ABSTRACT
PURPOSE: To assess the density values of breast lesions and breast tissue using non-contrast spiral breast CT (nc-SBCT) imaging. METHOD: In this prospective study women undergoing nc-SBCT between April-October 2023 for any purpose were included in case of: histologically proven malignant lesion (ML); fibroadenoma (FA) with histologic confirmation or stability > 24 months (retrospectively); cysts with ultrasound correlation; and women with extremely dense breast (EDB) and no sonographic findings. Three regions of interest were placed on each lesion and 3 different area of EDB. The evaluation was performed by two readers (R1 and R2). Kruskal-Wallis test, intraclass correlation (ICC) and ROC analysis were used. RESULTS: 40 women with 12 ML, 10 FA, 15 cysts and 9 with EDB were included. Median density values and interquartile ranges for R1 and R2 were: 60.2 (53.3-67.3) and 62.5 (55.67-76.3) HU for ML; 46.3 (41.9-59.5) and 44.5 (40.5-59.8) HU for FA; 35.3 (24.3-46.0) and 39.7 (26.7-52.0) HU for cysts; and 28.7 (24.2-33.0) and 33.3 (31.7-36.8) HU for EDB. For both readers, densities were significantly different for ML versus EDB (p < 0.001) and cysts (p < 0.001) and for FA versus EDB (p=/<0.003). The AUC was 0.925 (95 %CI 0.858-0.993) for R1 and 0.942 (0.884-1.00) for R2 when comparing ML versus others and 0.792 (0.596-0.987) and 0.833 (0.659-1) when comparing ML versus FA. The ICC showed an almost perfect inter-reader (0.978) and intra-reader agreement (>0.879 for both readers). CONCLUSIONS: In nc-SBCT malignant lesions have higher density values compared to normal tissue and measurements of density values are reproducible between different readers.
ABSTRACT
BACKGROUND: After breast conserving surgery (BCS), surgical clips indicate the tumor bed and, thereby, the most probable area for tumor relapse. The aim of this study was to investigate whether a U-Net-based deep convolutional neural network (dCNN) may be used to detect surgical clips in follow-up mammograms after BCS. METHODS: 884 mammograms and 517 tomosynthetic images depicting surgical clips and calcifications were manually segmented and classified. A U-Net-based segmentation network was trained with 922 images and validated with 394 images. An external test dataset consisting of 39 images was annotated by two radiologists with up to 7 years of experience in breast imaging. The network's performance was compared to that of human readers using accuracy and interrater agreement (Cohen's Kappa). RESULTS: The overall classification accuracy on the validation set after 45 epochs ranged between 88.2% and 92.6%, indicating that the model's performance is comparable to the decisions of a human reader. In 17.4% of cases, calcifications have been misclassified as post-operative clips. The interrater reliability of the model compared to the radiologists showed substantial agreement (κreader1 = 0.72, κreader2 = 0.78) while the readers compared to each other revealed a Cohen's Kappa of 0.84, thus showing near-perfect agreement. CONCLUSIONS: With this study, we show that surgery clips can adequately be identified by an AI technique. A potential application of the proposed technique is patient triage as well as the automatic exclusion of post-operative cases from PGMI (Perfect, Good, Moderate, Inadequate) evaluation, thus improving the quality management workflow.
ABSTRACT
In DCE-MRI, the degree of contrast uptake in normal fibroglandular tissue, i.e., background parenchymal enhancement (BPE), is a crucial biomarker linked to breast cancer risk and treatment outcome. In accordance with the Breast Imaging Reporting & Data System (BI-RADS), it should be visually classified into four classes. The susceptibility of such an assessment to inter-reader variability highlights the urgent need for a standardized classification algorithm. In this retrospective study, the first post-contrast subtraction images for 27 healthy female subjects were included. The BPE was classified slice-wise by two expert radiologists. The extraction of radiomic features from segmented BPE was followed by dataset splitting and dimensionality reduction. The latent representations were then utilized as inputs to a deep neural network classifying BPE into BI-RADS classes. The network's predictions were elucidated at the radiomic feature level with Shapley values. The deep neural network achieved a BPE classification accuracy of 84 ± 2% (p-value < 0.00001). Most of the misclassifications involved adjacent classes. Different radiomic features were decisive for the prediction of each BPE class underlying the complexity of the decision boundaries. A highly precise and explainable pipeline for BPE classification was achieved without user- or algorithm-dependent radiomic feature selection.
