Development and Evaluation of Machine Learning in Whole-Body Magnetic Resonance Imaging for Detecting Metastases in Patients With Lung or Colon Cancer: A Diagnostic Test Accuracy Study

OBJECTIVES

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.

MATERIALS AND METHODS

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.

RESULTS

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).

CONCLUSIONS

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.

T2-weighted cardiac magnetic resonance image and myocardial biomarker in hypertrophic cardiomyopathy

The phenomenon of high signal intensity on T2-weighted imaging of cardiac magnetic resonance in hypertrophic cardiomyopathy (HCM) has been previously studied. However, the underlying histopathologic mechanism remains unclear. Elevated cardiac troponin can be detected in some HCM patients. A reasonable hypothesis is that high myocardial T2 signal is a potential marker of myocardial injury in HCM. We sought to investigate the association between cardiac troponin and the extent of high T2 signals in HCM patients.Forty-four HCM patients underwent 3.0T cardiac magnetic resonance scanning. On T2-weighted images, the number of segments with high-signal intensity (myocardium-to-skeletal muscle signal intensity ratio >2) and the percentage of high-signal area (>2 standard deviation above the remote tissue) were measured in 16 myocardial segments along the LV mid-myocardial circumference on 3 short-axis images. The level of high-sensitivity cardiac troponin T (hs-cTnT) was also assessed.Myocardial high T2 signals were identified in 33 (75%) patients and 144 (20.5%) segments. Elevated hs-cTnT was observed in 28 (63.6%) patients. The Cochran-Armitage test showed a statistically significant trend of increasing levels of hs-cTnT with elevated number of segments with myocardial high T2 signal (P = .002). Further, the percentage of myocardium with high T2 signal was significantly associated with the hs-cTnT level (Pearson correlation: r = 0.388, P = .009).Myocardium with high T2 signals was very common in patients with HCM.Its extent is related with the level of plasma hs-cTnT.

Characterization of small (≤3 cm) hepatic lesions with atypical enhancement feature and hypointensity in hepatobiliary phase of gadoxetic acid-enhanced MRI in cirrhosis: A STARD-compliant article

It is difficult to characterize the nodular lesions in cirrhotic liver if typical enhancement pattern is not present on dynamic contrast-enhanced imagings. Although the signal intensity of the hepatobiliary phase in gadoxetic acid-enhanced magnetic resonance imaging (MRI) is helpful for characterization of the lesions, some dysplastic nodules may also exhibit low signal intensity in the hepatobiliary phase. We aimed to assess the usefulness of gadoxetic acid (Gd-EOB-DTPA)-enhanced MRI including diffusion-weighted imaging (DWI) for differentiation between atypical small hepatocellular carcinomas (HCCs) and dysplastic nodules showing low signal intensity (SI) in the hepatobiliary phase, and to evaluate the MRI findings in determining the histological grade of atypical HCCs in patients with cirrhosis.A total of 43 cirrhotic patients with a small (≤3 cm) liver nodule (n = 25, HCC; n = 18, dysplastic nodule) who underwent Gd-EOB-DTPA-enhanced MRI and pathologic confirmation were retrospectively reviewed. Atypical HCC was defined as not showing arterial hyperenhancement and delayed washout on dynamic MRI.High SI on both T2WI and DWI (sensitivity 80.0%, specificity 100%, positive predictive value 100%, negative predictive value 78.3%) was the most specific feature to differentiate atypical HCCs from dysplastic nodules. High SI on both T2WI and DWI (100% vs 61.5%, P = .039) or low SI on pre-enhanced T1WI (83.3% vs 30.8%, P = .021) was more frequent observed in Edmonson grade II-III HCCs compared with those in grade I HCCs.The combination of DWI and T2WI is most useful for the differentiation of atypical small HCCs from dysplastic nodules showing low SI in the hepatobiliary phase. Combination of DWI and T2WI or pre-enhanced T1WI seems to be useful for predicting the histological grade of atypical HCCs.

