Differentiating between benign and malignant breast lesions using dual-energy CT-based model: development and validation

OBJECTIVES

To develop and validate a dual-energy CT (DECT)-based model for noninvasively differentiating between benign and malignant breast lesions detected on DECT.

MATERIALS AND METHODS

This study prospectively enrolled patients with suspected breast cancer who underwent dual-phase contrast-enhanced DECT from July 2022 to July 2023. Breast lesions were randomly divided into the training and test cohorts at a ratio of 7:3. Clinical characteristics, DECT-based morphological features, and DECT quantitative parameters were collected. Univariate analyses and multivariate logistic regression were performed to determine independent predictors of benign and malignant breast lesions. An individualized model was constructed. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic ability of the model, whose calibration and clinical usefulness were assessed by calibration curve and decision curve analysis.

RESULTS

This study included 200 patients (mean age, 49.9 ± 11.9 years; age range, 22-83 years) with 222 breast lesions. Age, lesion shape, and the effective atomic number (Zeff) in the venous phase were significant independent predictors of breast lesions (all p < 0.05). The discriminative power of the model incorporating these three factors was high, with AUCs of 0.844 (95%CI 0.764-0.925) and 0.791 (95% CI 0.647-0.935) in the training and test cohorts, respectively. The constructed model showed a preferable fitting (all p > 0.05 by the Hosmer-Lemeshow test) and provided enhanced net benefits than simple default strategies within a wide range of threshold probabilities in both cohorts.

CONCLUSION

The DECT-based model showed a favorable diagnostic performance for noninvasive differentiation between benign and malignant breast lesions detected on DECT.

CRITICAL RELEVANCE STATEMENT

The combination of clinical and morphological characteristics and DECT-derived parameter have the potential to identify benign and malignant breast lesions and it may be useful for incidental breast lesions on DECT to decide if further work-up is needed.

KEY POINTS

It is important to characterize incidental breast lesions on DECT for patient management. DECT-based model can differentiate benign and malignant breast lesions with good performance. DECT-based model is a potential tool for distinguishing breast lesions detected on DECT.

Diagnostic Accuracy of Contrast-Enhanced Thoracic Photon-Counting Computed Tomography for Opportunistic Locoregional Staging of Breast Cancer Compared With Digital Mammography: A Prospective Trial

OBJECTIVE

Accurate locoregional staging is crucial for effective breast cancer treatment. Photon-counting computed tomography (PC-CT) is an emerging technology with high spatial resolution and the ability to depict uptake of contrast agents in tissues, making it a promising tool for breast cancer imaging. The aim of this study was to establish the feasibility of locoregional staging of breast cancer through contrast-enhanced thoracic PC-CT, assess its diagnostic performance, and compare it with that of digital mammography (DM).

MATERIALS AND METHODS

Patients with newly diagnosed breast cancer, DM, and indication of thoracic CT staging were prospectively enrolled in this clinical cohort study over a period of 6 months. Participants underwent contrast-enhanced thoracic PC-CT and breast magnetic resonance imaging in prone position. After blinding to patient data, 2 radiologists independently rated PC-CT and DM regarding the following 6 characteristics: (1) diameter of the largest mass lesion, (2) infiltration of cutis/pectoral muscle/thoracic wall, (3) number of mass lesions, (4) presence/absence of adjacent ductal carcinoma in situ (DCIS), (5) tumor conspicuity, and (6) diagnostic confidence. Reference standard was generated from consensus reading of magnetic resonance imaging combined with all histopathological/clinical data by an independent adjudication committee applying TNM eighth edition.

RESULTS

Among 32 enrolled female subjects (mean ± SD age, 59 ± 13.0 years), diagnostic accuracy for T-classification was higher for PC-CT compared with DM (0.94 vs 0.50, P < 0.01). Moreover, the correlation of the number of detected tumor masses with the reference standard was stronger for PC-CT than for DM (0.72 vs 0.50, P < 0.01). We observed that PC-CT significantly ( P < 0.04) outperformed DM regarding not only sensitivity (0.83 and 0.25, respectively) but also specificity (0.99 and 0.80, respectively) for adjacent DCIS. The κ values for interreader reliability were higher for PC-CT compared with DM (mean 0.88 vs 0.54, respectively; P = 0.01).