ABSTRACT
Background and Purpose: The feasibility of acquiring diffusion-weighted imaging (DWI) images on an MR-Linac for quantitative response assessment during radiotherapy was explored. DWI data obtained with a Spin Echo Echo Planar Imaging sequence adapted for a 0.35 T MR-Linac were examined and compared with DWI data from a conventional 3 T scanner. Materials and Methods: Apparent diffusion coefficient (ADC) measurements and a distortion correction technique were investigated using DWI-calibrated phantoms and in the brains of seven volunteers. All DWI utilized two phase-encoding directions for distortion correction and off-resonance field estimation. ADC maps in the brain were analyzed for automatically segmented normal tissues. Results: Phantom ADC measurements on the MR-Linac were within a 3 % margin of those recorded by the 3 T scanner. The maximum distortion observed in the phantom was 2.0 mm prior to correction and 1.1 mm post-correction on the MR-Linac, compared to 6.0 mm before correction and 3.6 mm after correction at 3 T. In vivo, the average ADC values for gray and white matter exhibited variations of 14 % and 4 %, respectively, for different selections of b-values on the MR-Linac. Distortions in brain images before correction, estimated through the off-resonance field, reached 2.7 mm on the MR-Linac and 12 mm at 3 T. Conclusion: Accurate ADC measurements are achievable on a 0.35 T MR-Linac, both in phantom and in vivo. The selection of b-values significantly influences ADC values in vivo. DWI on the MR-Linac demonstrated lower distortion levels, with a maximum distortion reduced to 1.1 mm after correction.
ABSTRACT
Techniques of artificial intelligence (AI) are increasingly used in the treatment of patients, such as providing a diagnosis in radiological imaging, improving workflow by triaging patients or providing an expert opinion based on clinical symptoms; however, such AI techniques also hold intrinsic risks as AI algorithms may point in the wrong direction and constitute a black box without explaining the reason for the decision-making process.This article outlines a case where an erroneous ChatGPT diagnosis, relied upon by the patient to evaluate symptoms, led to a significant treatment delay and a potentially life-threatening situation. With this case, we would like to point out the typical risks posed by the widespread application of AI tools not intended for medical decision-making.
Subject(s)
Ischemic Attack, Transient , Humans , Ischemic Attack, Transient/diagnosis , Ischemic Attack, Transient/etiology , Artificial Intelligence , Delayed Diagnosis , Algorithms , Clinical Decision-MakingABSTRACT
OBJECTIVES: Development of automated segmentation models enabling standardized volumetric quantification of fibroglandular tissue (FGT) from native volumes and background parenchymal enhancement (BPE) from subtraction volumes of dynamic contrast-enhanced breast MRI. Subsequent assessment of the developed models in the context of FGT and BPE Breast Imaging Reporting and Data System (BI-RADS)-compliant classification. METHODS: For the training and validation of attention U-Net models, data coming from a single 3.0-T scanner was used. For testing, additional data from 1.5-T scanner and data acquired in a different institution with a 3.0-T scanner was utilized. The developed models were used to quantify the amount of FGT and BPE in 80 DCE-MRI examinations, and a correlation between these volumetric measures and the classes assigned by radiologists was performed. RESULTS: To assess the model performance using application-relevant metrics, the correlation between the volumes of breast, FGT, and BPE calculated from ground truth masks and predicted masks was checked. Pearson correlation coefficients ranging from 0.963 ± 0.004 to 0.999 ± 0.001 were achieved. The Spearman correlation coefficient for the quantitative and qualitative assessment, i.e., classification by radiologist, of FGT amounted to 0.70 (p < 0.0001), whereas BPE amounted to 0.37 (p = 0.0006). CONCLUSIONS: Generalizable algorithms for FGT and BPE segmentation were developed and tested. Our results suggest that when assessing FGT, it is sufficient to use volumetric measures alone. However, for the evaluation of BPE, additional models considering voxels' intensity distribution and morphology are required. CRITICAL RELEVANCE STATEMENT: A standardized assessment of FGT density can rely on volumetric measures, whereas in the case of BPE, the volumetric measures constitute, along with voxels' intensity distribution and morphology, an important factor. KEY POINTS: ⢠Our work contributes to the standardization of FGT and BPE assessment. ⢠Attention U-Net can reliably segment intricately shaped FGT and BPE structures. ⢠The developed models were robust to domain shift.
