PURPOSEThe aim of this study was to assess the potential of machine learning with multiparametric magnetic resonance imaging (mpMRI) for the early prediction of pathological complete response (pCR) ...to neoadjuvant chemotherapy (NAC) and of survival outcomes in breast cancer patients.
MATERIALS AND METHODSThis institutional review board–approved prospective study included 38 women (median age, 46.5 years; range, 25–70 years) with breast cancer who were scheduled for NAC and underwent mpMRI of the breast at 3 T with dynamic contrast-enhanced (DCE), diffusion-weighted imaging (DWI), and T2-weighted imaging before and after 2 cycles of NAC. For each lesion, 23 features were extractedqualitative T2-weighted and DCE-MRI features according to BI-RADS (Breast Imaging Reporting and Data System), quantitative pharmacokinetic DCE features (mean plasma flow, volume distribution, mean transit time), and DWI apparent diffusion coefficient (ADC) values. To apply machine learning to mpMRI, 8 classifiers including linear support vector machine, linear discriminant analysis, logistic regression, random forests, stochastic gradient descent, decision tree, adaptive boosting, and extreme gradient boosting (XGBoost) were used to rank the features. Histopathologic residual cancer burden (RCB) class (with RCB 0 being a pCR), recurrence-free survival (RFS), and disease-specific survival (DSS) were used as the standards of reference. Classification accuracy with area under the receiving operating characteristic curve (AUC) was assessed using all the extracted qualitative and quantitative features for pCR as defined by RCB class, RFS, and DSS using recursive feature elimination. To overcome overfitting, 4-fold cross-validation was used.
RESULTSMachine learning with mpMRI achieved stable performance as shown by mean classification accuracies for the prediction of RCB class (AUC, 0.86) and DSS (AUC, 0.92) based on XGBoost and the prediction of RFS (AUC, 0.83) with logistic regression. The XGBoost classifier achieved the most stable performance with high accuracies compared with other classifiers. The most relevant features for the prediction of RCB class were as followschanges in lesion size, complete pattern of shrinkage, and mean transit time on DCE-MRI; minimum ADC on DWI; and peritumoral edema on T2-weighted imaging. The most relevant features for prediction of RFS were as followsvolume distribution, mean plasma flow, and mean transit time; DCE-MRI lesion size; minimum, maximum, and mean ADC with DWI. The most relevant features for prediction of DSS were as followslesion size, volume distribution, and mean plasma flow on DCE-MRI, and maximum ADC with DWI.
CONCLUSIONSMachine learning with mpMRI of the breast enables early prediction of pCR to NAC as well as survival outcomes in breast cancer patients with high accuracy and thus may provide valuable predictive information to guide treatment decisions.
The shoulder, a very complex joint, offers a wide range of pathologies. Intraarticular abnormalities and rotator cuff injuries are mainly assessed and diagnosed by magnetic resonance arthrography ...(MRA). In contrast to this well-established gold standard, high-resolution ultrasound (US) offers an additional easy and excellent modality to assess the shoulder joint. Therefore, the purpose of this study was to evaluate in which anatomic structures and pathologies comparable results of US and MRA could be achieved.
In this IRB-approved prospective study 67 patients with clinically suspected labral lesions, rotator cuff rupture, or injury of the long head of the biceps (LHB) tendon were enrolled. Each participant was examined with high resolution US, and directly followed by MRA at 3 Tesla with a standard sequence protocol. To evaluate the agreement of the diagnostic performance between US and MRA a weighted kappa statistic was used.
Both of the investigated modalities yielded a moderate to almost perfect agreement in assessing a wide range of shoulder joint pathologies. For the rotator cuff, consistency was found in 71.64% for the supraspinatus tendon, in 95.52% for the infraspinatus tendon, in 83.58% for the subscapularis tendon, and in 98.51% for the teres minor tendon. The diagnostic accuracy between both modalities was 80.60% for the LHB tendon, 77.61% for the posterior labroligamentous complex, 83.58% for the acromioclavicular joint, and 91.04% for the assessment of osseous irregularities and impaction fractures.