CONCLUSIONS

Photon-counting computed tomography outperformed DM in T-classification and provided higher diagnostic accuracy for the detection of adjacent DCIS. Therefore, opportunistic locoregional staging of breast cancer in contrast-enhanced thoracic PC-CT is feasible and could overcome limitations of DM with the potential to improve patient management.

Assignment of the biological value of solid breast masses based on quantitative evaluations of spectral CT examinations using electron density mapping, Zeffective mapping and iodine mapping

OBJECTIVE

We aimed to asses, in a clinical setting, whether the newly available quantitative evaluation of electron density (ED) in spectral CT examinations of the breast provide information on the biological identity of solid breast masses and whether ED maps yield added value to the diagnostic information of iodine maps and Zeff maps calculated from the same CT image datasets.

METHODS

All patients at the University Breast Cancer Center who underwent a clinically indicated Dual Layer Computed Tomography (DLCT) examination for staging of invasive breast cancer from 2018 to 2020 were prospectively included. Iodine concentration maps, Zeff maps and ED maps were automatically reconstructed from the DLCT datasets. Region of interest (ROI) based evaluations in the breast target lesions and in the aorta were performed semi-automatically in identical anatomical positions using dedicated evaluation software. Case-by-case evaluations were carried independently by 2 of 4 radiologists for each examination, respectively. Statistical analysis derived from the ROIs was done by calculating ROC/AUC curves and Youden indices.

RESULTS

The evaluations comprised 166 DLCT examinations. In the ED maps the measurements in the breast target lesions yielded Youden cutpoints of 104.0% (reader 1) and 103.8% (reader 2) resulting in AUCs of 0.63 and 0.67 at the empirical cutpoints. The variables "Zeff" and "iodine content" derived from the target lesions showed superior diagnostical results, with a Youden cutpoint of 8.0 mg/ml in the iodine maps and cutpoints of 1.1/1.2 in the Zeff maps the AUCs ranging from 0.84 to 0.85 (p = 0.023 to <0.000). The computational combination of Zeff and ED measurements in the target lesions yielded a slight AUC increase (readers 1: 0.85-0.87; readers 2: 0.84-0.94). The ratios of the measured values in the target lesions normalized to the values measured in the aorta showed comparable results. The AUCs of ED derived from the cutpoints showed inferior results to those derived from the Zeff maps and iodine maps (ED: 0.64 and 0.66 for reader 1 and 2; Zeff: 0.86 for both readers; iodine content: 0.89 and 0.86 for reader 1 and 2, respectively). The computational combination of the ED results and the Zeff measurements did not lead to a clinically relevant diagnostic gain with AUCs ranging from 0.86 to 0.88.

CONCLUSIONS

Quantitative assessments of Zeff, iodine content and ED all targeting the physical and chemical aspects of iodine uptake in solid breast masses confirmed diagnostically robust cutpoints for the differentiation of benign and malignant findings (Zeff < 7.7, iodine content of <0.8 mg/ml). The evaluations of the ED did not indicate any added diagnostic value beyond the quantitative assessments of Zeff and iodine content. Further research is warranted to develop suitable clinical indications for the use of ED maps.

Correlation of CT-data derived from multiparametric dual-layer CT-maps with immunohistochemical biomarkers in invasive breast carcinomas

OBJECTIVE

To examine the correlation of quantitative measurements from material decomposition maps calculated from dual-layer CT (DLCT)-image datasets with immunohistochemical biomarkers of invasive breast carcinomas.