ABSTRACT
OBJECTIVES: Breast density is considered an independent risk factor for the development of breast cancer. This study aimed to quantitatively assess the percent breast density (PBD) and the mammary glands volume (MGV) according to the patient's age and breast quadrant. We propose a regression model to estimate PBD and MGV as a function of the patient's age. METHODS: The breast composition in 1027 spiral breast CT (BCT) datasets without soft tissue masses, calcifications, or implants from 517 women (57 ± 8 years) were segmented. The breast tissue volume (BTV), MGV, and PBD of the breasts were measured in the entire breast and each of the four quadrants. The three breast composition features were analyzed in the seven age groups, from 40 to 74 years in 5-year intervals. A logarithmic model was fitted to the BTV, and a multiplicative inverse model to the MGV and PBD as a function of age was established using a least-squares method. RESULTS: The BTV increased from 545 ± 345 to 676 ± 412 cm3, and the MGV and PBD decreased from 111 ± 164 to 57 ± 43 cm3 and from 21 ± 21 to 11 ± 9%, respectively, from the youngest to the oldest group (p < 0.05). The average PBD over all ages were 14 ± 13%. The regression models could predict the BTV, MGV, and PBD based on the patient's age with residual standard errors of 386 cm3, 67 cm3, and 13%, respectively. The reduction in MGV and PBD in each quadrant followed the ones in the entire breast. CONCLUSIONS: The PBD and MGV computed from BCT examinations provide important information for breast cancer risk assessment in women. The study quantified the breast mammary gland reduction and density decrease over the entire breast. It established mathematical models to estimate the breast composition features-BTV, MGV, and PBD, as a function of the patient's age.
ABSTRACT
Myoglobin (MB) is expressed in different cancer types and may act as a tumor suppressor in breast cancer. The mechanisms by which basal MB expression level impacts murine mammary tumorigenesis are unclear. We investigated how MB expression in breast cancer influences proliferation, metastasis, tumor hypoxia, and chemotherapy treatment in vivo. We crossed PyMT and WapCreTrp53flox mammary cancer mouse models that differed in tumor grade/type and onset of mammary carcinoma with MB knockout mice. The loss of MB in WapCre;Trp53flox mice did not affect tumor development and progression. On the other hand, loss of MB decreased tumor growth and increased tissue hypoxia as well as the number of lung metastases in PyMT mice. Furthermore, Doxorubicin therapy prevented the stronger metastatic propensity of MB-deficient tumors in PyMT mice. This suggests that, although MB expression predicts improved prognosis in breast cancer patients, MB-deficient tumors may still respond well to first-line therapies. We propose that determining the expression level of MB in malignant breast cancer biopsies will improve tumor stratification, outcome prediction, and personalized therapy in cancer patients.
Subject(s)
Carcinoma , Myoglobin , Animals , Mice , Myoglobin/genetics , Biopsy , Disease Models, Animal , Hypoxia/genetics , Mice, KnockoutABSTRACT
INTRODUCTION AND HYPOTHESIS: The purpose was to investigate the safety and feasibility of transurethral injections of autologous muscle precursor cells (MPCs) into the external urinary sphincter (EUS) to treat stress urinary incontinence (SUI) in female patients. METHODS: Prospective and randomised phase I clinical trial. Standardised 1-h pad test, International Consultation on Incontinence Questionnaire-Urinary Incontinence Short Form (ICIQ-UI-SF), urodynamic study, and MRI of the pelvis were performed at baseline and 6 months after treatment. MPCs gained through open muscle biopsy were transported to a GMP facility for processing and cell expansion. The final product was injected into the EUS via a transurethral ultrasound-guided route. Primary outcomes were defined as any adverse events (AEs) during follow-up. Secondary outcomes were functional, questionnaire, and radiological results. RESULTS: Ten female patients with SUI grades I-II were included in the study and 9 received treatment. Out of 8 AEs, 3 (37.5%) were potentially related to treatment and treated conservatively: 1 urinary tract infection healed with antibiotics treatment, 1 dysuria and 1 discomfort at biopsy site. Functional urethral length under stress was 25 mm at baseline compared with 30 mm at 6 months' follow-up (p=0.009). ICIQ-UI-SF scores improved from 7 points at baseline to 4 points at follow-up (p=0.035). MRI of the pelvis revealed no evidence of tumour or necrosis, whereas the diameter of the EUS muscle increased from 1.8 mm at baseline to 1.9 mm at follow-up (p=0.009). CONCLUSION: Transurethral injections of autologous MPCs into the EUS for treatment of SUI in female patients can be regarded as safe and feasible. Only a minimal number of expected and easily treatable AEs were documented.