Cardiovascular magnetic resonance techniques and findings in children with myocarditis: a multicenter retrospective study

BACKGROUND

Cardiovascular magnetic resonance (CMR) is increasingly used to diagnose myocarditis in adults but its use in children is not well-established. We sought to describe the presentation, CMR protocol and findings, and outcomes in a multicenter cohort of children with myocarditis.

METHODS

Thirteen hospitals retrospectively identified patients meeting the following inclusion criteria: 1) diagnosis of myocarditis by the managing physicians, 2) age <21 years, 3) CMR examination within 30 days of presentation, and 4) no congenital heart disease. Clinical data and test results, including CMR findings, were abstracted from the medical record.

RESULTS

For the 143 patients meeting inclusion criteria, the median age was 16.0 years (range, 0.1-20.3) and 139 (97 %) were hospitalized at the time of CMR. The median time from presentation to CMR was 2 days (0-28). The median left ventricular ejection fraction at CMR was 56 % (10-74), with 29 (20 %) below 45 %. The median right ventricular ejection fraction was 54 % (15-72), with 11 (8 %) below 40 %. There was significant variability among centers in the types of tissue characterization techniques employed (p < 0.001). Overall, late gadolinium enhancement (LGE) was used in 100 % of studies, followed by T2-weighted imaging (T2W) in 69 %, first-pass contrast perfusion (FPP) in 48 %, and early gadolinium enhancement (EGE) in 28 %. Abnormalities were most common with LGE (81 %), followed by T2W (74 %), EGE (55 %), and FPP (8 %). The CMR study was interpreted as positive for myocarditis in 117 patients (82 %), negative in 18 (13 %), and equivocal in 7 (5 %), yielding a sensitivity of 82 %. At a median follow-up of 7.1 months (0-87), all patients were alive and 5 had undergone cardiac transplantation. CMR parameters at presentation associated with persistent left ventricular dysfunction were larger left ventricular end-diastolic volume and lower left and right ventricular ejection fraction but not abnormal LGE.

CONCLUSIONS

Despite significant practice variation in imaging protocol among centers, CMR had a high sensitivity for the diagnosis of myocarditis in pediatric patients. Abnormalities were most often seen with LGE followed by T2W, EGE, and FPP. These findings should be useful in designing future prospective studies.

Myocardial area at risk and salvage measured by T2-weighted cardiovascular magnetic resonance: reproducibility and comparison of two T2-weighted protocols

BACKGROUND

Late gadolinium enhancement (LGE) and T2-weighted cardiovascular magnetic resonance (CMR) provides a means to measure myocardial area at risk (AAR) and salvage. Several T2-weighted CMR sequences are in use, but there is no consensus in terms of which sequence to be the preferred. Therefore, the aim of the present study was to: (1) Assess the reproducibility and (2) compare the two most frequently used T2-weighted CMR protocols for measuring AAR and salvage.

METHODS

91 patients with ST-elevation myocardial infarction treated with primary percutaneous coronary intervention underwent a CMR scan 1-7 days after initial treatment. Two different T2-weighted protocols, varying in slice thickness and echo time (TE), were applied covering the entire left ventricle (LV) (protocol 1: TE 65 msec and slice thickness 15 mm; protocol 2: TE 100 msec and slice thickness of 8 mm). On a second scan performed 3 months later, infarct size was assessed with a standard LGE sequence. The two protocols were compared in terms of AAR and salvage index. Furthermore, intra- and interobserver reproducibility were assessed.

RESULTS

Protocol 1 measures a larger AAR and salvage index than protocol 2 with a mean difference in AAR of 1±8%LV (p<0.01) and 6±12 g (p<0.01) and salvage index of 0.04±0.12 (p<0.01). Both protocols had a high intra- and interobserver reproducibility with acceptable limits of agreement (6-8%LV and 6-12 g in AAR and 0.06-0.08 in salvage index).

CONCLUSIONS

We report acceptable reproducibility for AAR and salvage index measured by T2-weighted images. Thus CMR is a reliable tool for measuring AAR and salvage index. Protocol 2 (8 mm slice thickness and 100 msec TE) measures slightly smaller AAR than protocol 1 (15 mm slice thickness and 65 msec TE), but the present study does not allow for a clear recommendation of either of the protocols.