MATERIAL AND METHODS

All patients at the University Breast Cancer Center who underwent a clinically indicated dual-layer CT-scan for staging of invasive ductal breast carcinoma from 01/2016 to 07/2020 were prospectively included. Iodine concentration maps and maps of the effective atomic numbers (Zeffective) were reconstructed from the image datasets. ROI-based evaluations of the index tumors and predefined references tissues for normalization were performed semi-automatically in identical anatomical positions using dedicated evaluation software. Statistical analysis was essentially descriptive using Spearmańs rank correlation and (multivariable) partial correlation.

RESULTS

Bivariate showed statistically significant correlations of iodine contents (r = -0.154/-0.202/0.180, p = 0.039/0.006/0.015), and Zeffective-values (r = -0.158/-0.199/0.179, p = 0.034/0.007/0.016) for all 184 carcinomas and the subgroup of 168 invasive ductal carcinomas. The results were confirmed by multivariate analyses with "age", "diameter" and "ACR-grade" as possible confounders. Normalization of the measured target values with those in the aorta confirmed significant correlations of iodine content and Zeffective compared to Estrogen (r = 0.174, p = 0.019), Progesteron (r = 0.168/0.177, p = 0.024/0.017), and HER2 receptor expression (r = -0.222/-0.184, p = 0.003/0.013). All CT-parameters showed significant correlations with immunohistochemical subtyping (r = 0.191/0.192, p = 0.010).

CONCLUSIONS

Our preliminary results indicate that iodine content and Zeffective-values derived from DLCT-examinations correlate with hormone receptor expression in invasive breast carcinomas. Assignments to benign entities already seam feasible in clinical routine CT-diagnostics. After further investigations iodine content and Zeffective may be translated as diagnostical and prognostical biomarkers into clinical routine in the long term.

Possibility to discriminate benign from malignant breast lesions detected on dual-layer spectral CT-evaluation

OBJECTIVES

Intramammary mass lesions are reportedly present in up to 5.8% of all contrast enhanced CT-examinations of the female chest. We aimed to assess whether their biological relevance can be estimated using spectral CT-datasets.

METHODS

In this bicentric retrospective study patients with breast masses visualized on spectral CT-examinations from 07/2017 to 06/2019 were included. Lesions were characterized as malignant or benign based on histology and/or a stable follow-up of >2 years. Conventional CT-images, iodine density-maps, virtual monoenergetic-images (40 keV, 100 keV) and Zeffective-maps were evaluated by two independent readers. Statistical analysis derived from the Regions of interest (ROIs) was done by calculating the Areas under the Receiver operating characteristic (ROC) curve (AUC) and Youden-indices.

RESULTS

106 breast masses (malignant/benign: 81/25, 76.4%/23.6%) were included. The mean AUCs of the variables "iodine content" (reader 1/2:0.97;0.98), "monoenergetic curve-slope" (0.97;0.96) and "Zeffective" (0.98;0.98) measured in the target lesions (TL) showed superior results compared to those derived from the variable "density" (0.92;0.93) (p < 0.001). The ratios "TL to aorta" calculated for the variables "iodine content", "monoenergetic curve-slope" and "Zeffective" showed superior results compared to normal breast tissue and muscle (p < 0.001). The optimal cutpoint for the "iodine content" in the TL was 0.7-0.9 mg/ml (sensitivity 96.6%, specificity 91.7%). The best diagnostic results were achieved by normalizing the iodine content in the TL to that in the aorta (optimal cutpoint 0.1, sensitivity 95.5%, 98.9%, specificity 91.7%).

CONCLUSIONS

Our preliminary results suggest that spectral CT-datasets might allow to estimate the biological dignity of breast masses detected on clinically indicated chest-examinations.