Subject(s)
Urinary Incontinence, Stress , Urinary Incontinence , Humans , Female , Urinary Incontinence, Stress/therapy , Prospective Studies , Urethra/diagnostic imaging , Muscles , Treatment OutcomeABSTRACT
OBJECTIVE: In this study, we investigate the feasibility of a deep Convolutional Neural Network (dCNN), trained with mammographic images, to detect and classify microcalcifications (MC) in breast-CT (BCT) images. METHODS: This retrospective single-center study was approved by the local ethics committee. 3518 icons generated from 319 mammograms were classified into three classes: "no MC" (1121), "probably benign MC" (1332), and "suspicious MC" (1065). A dCNN was trained (70% of data), validated (20%), and tested on a "real-world" dataset (10%). The diagnostic performance of the dCNN was tested on a subset of 60 icons, generated from 30 mammograms and 30 breast-CT images, and compared to human reading. ROC analysis was used to calculate diagnostic performance. Moreover, colored probability maps for representative BCT images were calculated using a sliding-window approach. RESULTS: The dCNN reached an accuracy of 98.8% on the "real-world" dataset. The accuracy on the subset of 60 icons was 100% for mammographic images, 60% for "no MC", 80% for "probably benign MC" and 100% for "suspicious MC". Intra-class correlation between the dCNN and the readers was almost perfect (0.85). Kappa values between the two readers (0.93) and the dCNN were almost perfect (reader 1: 0.85 and reader 2: 0.82). The sliding-window approach successfully detected suspicious MC with high image quality. The diagnostic performance of the dCNN to classify benign and suspicious MC was excellent with an AUC of 93.8% (95% CI 87, 4%-100%). CONCLUSION: Deep convolutional networks can be used to detect and classify benign and suspicious MC in breast-CT images.
Subject(s)
Breast Diseases , Neural Networks, Computer , Humans , Retrospective Studies , Mammography/methods , Tomography, X-Ray Computed , ROC CurveABSTRACT
PURPOSE: To compare the positive predictive values (PPVs) of BI-RADS categories used to assess pure mammographic calcifications in women with and without a previous history of breast cancer (PHBC). MATERIALS AND METHODS: In this retrospective study, all consecutive pure mammographic calcifications (n = 320) undergoing a stereotactic biopsy between 2016 and 2018 were identified. Mammograms were evaluated in consensus by two radiologists according to BI-RADS and blinded to patient history and pathology results. Final pathologic results were used as the standard of reference. PPV of BI-RADS categories were compared between the two groups. Data were evaluated using standard statistics, Mann-Whitney U tests and Chi-square tests. RESULTS: Two hundred sixty-eight patients (274 lesions, median age 54 years, inter-quartile range, 50-65 years) with a PHBC (n = 46) and without a PHBC (n = 222) were included. Overall PPVs were the following: BI-RADS 2, 0% (0 of 56); BI-RADS 3, 9.1% (1 of 11); BI-RADS 4a, 16.2% (6 of 37); BI-RADS 4b, 37.5% (48 of 128); BI-RADS 4c, 47.3% (18 of 38) and BI-RADS 5, 100% (4 of 4). The PPV of BI-RADS categories was similar in patients with and without a PHBC (P = .715). Calcifications were more often malignant in patients with a PHBC older than 10 years (47.3%, 9 of 19) compared to 1-2 years (25%, 1 of 4), 2-5 years (20%, 2 of 10) and 5-10 years (0%, of 13) from the first breast cancer (P = .005). CONCLUSION: PPV of mammographic calcifications is similar in women with or without PHBC when BI-RADS classification is strictly applied. A higher risk of malignancy was observed in patients with a PHBC longer than 10 years.
