Efficient Segmentation for Left Atrium With Convolution Neural Network Based on Active Learning in Late Gadolinium Enhancement Magnetic Resonance Imaging

BACKGROUND

To propose fully automatic segmentation of left atrium using active learning with limited dataset in late gadolinium enhancement in cardiac magnetic resonance imaging (LGE-CMRI).

METHODS

An active learning framework was developed to segment the left atrium in cardiac LGE-CMRI. Patients (n = 98) with atrial fibrillation from the Korea University Anam Hospital were enrolled. First, 20 cases were delineated for ground truths by two experts and used for training a draft model. Second, the 20 cases from the first step and 50 new cases, corrected in a human-in-the-loop manner after predicting using the draft model, were used to train the next model; all 98 cases (70 cases from the second step and 28 new cases) were trained. An additional 20 LGE-CMRI were evaluated in each step.

RESULTS

The Dice coefficients for the three steps were 0.85 ± 0.06, 0.89 ± 0.02, and 0.90 ± 0.02, respectively. The biases (95% confidence interval) in the Bland-Altman plots of each step were 6.36% (-14.90-27.61), 6.21% (-9.62-22.03), and 2.68% (-8.57-13.93). Deep active learning-based annotation times were 218 ± 31 seconds, 36.70 ± 18 seconds, and 36.56 ± 15 seconds, respectively.

CONCLUSION

Deep active learning reduced annotation time and enabled efficient training on limited LGE-CMRI.

Guideline-concordant utilization of magnetic resonance imaging in adults receiving chiropractic manipulative therapy vs other care for radicular low back pain: a retrospective cohort study

BACKGROUND

Lumbar magnetic resonance imaging (LMRI) is often performed early in the course of care, which can be discordant with guidelines for non-serious low back pain. Our primary hypothesis was that adults receiving chiropractic spinal manipulative therapy (CSMT) for incident radicular low back pain (rLBP) would have reduced odds of early LMRI over 6-weeks' follow-up compared to those receiving other care (a range of medical care, excluding CSMT). As a secondary hypothesis, CSMT recipients were also expected to have reduced odds of LMRI over 6-months' and 1-years' follow-up.

METHODS

A national 84-million-patient health records database including large academic healthcare organizations (TriNetX) was queried for adults age 20-70 with rLBP newly-diagnosed between January 31, 2012 and January 31, 2022. Receipt or non-receipt of CSMT determined cohort allocation. Patients with prior lumbar imaging and serious pathology within 90 days of diagnosis were excluded. Propensity score matching controlled for variables associated with LMRI utilization (e.g., demographics). Odds ratios (ORs) of LMRI over 6-weeks', 6-months', and 1-years' follow-up after rLBP diagnosis were calculated.

RESULTS

After matching, there were 12,353 patients per cohort (mean age 50 years, 56% female), with a small but statistically significant reduction in odds of early LMRI in the CSMT compared to other care cohort over 6-weeks' follow-up (9%, 10%, OR [95% CI] 0.88 [0.81-0.96] P = 0.0046). There was a small but statistically significant increase in odds of LMRI among patients in the CSMT relative to the other care cohort over 6-months' (12%, 11%, OR [95% CI] 1.10 [1.02-1.19], P < 0.0174) and 1-years' follow-up (14%, 12%, OR [95% CI] 1.21 [1.13-1.31], P < 0.0001).

CONCLUSIONS

These results suggest that patients receiving CSMT for newly-diagnosed rLBP are less likely to receive early LMRI than patients receiving other care. However, CSMT recipients have a small increase in odds of LMRI over the long-term. Both cohorts in this study had a relatively low rate of early LMRI, possibly because the data were derived from academic healthcare organizations. The relationship of these findings to other patient care outcomes and cost should be explored in a future randomized controlled trial.

REGISTRATION

Open Science Framework ( https://osf.io/t9myp ).

Volume estimation of brain ventricles using Cavalieri's principle and Atlas-based methods in Alzheimer disease: Consistency between methods

Automatic estimations of brain ventricles are needed to assess disease progression in neurodegenerative disorders such as Alzheimer Disease (AD). The objectives of this study are to evaluate the diagnostic performances of an automated volumetric assessment tool in estimating lateral ventricle volumes in AD and to compare this with Cavalieri's principle, which is accepted as the gold standard method. This is across-sectional volumetric study including 25 Alzheimer patients and 25 healthy subjects undergoing magnetic resonance images (MRI) with a 3D turbo spin echo sequence at 1.5 Tesla. The Atlas-based method incorporated MRIStudio software to automatically measure he volumes of brain ventricles. To compare the corresponding measurements, we used manual point-counting and semi-automatic planimetry methods based on Cavalieri's principle. Bland-Altman test results indicated an excellent agreement between Cavalieri's principle and the Atlas-based method in all volumetric measurements (p < 0.05). We obtained a 64% sensitivity and 92% specificity for lateral ventricular volumes according to the Atlas-based method. AD subjects had significantly larger left and right lateral ventricle volume (LVV) when compared to control subjects in respect to three volumetric methods (p < 0.01). Lateral ventricle-to-brain ratio (VBR) statistically increased 49.23% in measurements done with the point-counting method, 45.12% with the planimetry method, and 45.49% with the Atlas-based method in AD patients (p < 0.01). As a result, the Atlas-based method may be used instead of manual volumetry to estimate brain volumes. Additionally, this method provides rapid and accurate estimations of brain ventricular volumes in-vivo examination of MRI.