Giant epidermal inclusion cyst with infection arising within the breast parenchyma: a case report

Epidermal inclusion cysts (EICs) of the breast develop in the deep breast parenchyma, and they are very rare. Only about 10 cases have been reported in the English-language literature to date. In this report, we present a rare case of a giant EIC with infection arising within the deep breast parenchyma. Unlike a typical EIC of the breast, the EIC in the present case was a cystic and solid lesion containing a large amount of liquid within the cyst and popcorn-like calcification in the wall. In this report, we describe the contrast-enhanced spectral mammography (CESM), ultrasonography, and computed tomography findings and provide a reference for the diagnosis of EICs. To the best of our knowledge, this is the first report of the CESM findings of an EIC. Our case illustrates that CESM has excellent performance similar to that of magnetic resonance imaging and is much more effective than conventional digital mammography. Additionally, our case indicates that precise correlation of CESM with ultrasonography findings contributes to the diagnosis of EICs. This rare case with multiple imaging findings will increase the awareness of EICs in the breast parenchyma.

Application of Machine Learning Algorithms in Breast Cancer Diagnosis and Classification

Breast cancer continues to be the most frequent cancer in females, affecting about one in 8 women and causing the highest number of cancer-related deaths in females worldwide despite remarkable progress in early diagnosis, screening, and patient management. All breast lesions are not malignant, and all the benign lesions do not progress to cancer. However, the accuracy of diagnosis can be increased by a combination or preoperative tests such as physical examination, mammography, fine-needle aspiration cytology, and core needle biopsy. Despite some limitations, these procedures are more accurate, reliable, and acceptable, when compared with a single adopted diagnostic procedure. Recent studies have shown that breast cancer can be accurately predicted and diagnosed using machine learning (ML) technology. The objective of this study was to explore the application of ML approaches to classify breast cancer based on feature values generated from a digitized image of a fine-needle aspiration (FNA) of a breast mass. To achieve this objective, we used ML algorithms, collected a scientific dataset of 569 breast cancer patients from Kaggle (https://www.kaggle.com/uciml/breast-cancer-wisconsin-data), analyze and interpreted the data based on ten real-valued features of a breast mass FNA including the radius, texture, perimeter, area, smoothness, compactness, concavity, concave points, symmetry, and fractal dimension. Among the 569 patients tested, 63% were diagnosed with benign breast cancer and 37% were diagnosed with malignant breast cancer. Benign tumors grow slowly and do not spread while malignant tumors grow rapidly and spread to other parts of the body.

Applications of Computational Methods in Biomedical Breast Cancer Imaging Diagnostics: A Review

With the exponential increase in new cases coupled with an increased mortality rate, cancer has ranked as the second most prevalent cause of death in the world. Early detection is paramount for suitable diagnosis and effective treatment of different kinds of cancers, but this is limited to the accuracy and sensitivity of available diagnostic imaging methods. Breast cancer is the most widely diagnosed cancer among women across the globe with a high percentage of total cancer deaths requiring an intensive, accurate, and sensitive imaging approach. Indeed, it is treatable when detected at an early stage. Hence, the use of state of the art computational approaches has been proposed as a potential alternative approach for the design and development of novel diagnostic imaging methods for breast cancer. Thus, this review provides a concise overview of past and present conventional diagnostics approaches in breast cancer detection. Further, we gave an account of several computational models (machine learning, deep learning, and robotics), which have been developed and can serve as alternative techniques for breast cancer diagnostics imaging. This review will be helpful to academia, medical practitioners, and others for further study in this area to improve the biomedical breast cancer imaging diagnosis.