Subject(s)
Breast Neoplasms , Calcinosis , Female , Humans , Middle Aged , Breast Neoplasms/pathology , Retrospective Studies , Mammography/methods , Biopsy , Predictive Value of TestsABSTRACT
BACKGROUND: Spiral breast computed tomography (BCT) equipped with a photon-counting detector (PCD) is a new radiological modality allowing for the compression-free acquisition of high-resolution 3-D datasets of the breast. Optimized dose exposu04170/re setups according to breast size were previously proposed but could not effectively be applied in a clinical environment due to ambiguity in measuring breast size. PURPOSE: This study aims to report the standard radiation dose values in a large cohort of patients examined with BCT, and to provide a mathematical model to estimate radiation dose based on morphological features of the breast. METHODS: This retrospective study was conducted on 1657 BCT examinations acquired between 2018 and 2021 from 829 participants (57 ± 10 years, all female). Applying a dedicated breast tissue segmentation algorithm and Monte Carlo (MC) simulation, mean absorbed dose (MAD), mean glandular dose (MGD), mean skin dose (MSD), maximum glandular dose (maxGD), and maximum skin dose (maxSD) were calculated and related to morphological features such as breast volume, effective diameter, breast length, skin volume, and glandularity. Effective dose (ED) was calculated by applying the corresponding beam and tissue weighting factors, 1 Sv/Gy and 0.12 per breast. Relevant morphological features predicting dose values were identified based on the Spearman's rank correlation coefficient. Exponential or bi-exponential models predicting the dose values as a function of morphological features were fitted by using a non-linear least squares (LS) method. The models were validated by assessing R2 and residual standard error (RSE). RESULTS: The most relevant morphological features for radiation dose estimation were the breast volume (correlation coefficient: -0.8), diameter (-0.7), and length (-0.6). The glandularity presented a weak-positive correlation (0.4) with MGD and maxGD due to the inhomogeneous distribution of the glandularity and absorbed dose in the 3-D breast volume. The standard MGDs were calculated to be 7.3 ± 0.7, 6.5 ± 0.3, and 5.9 ± 0.3 mGy, MADs to 7.6 ± 0.8, 6.8 ± 0.3, and 6.2 ± 0.3 mGy, maxSDs to 19.9 ± 1.6, 19.5 ± 0.5, and 18.9 ± 0.5 mGy, and EDs to 0.88 ± 0.08, 0.78 ± 0.04, and 0.72 ± 0.04 mSv for small, medium, and large breasts with average breast lengths of 5.9 ± 1.6, 8.7 ± 1.3, and 12.2 ± 2.0 cm, respectively. The estimated glandularity - 23.1 ± 16.9, 12.5 ± 11.4, and 6.9 ± 7.3% from small to large breasts. The mathematical models were able to estimate the MAD, MGD, MSD, and maxSD as a function of each morphological feature with only upto 0.5 mGy RSE. CONCLUSION: We presented the typical morphological features and standard dose values according to the breast size acquired from a large patient cohort. We established radiation dose estimation models allowing accurate estimation of dose values including MGD with an acceptable RSE based on each of the easily measured morphological features of the breast. Clinicians could use the breast length to operate as a dosimetric alert of the scanner prior to a BCT scan. Radiation exposure for BCT was lower than diagnostic mammography (MG) and cone-beam breast CT (BCT).
Subject(s)
Breast , Mammography , Humans , Female , Retrospective Studies , Radiation Dosage , Monte Carlo Method , Phantoms, Imaging , Breast/diagnostic imaging , Mammography/methods , Tomography, Spiral ComputedABSTRACT
OBJECTIVES: In this retrospective, single-center study we investigate the changes of radiomics features during dynamic breast-MRI for healthy tissue compared to benign and malignant lesions. METHODS: 60 patients underwent breast-MRI using a dynamic 3D gradient-echo sequence. Changes of 34 texture features (TF) in 30 benign and 30 malignant lesions were calculated for 5 dynamic datasets and corresponding 4 subtraction datasets. Statistical analysis was performed with ANOVA, and systematic changes in features were described by linear and polynomial regression models. RESULTS: ANOVA revealed significant differences (p < 0.05) between normal tissue and lesions in 13 TF, compared to 9 TF between benign and malignant lesions. Most TF showed significant differences in early dynamic and subtraction datasets. TF associated with homogeneity were suitable to discriminate between healthy parenchyma and lesions, whereas run-length features were more suitable to discriminate between benign and malignant lesions. Run length nonuniformity (RLN) was the only feature able to distinguish between all three classes with an AUC of 88.3%. Characteristic changes were observed with a systematic increase or decrease for most TF with mostly polynomial behavior. Slopes showed earlier peaks in malignant lesions, compared to benign lesions. Mean values for the coefficient of determination were higher during subtraction sequences, compared to dynamic sequences (benign: 0.98 vs 0. 72; malignant: 0.94 vs 0.74). CONCLUSIONS: TF of breast lesions follow characteristic patterns during dynamic breast-MRI, distinguishing benign from malignant lesions. Early dynamic and subtraction datasets are particularly suitable for texture analysis in breast-MRI. Features associated with tissue homogeneity seem to be indicative of benign lesions.