High resolution US is a reliable imaging modality for the rotator cuff, the LHB tendon, and the acromioclavicular joint, so for these structures we recommend a preference for US over MRA based on its diagnostic accuracy, comfortability, cost effectiveness, and availability. If the diagnosis remains elusive, for all other intraarticular structures we recommend MRA for further diagnostic assessment.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
OBJECTIVESThe aim of this study was to evaluate breast multiparametric ultrasound (mpUS) and its potential to reduce unnecessary breast biopsies with 1, 2, or 3 additional quantitative parameters ...(Doppler, elastography, and contrast-enhanced ultrasound CEUS) to B-mode and investigate possible variations with different reader experience.
MATERIALS AND METHODSThis prospective study included 124 women (age range, 18–82 years; mean, 52 years), each with 1 new breast lesion, scheduled for ultrasound-guided biopsy between October 2015 and September 2016. Each lesion was examined with B-mode, elastography (Virtual Touch IQ VTIQ), Doppler, and CEUS, and different quantitative parameters were recorded for each modality. Four readers (2 experienced breast radiologists and 2 in-training) independently evaluated B-mode images of each lesion and assigned a BI-RADS (Breast Imaging Reporting and Data System) score. Using the area under the receiver operating characteristic curve (AUC), the most accurate quantitative parameter for each modality was chosen. These were then combined with the BI-RADS scores of all readers. Descriptive statistics and AUC were used to evaluate the diagnostic performance of mpUS.
RESULTSSixty-five lesions were malignant. MpUS with B-mode and 2 additional quantitative parameters (VTIQ and CEUS or Doppler) showed the highest diagnostic performance for all readers (averaged AUCs, 0.812–0.789 respectively vs 0.683 for B-mode, P = 0.0001). Both combinations significantly reduced the number of false-positive findings up to 46.9% (P < 0.0001).
CONCLUSIONSQuantitative mpUS with 2 different triple assessment modalities (B-mode, VTIQ elastography, CEUS, or Doppler) shows the best diagnostic performance for breast cancer diagnosis and leads to a significant reduction of false-positive biopsy recommendations, for both experienced and inexperienced readers.
Objectives
To investigate the influence of region-of-interest (ROI) placement and different apparent diffusion coefficient (ADC) parameters on ADC values, diagnostic performance, reproducibility and ...measurement time in breast tumours.
Methods
In this IRB-approved, retrospective study, 149 histopathologically proven breast tumours (109 malignant, 40 benign) in 147 women (mean age 53.2) were investigated. Three radiologists independently measured minimum, mean and maximum ADC, each using three ROI placement approaches:1 – small 2D-ROI, 2 – large 2D-ROI and 3 – 3D-ROI covering the whole lesion. One reader performed all measurements twice. Median ADC values, diagnostic performance, reproducibility, and measurement time were calculated and compared between all combinations of ROI placement approaches and ADC parameters.
Results
Median ADC values differed significantly between the ROI placement approaches (
p
< .001). Minimum ADC showed the best diagnostic performance (AUC .928–.956), followed by mean ADC obtained from 2D ROIs (.926–.94). Minimum and mean ADC showed high intra- (ICC .85–.94) and inter-reader reproducibility (ICC .74–.94). Median measurement time was significantly shorter for the 2D ROIs (
p
< .001).
Conclusions
ROI placement significantly influences ADC values measured in breast tumours. Minimum and mean ADC acquired from 2D-ROIs are useful for the differentiation of benign and malignant breast lesions, and are highly reproducible, with rapid measurement.
Key Points
• Region of interest placement significantly influences apparent diffusion coefficient of breast tumours.
• Minimum and mean apparent diffusion coefficient perform best and are reproducible.
• 2D regions of interest perform best and provide rapid measurement times.
Objectives
To determine whether 3D acquisitions provide equivalent image quality, lesion delineation quality and PI-RADS v2 performance compared to 2D acquisitions in T2-weighted imaging of the ...prostate at 3 T.
Methods
This IRB-approved, prospective study included 150 consecutive patients (mean age 63.7 years, 35–84 years; mean PSA 7.2 ng/ml, 0.4–31.1 ng/ml). Two uroradiologists (R1, R2) independently rated image quality and lesion delineation quality using a five-point ordinal scale and assigned a PI-RADS score for 2D and 3D T2-weighted image data sets. Data were compared using visual grading characteristics (VGC) and receiver operating characteristics (ROC)/area under the curve (AUC) analysis.