Dual-Energy CT for Locoregional Staging of Breast Cancer: Preliminary Results

OBJECTIVE. The objective of this study was to demonstrate the feasibility of dual-energy CT (DECT) for locoregional staging of breast cancer and differentiation of tumor histotypes. MATERIALS AND METHODS. From January 2016 to July 2017, a total of 31 patients (mean [± SD] age, 55.8 ± 14.8 years) with breast cancer diagnosed by needle biopsy who underwent preoperative contrast-enhanced DECT for staging purposes were selected from a retrospective review of institutional databases. Monochromatic images obtained at 40 and 70 keV were evaluated by two readers who determining the number of hypervascularized tumors present and the largest tumor diameter for each breast. The attenuation values and iodine concentration of tumors and normal breast tissue and the ratios of these findings in each tissue type were recorded. Cancers were classified as ductal carcinoma in situ, invasive ductal carcinoma, and invasive lobular carcinoma. The reference standard was the final pathologic finding after surgery. RESULTS. A total of 64 tumor lesions were found at histopathologic analysis versus 67 on DECT for 34 breasts (three bilateral cancers were included). Nonparametric statistics were used. The largest lesion diameter observed DECT was 33.2 ± 20.5 mm versus 31.8 ± 20.5 mm on pathologic analysis, and cancer distribution was correctly classified for 31 of 34 (91%) cases. ROC curves derived from lesion iodine concentration showed that the optimal thresholds for distinguishing infiltrating carcinomas (invasive lobular and ductal carcinomas) and from other lesions were 1.70 mg/mL (sensitivity, 94.9%; specificity, 93.0%; AUC value, 0.968). ROC curves derived from the ratio of the iodine concentration in lesions to that in normal breast parenchyma showed that 6.13 was the optimal threshold to distinguish invasive ductal carcinoma from other lesions (sensitivity, 87.0%; specificity, 81.1%; AUC value, 0.914). CONCLUSION. DECT is feasible and seems to be a reliable tool for locoregional staging of breast cancer.

Extra-abdominal dual-energy CT applications: a comprehensive overview

Unlike conventional computed tomography, dual-energy computed tomography is a relatively novel technique that exploits ionizing radiations at different energy levels. The separate radiation sets can be achieved through different technologies, such as dual source, dual layers or rapid switching voltage. Body tissue molecules vary for their specific atomic numbers and electron density, and the interaction with different sets of radiations results in different attenuations, allowing to their final distinction. In particular, iodine recognition and quantification have led to important information about intravenous contrast medium delivery within the body. Over the years, useful post-processing algorithms have also been validated for improving tissue characterization. For instance, contrast resolution improvement and metal artifact reduction can be obtained through virtual monoenergetic images, dose reduction by virtual non-contrast reconstructions and iodine distribution highlighting through iodine overlay maps. Beyond the evaluation of the abdominal organs, dual-energy computed tomography has also been successfully employed in other anatomical districts. Although lung perfusion is one of the most investigated, this evaluation has been extended to narrowly fields of application, such as musculoskeletal, head and neck, vascular and cardiac. The potential pool of information provided by dual-energy technology is already wide and not completely explored, yet. Therefore, its performance continues to raise increasing interest from both radiologists and clinicians.

Comparison of Lipid and Water Contents by Time-domain Diffuse Optical Spectroscopy and Dual-energy Computed Tomography in Breast Cancer Patients

We previously compared time-domain diffuse optical spectroscopy (TD-DOS) with magnetic resonance imaging (MRI) using various water/lipid phantoms. However, it is difficult to conduct similar comparisons in the breast, because of measurement differences due to modality-dependent differences in posture. Dual-energy computed tomography (DECT) examination is performed in the same supine position as a TD-DOS measurement. Therefore, we first verified the accuracy of the measured fat fraction of fibroglandular tissue in the normal breast on DECT by comparing it with MRI in breast cancer patients (n = 28). Then, we compared lipid and water signals obtained in TD-DOS and DECT from normal and tumor-tissue regions (n = 16). The TD-DOS breast measurements were carried out using reflectance geometry with a source–detector separation of 3 cm. A semicircular region of interest (ROI), with a transverse diameter of 3 cm and a depth of 2 cm that included the breast surface, was set on the DECT image. Although the measurement area differed between the modalities, the correlation coefficients of lipid and water signals between TD-DOS and DECT were rs = 0.58 (p < 0.01) and rs = 0.90 (p < 0.01), respectively. These results indicate that TD-DOS captures the characteristics of the lipid and water contents of the breast.