Subject(s)
Magnetic Resonance Imaging , Humans , Retrospective Studies , Radiography , BiomarkersABSTRACT
PURPOSE: To compare the subjective image quality assessment using B-CT and digital mammography in women with personal history of breast cancer (PHBC). METHOD: In this retrospective study 32 patients with PHBC were included. Each patient had undergone a B-CT examination and a previous mammogram in a time interval of less than 18 months between the two examinations. Two radiologists evaluated the two examinations independently with regard to the presence of lesions, BI-RADS classification, level of confidence for the overall exam interpretation, scar evaluation and image quality including image degradation due to clip artifacts. Level of confidence and image quality were assessed using a 5-point Likert scale. A p-value of less than 0.01 was considered statistically significant. RESULTS: Thirty-seven operated and 27 non-operated breasts were included. Confidence for the overall interpretation with B-CT was equal or superior to mammography in 63 cases (98.4 %) for reader 1 and in 58 cases (90.6 %) for reader 2 (p <.001). Confidence for scar evaluation with B-CT was equal or superior to mammography in all cases for reader 1 and in 34 cases (91.9 %) for readers 2 (p <.001). One case with local recurrence in B-CT was identified by both readers and no false positive findings were reported. A moderate to high image degradation due to beam-hardening artifacts has been reported by both readers in 29.4 % of cases due to surgical clips in the B-CT volume. CONCLUSIONS: B-CT in patients with PHBC provides high quality images that can be evaluated with confidence equal or superior to mammography.
Subject(s)
Breast Neoplasms , Female , Humans , Retrospective Studies , Breast Neoplasms/diagnostic imaging , Cicatrix , Tomography, X-Ray Computed , Mammography/methods , Breast/diagnostic imagingABSTRACT
Imaging of iron-based nanoparticles (NPs) remains challenging because of the presence of endogenous iron in tissues that is difficult to distinguish from exogenous iron originating from the NPs. Here, an analytical cascade for characterizing the biodistribution of biomedically relevant iron-based NPs from the organ scale to the cellular and subcellular scales is introduced. The biodistribution on an organ level is assessed by elemental analysis and quantification of magnetic iron by electron paramagnetic resonance, which allowed differentiation of exogenous and endogenous iron. Complementary to these bulk analysis techniques, correlative whole-slide optical and electron microscopy provided spatially resolved insight into the biodistribution of endo- and exogenous iron accumulation in macrophages, with single-cell and single-particle resolution, revealing coaccumulation of iron NPs with endogenous iron in splenic macrophages. Subsequent transmission electron microscopy revealed two types of morphologically distinct iron-containing structures (exogenous nanoparticles and endogenous ferritin) within membrane-bound vesicles in the cytoplasm, hinting at an attempt of splenic macrophages to extract and recycle iron from exogenous nanoparticles. Overall, this strategy enables the distinction of endo- and exogenous iron across scales (from cm to nm, based on the analysis of thousands of cells) and illustrates distribution on organ, cell, and organelle levels.
