two breast radiologists and classified according to the 2013 American College of Radiology Breast Imaging Reporting and Data System (ACR BI-RADS) lexicon. In a second step of the study, a combined oncologic and radiologic assessment based on clinical-pathological and radiological data was performed, in order to identify patients who may need adjuvant chemotherapy. Results were correlated with risk levels expressed by ODXRS, using the decision made on the basis of the ODXRS test as a gold standard. The χ2 test was used to evaluate associations between categorical variables, and significant ones were further investigated using logistic regression analyses. A total of 58 luminal-like, early-stage breast cancers were included. A positive correlation was found between ODXRS and tumor size (p = 0.003), staging (p = 0.001) and grading (p = 0.005), and between BI-RADS categories and ODXRS (p < 0.05 for both readers), the latter being confirmed at multivariate regression analysis. Moreover, BI-RADS categories proved to be positive predictors of the therapeutic decision taken after performing an ODXRS assay. A statistically significant association was also found between the therapeutic decision based on the ODXRS and the results of combined onco-radiologic assessment (p < 0.001). Our study suggests that there is a correlation between BI-RADS categories at MRI and ODXRS and that a combined onco-radiological assessment may predict the decision made on the basis of the results of ODXRS genomic test.

Predicting hormone receptors and PAM50 subtypes of breast cancer from multi-scale lesion images of DCE-MRI with transfer learning technique

BACKGROUND

The recent development of artificial intelligence (AI) technologies coupled with medical imaging data has gained considerable attention, and offers a non-invasive approach for cancer diagnosis and prognosis. In this context, improved breast cancer (BC) molecular characteristics assessment models are foreseen to enable personalized strategies with better clinical outcomes compared to existing screening strategies. And it is a promising approach to developing models for hormone receptors (HR) and subtypes of BC patients from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data.

METHODS

In this institutional review board-approved study, 174 BC patients with both DCE-MRI and RNA-seq data in the local database were analyzed. Slice images from tumor lesions and multi-scale peri-tumor regions were used as model inputs, and five representative pre-trained transfer learning (TF) networks, such as Inception-v3 and Xception, were employed to establish prediction models. A comprehensive analysis was performed using five-fold cross-validation to avoid overfitting, and accuracy (ACC) and area under the receiver operating characteristic curve (AUROC) to evaluate model performance.

RESULTS

Xception achieved the superior results when using solely tumor regions, with highest AUROCs of 0.844 (95% CI: [0.841, 0.847]) and 0.784 (95% CI: [0.781, 0.788]) for estrogen receptor (ER) and progesterone receptor (PR), respectively, and best ACC of 0.467 (95% CI: [0.462, 0.470]) for PAM50 subtypes. A significant improvement in the model performance was observed when images of the peri-tumor region were included, with optimal results achieved using images of the tumor and the 10 mm peri-tumor regions. Xception-based TF models performed most effectively in predicting ER and PR statuses, with the AUROCs were 0.942 (95% CI: [0.940, 0.944]) and 0.920 (95% CI: [0.917, 0.922]), respectively, whereas for PAM50 subtypes, the Inception-v3-based network yielded the highest ACC as 0.742 (95% CI: [0.738, 0.746]).

CONCLUSIONS

Transfer learning analysis based on DCE-MRI data of tumor and peri-tumor regions was helpful to the non-invasive assessment of molecular characteristics of BC.