Results
Image quality was similarly good to excellent for 2D T2w (mean score R1, 4.3 ± 0.81; R2, 4.7 ± 0.83) and 3D T2w (mean score R1, 4.3 ± 0.82; R2, 4.7 ± 0.69),
p
= 0.269. Lesion delineation was rated good to excellent for 2D (mean score R1, 4.16 ± 0.81; R2, 4.19 ± 0.92) and 3D T2w (R1, 4.19 ± 0.94; R2, 4.27 ± 0.94) without significant differences (
p
= 0.785). ROC analysis showed an equivalent performance for 2D (AUC 0.580–0.623) and 3D (AUC 0.576–0.629) T2w (
p
> 0.05, respectively).
Conclusions
Three-dimensional acquisitions demonstrated equivalent image and lesion delineation quality, and PI-RADS v2 performance, compared to 2D in T2-weighted imaging of the prostate. Three-dimensional T2-weighted imaging could be used to considerably shorten prostate MRI protocols in clinical practice.
Key points
• 3D shows equivalent image quality and lesion delineation compared to 2D T2w.
• 3D T2w and 2D T2w image acquisition demonstrated comparable diagnostic performance.
• Using a single 3D T2w acquisition may shorten the protocol by 40%.
• Combined with short DCE, multiparametric protocols of 10 min are feasible.
Purpose
To investigate the impact of a scoring system (
Tree
) on inter-reader agreement and diagnostic performance in breast MRI reading.
Materials and methods
This IRB-approved, single-centre study ...included 100 patients with 121 consecutive histopathologically verified lesions (52 malignant, 68 benign). Four breast radiologists with different levels of MRI experience and blinded to histopathology retrospectively evaluated all examinations. Readers independently applied two methods to classify breast lesions: BI-RADS and
Tree
. BI-RADS provides a reporting lexicon that is empirically translated into likelihoods of malignancy;
Tree
is a scoring system that results in a diagnostic category. Readings were compared by ROC analysis and kappa statistics.
Results
Inter-reader agreement was substantial to almost perfect (kappa: 0.643–0.896) for
Tree
and moderate (kappa: 0.455–0.657) for BI-RADS. Diagnostic performance using
Tree
(AUC: 0.889–0.943) was similar to BI-RADS (AUC: 0.872–0.953). Less experienced radiologists achieved AUC: improvements up to 4.7 % using
Tree
(
P
-values: 0.042–0.698); an expert’s performance did not change (
P
= 0.526). The least experienced reader improved in specificity using
Tree
(16 %,
P
= 0.001). No further sensitivity and specificity differences were found (
P
> 0.1).
Conclusion
The
Tree
scoring system improves inter-reader agreement and achieves a diagnostic performance similar to that of BI-RADS. Less experienced radiologists, in particular, benefit from
Tree
.
Key Points
•
The Tree scoring system shows high diagnostic accuracy in mass and non-mass lesions
.
•
The Tree scoring system reduces inter-reader variability related to reader experience
.
•
The Tree scoring system improves diagnostic accuracy in non-expert readers
.
Purpose
The purpose of this study was to compare three different biopsy devices on false-negative and underestimation rates in MR-guided, vacuum-assisted breast biopsy (VABB) of MRI-only lesions.
...Methods
This retrospective, single-center study was IRB-approved. Informed consent was waived. 467 consecutive patients underwent 487 MR-guided VABB using three different 8-10-gauge-VABB devices (Atec-9-gauge,A; Mammotome-8-gauge,M; Vacora-10-gauge,V). VABB data (lesion-type, size, biopsy device, histopathology) were compared to final diagnosis (surgery,
n
= 210 and follow-up,
n
= 277). Chi-square, and Kruskal–Wallis tests were applied.
P
values < 0.05 were considered significant.
Results
Final diagnosis was malignant in 104 (21.4 %), high risk in 64 (13.1 %) and benign in 319 (65.5 %) cases. Eleven of 328 (3.4 %) benign-rated lesions were false-negative (1/95, 1.1 %, A; 2/73, 2.7 %, M; 8/160 5.0 % V;
P
= 0.095). Eleven high-risk (11/77, 14.3 %) lesions proved to be malignant (3/26, 11.5 % A; 4/12, 33.3 % M; 4/39, 10.3 % V;
P
= 0.228). Five of 34 (14.7 %) DCIS were upgraded to invasive cancer (2/15, 13.3 %, A; 1/6, 16.6 % M; 2/13, 15.3 %, V;
P
= 0.977). Lesion size (
P
= 0.05) and type (mass vs. non-mass,
P
= 0.107) did not differ significantly.