Subject(s)
Iron , Macrophages , Tissue Distribution , Microscopy, Electron , Microscopy, Electron, TransmissionABSTRACT
INTRODUCTION: The quantification of the amount of the glandular tissue and breast density is important to assess breast cancer risk. Novel photon-counting breast computed tomography (CT) technology has the potential to quantify them. For accurate analysis, a dedicated method to segment the breast components-the adipose and glandular tissue, skin, pectoralis muscle, skinfold section, rib, and implant-is required. We propose a fully automated breast segmentation method for breast CT images. METHODS: The framework consists of four parts: (1) investigate, (2) segment the components excluding adipose and glandular tissue, (3) assess the breast density, and (4) iteratively segment the glandular tissue according to the estimated density. For the method, adapted seeded watershed and region growing algorithm were dedicatedly developed for the breast CT images and optimized on 68 breast images. The segmentation performance was qualitatively (five-point Likert scale) and quantitatively (Dice similarity coefficient [DSC] and difference coefficient [DC]) demonstrated according to human reading by experienced radiologists. RESULTS: The performance evaluation on each component and overall segmentation for 17 breast CT images resulted in DSCs ranging 0.90-0.97 and in DCs 0.01-0.08. The readers rated 4.5-4.8 (5 highest score) with an excellent inter-reader agreement. The breast density varied by 3.7%-7.1% when including mis-segmented muscle or skin. CONCLUSION: The automatic segmentation results coincided with the human expert's reading. The accurate segmentation is important to avoid the significant bias in breast density analysis. Our method enables accurate quantification of the breast density and amount of the glandular tissue that is directly related to breast cancer risk.
Subject(s)
Breast Neoplasms , Breast , Humans , Female , Breast/diagnostic imaging , Tomography, X-Ray Computed/methods , Breast Density , Algorithms , Breast Neoplasms/diagnostic imagingABSTRACT
OBJECTIVE: To investigate aspects of image quality, feasibility and patient comfort in dedicated spiral breast computed tomography (B-CT) in a large patient cohort. METHODS: This retrospective study was approved by the institutional review board. 2418 B-CT scans from 1222 women examined between 04/16/2019 and 04/13/2022 were analyzed. Patients evaluated their comfort during the examination, radiographers carrying out the scans evaluated the patient's mobility and usability of the B-CT device, whereas radiologists assessed lesion contrast, detectability of calcifications, breast coverage and overall image quality. For semi-quantitative assessment, a Likert-Scale was used and statistical significance and correlations were calculated using ANOVAs and Spearman tests. RESULTS: Comfort, mobility and usability of the B-CT were rated each with either "no" or "negligible" complaints in >99%. Image quality was rated with "no" or "negligible complaints" in 96.7%. Lesion contrast and detectability of calcifications were rated either "optimal" or "good" in 92.6% and 98.4%. "Complete" and "almost complete" breast coverage were reported in 41.9%, while the pectoral muscle was found not to be covered in 56.0%. Major parts of the breast were not covered in 2.1%. Some variables were significantly correlated, such as age with comfort (ρ = -0.168, p < .001) and mobility (ρ = -0.172, p < .001) as well as patient weight with lesion contrast (ρ = 0.172, p < .001) and breast coverage (ρ = -0.109, p < .001). CONCLUSIONS: B-CT provides high image quality and contrast of soft tissue lesions as well as calcifications, while covering the pre-pectoral areas of the breast remains challenging. B-CT is easy to operate for the radiographer and comfortable for the majority of women.
Subject(s)
Calcinosis , Mammography , Breast/diagnostic imaging , Female , Humans , Mammography/methods , Patient Comfort , Retrospective Studies , Tomography, Spiral Computed/methodsABSTRACT
The aim of this study was to investigate the potential of a machine learning algorithm to classify breast cancer solely by the presence of soft tissue opacities in mammograms, independent of other morphological features, using a deep convolutional neural network (dCNN). Soft tissue opacities were classified based on their radiological appearance using the ACR BI-RADS atlas. We included 1744 mammograms from 438 patients to create 7242 icons by manual labeling. The icons were sorted into three categories: "no opacities" (BI-RADS 1), "probably benign opacities" (BI-RADS 2/3) and "suspicious opacities" (BI-RADS 4/5). A dCNN was trained (70% of data), validated (20%) and finally tested (10%). A sliding window approach was applied to create colored probability maps for visual impression. Diagnostic performance of the dCNN was compared to human readout by experienced radiologists on a "real-world" dataset. The accuracies of the models on the test dataset ranged between 73.8% and 89.8%. Compared to human readout, our dCNN achieved a higher specificity (100%, 95% CI: 85.4-100%; reader 1: 86.2%, 95% CI: 67.4-95.5%; reader 2: 79.3%, 95% CI: 59.7-91.3%), and the sensitivity (84.0%, 95% CI: 63.9-95.5%) was lower than that of human readers (reader 1:88.0%, 95% CI: 67.4-95.4%; reader 2:88.0%, 95% CI: 67.7-96.8%). In conclusion, a dCNN can be used for the automatic detection as well as the standardized and observer-independent classification of soft tissue opacities in mammograms independent of the presence of microcalcifications. Human decision making in accordance with the BI-RADS classification can be mimicked by artificial intelligence.
ABSTRACT
INTRODUCTION: Magnetic resonance enterography (MRE) is useful for detecting bowel strictures, whereas a number of imaging biomarkers may reflect severity of fibrosis burden in Crohn's disease (CD). This study aimed to verify the association of MRE metrics with histologic fibrosis independent of inflammation. METHODS: This prospective European multicenter study performed MRE imaging on 60 patients with CD with bowel strictures before surgical resection. Locations of 61 histological samples were annotated on MRE examinations, followed by central readings using the Chiorean score and measurement of delayed gain of enhancement (DGE), magnetization transfer ratio, T2-weighted MRI sequences (T2R), apparent diffusion coefficient (ADC), and the magnetic resonance index of activity (MaRIA). Correlations of histology and MRE metrics were assessed. Least Absolute Shrinkage and Selection Operator and receiver operator characteristic (ROC) curve analyses were used to select composite MRE scores predictive of histology and to estimate their predictive value. RESULTS: ADC and MaRIA correlated with fibrosis (R = -0.71, P < 0.0001, and 0.59, P < 0.001) and more moderately with inflammation (R = -0.35, P < 0.01, and R = 0.53, P < 0.001). Lower or no correlations of fibrosis or inflammation were found with DGE, magnetization transfer ratio, or T2R. Least Absolute Shrinkage and Selection Operator and ROC identified a composite score of MaRIA, ADC, and DGE as a very good predictor of histologic fibrosis (ROC area under the curve = 0.910). MaRIA alone was the best predictor of histologic inflammation with excellent performance in identifying active histologic inflammation (ROC area under the curve = 0.966). DISCUSSION: MRE-based scores for histologic fibrosis and inflammation may assist in the characterization of CD stenosis and enable development of fibrosis-targeted therapies and clinical treatment of stenotic patients.
Subject(s)
Crohn Disease , Constriction, Pathologic/diagnostic imaging , Crohn Disease/complications , Crohn Disease/diagnostic imaging , Fibrosis , Humans , Inflammation/diagnostic imaging , Magnetic Resonance Imaging/methods , Magnetic Resonance Spectroscopy , Prospective StudiesABSTRACT
BACKGROUND: We investigated whether features derived from texture analysis (TA) can distinguish breast density (BD) in spiral photon-counting breast computed tomography (PC-BCT). METHODS: In this retrospective single-centre study, we analysed 10,000 images from 400 PC-BCT examinations of 200 patients. Images were categorised into four-level density scale (a-d) using Breast Imaging Reporting and Data System (BI-RADS)-like criteria. After manual definition of representative regions of interest, 19 texture features (TFs) were calculated to analyse the voxel grey-level distribution in the included image area. ANOVA, cluster analysis, and multinomial logistic regression statistics were used. A human readout then was performed on a subset of 60 images to evaluate the reliability of the proposed feature set. RESULTS: Of the 19 TFs, 4 first-order features and 7 second-order features showed significant correlation with BD and were selected for further analysis. Multinomial logistic regression revealed an overall accuracy of 80% for BD assessment. The majority of TFs systematically increased or decreased with BD. Skewness (rho -0.81), as a first-order feature, and grey-level nonuniformity (GLN, -0.59), as a second-order feature, showed the strongest correlation with BD, independently of other TFs. Mean skewness and GLN decreased linearly from density a to d. Run-length nonuniformity (RLN), as a second-order feature, showed moderate correlation with BD, but resulted in redundant being correlated with GLN. All other TFs showed only weak correlation with BD (range -0.49 to 0.49, p < 0.001) and were neglected. CONCLUSION: TA of PC-BCT images might be a useful approach to assess BD and may serve as an observer-independent tool.