Conclusion
MR-guided VABB is an accurate method for diagnosis of MRI-only lesions. No significant differences on false-negative and underestimation rates were observed between three different biopsy devices.
Key Points
•
MR-guided VABB is an accurate procedure for the diagnosis of MRI-only lesions.
•
Similar false-negative and underestimation rates allow all three different MR-guided VABB devices for clinical application.
•
High-risk lesions should undergo surgery due to a substantial underestimation rate.
•
Agreement between MR-guided VABB and final diagnosis (benign/malignant) was 95.5% (465/487).
Whole-body magnetic resonance imaging (WB-MRI) has been demonstrated to be efficient and cost-effective for cancer staging. The study aim was to develop a machine learning (ML) algorithm to improve ...radiologists' sensitivity and specificity for metastasis detection and reduce reading times.
A retrospective analysis of 438 prospectively collected WB-MRI scans from multicenter Streamline studies (February 2013-September 2016) was undertaken. Disease sites were manually labeled using Streamline reference standard. Whole-body MRI scans were randomly allocated to training and testing sets. A model for malignant lesion detection was developed based on convolutional neural networks and a 2-stage training strategy. The final algorithm generated lesion probability heat maps. Using a concurrent reader paradigm, 25 radiologists (18 experienced, 7 inexperienced in WB-/MRI) were randomly allocated WB-MRI scans with or without ML support to detect malignant lesions over 2 or 3 reading rounds. Reads were undertaken in the setting of a diagnostic radiology reading room between November 2019 and March 2020. Reading times were recorded by a scribe. Prespecified analysis included sensitivity, specificity, interobserver agreement, and reading time of radiology readers to detect metastases with or without ML support. Reader performance for detection of the primary tumor was also evaluated.
Four hundred thirty-three evaluable WB-MRI scans were allocated to algorithm training (245) or radiology testing (50 patients with metastases, from primary 117 colon n = 117 or lung n = 71 cancer). Among a total 562 reads by experienced radiologists over 2 reading rounds, per-patient specificity was 86.2% (ML) and 87.7% (non-ML) (-1.5% difference; 95% confidence interval CI, -6.4%, 3.5%; P = 0.39). Sensitivity was 66.0% (ML) and 70.0% (non-ML) (-4.0% difference; 95% CI, -13.5%, 5.5%; P = 0.344). Among 161 reads by inexperienced readers, per-patient specificity in both groups was 76.3% (0% difference; 95% CI, -15.0%, 15.0%; P = 0.613), with sensitivity of 73.3% (ML) and 60.0% (non-ML) (13.3% difference; 95% CI, -7.9%, 34.5%; P = 0.313). Per-site specificity was high (>90%) for all metastatic sites and experience levels. There was high sensitivity for the detection of primary tumors (lung cancer detection rate of 98.6% with and without ML 0.0% difference; 95% CI, -2.0%, 2.0%; P = 1.00, colon cancer detection rate of 89.0% with and 90.6% without ML -1.7% difference; 95% CI, -5.6%, 2.2%; P = 0.65). When combining all reads from rounds 1 and 2, reading times fell by 6.2% (95% CI, -22.8%, 10.0%) when using ML. Round 2 read-times fell by 32% (95% CI, 20.8%, 42.8%) compared with round 1. Within round 2, there was a significant decrease in read-time when using ML support, estimated as 286 seconds (or 11%) quicker ( P = 0.0281), using regression analysis to account for reader experience, read round, and tumor type. Interobserver variance suggests moderate agreement, Cohen κ = 0.64; 95% CI, 0.47, 0.81 (with ML), and Cohen κ = 0.66; 95% CI, 0.47, 0.81 (without ML).
There was no evidence of a significant difference in per-patient sensitivity and specificity for detecting metastases or the primary tumor using concurrent ML compared with standard WB-MRI. Radiology read-times with or without ML support fell for round 2 reads compared with round 1, suggesting that readers familiarized themselves with the study reading method. During the second reading round, there was a significant reduction in reading time when using ML support.