Whole-body MRI in children aged 6-18 years. Reliability of identifying and grading high signal intensity changes within bone marrow

Machine Learning Algorithm: Texture Analysis in CNO and Application in Distinguishing CNO and Bone Marrow Growth-Related Changes on Whole-Body MRI

OBJECTIVE

The purpose of this study is to analyze the texture characteristics of chronic non-bacterial osteomyelitis (CNO) bone lesions, identified as areas of altered signal intensity on short tau inversion recovery (STIR) sequences, and to distinguish them from bone marrow growth-related changes through Machine Learning (ML) and Deep Learning (DL) analysis.

MATERIALS AND METHODS

We included a group of 66 patients with confirmed diagnosis of CNO and a group of 28 patients with suspected extra-skeletal systemic disease. All examinations were performed on a 1.5 T MRI scanner. Using the opensource 3D Slicer software version 4.10.2, the ROIs on CNO lesions and on the red bone marrow were sampled. Texture analysis (TA) was carried out using Pyradiomics. We applied an optimization search grid algorithm on nine classic ML classifiers and a Deep Learning (DL) Neural Network (NN). The model's performance was evaluated using Accuracy (ACC), AUC-ROC curves, F1-score, Positive Predictive Value (PPV), Mean Absolute Error (MAE) and Root-Mean-Square Error (RMSE). Furthermore, we used Shapley additive explanations to gain insight into the behavior of the prediction model.

RESULTS

Most predictive characteristics were selected by Boruta algorithm for each combination of ROI sequences for the characterization and classification of the two types of signal hyperintensity. The overall best classification result was obtained by the NN with ACC = 0.91, AUC = 0.93 with 95% CI 0.91-0.94, F1-score = 0.94 and PPV = 93.8%. Between classic ML methods, ensemble learners showed high model performance; specifically, the best-performing classifier was the Stack (ST) with ACC = 0.85, AUC = 0.81 with 95% CI 0.8-0.84, F1-score = 0.9, PPV = 90%.

CONCLUSIONS

Our results show the potential of ML methods in discerning edema-like lesions, in particular by distinguishing CNO lesions from hematopoietic bone marrow changes in a pediatric population. The Neural Network showed the overall best results, while a Stacking classifier, based on Gradient Boosting and Random Forest as principal estimators and Logistic Regressor as final estimator, achieved the best results between the other ML methods.

Pediatric whole-body magnetic resonance imaging: comparison of STIR and T2 Dixon sequences in the detection and grading of high signal bone marrow changes

OBJECTIVES

To compare short time inversion recovery (STIR) and T2 Dixon in the detection and grading of high signal intensity areas in bone marrow on whole-body MRI in healthy children.

METHODS

Prospective study, including whole-body 1.5-T MRIs from 77 healthy children. Two experienced radiologists in consensus identified and graded areas of high bone marrow signal on STIR and T2-weighted (T2W) turbo spin echo (TSE) Dixon images (presence, extension) in two different sessions at an interval of at least 3 weeks. In a third session, a third observer joined the two readers for an additional consensus reading with all sequences available (substitute gold standard).

RESULTS

Four hundred ninety of 545 (89.9%) high signal areas were visible on both sequences, while 27 (5.0%) were visible on STIR only and 28 (5.1%) on T2W Dixon only. Twenty-four of 27 (89%) lesions seen on STIR only, and 25/28 (89%) seen on T2W Dixon only, were graded as mildly increased signal intensity. The proportion of true positive high signal lesions was higher for the T2W Dixon images as compared to STIR (74.2% vs. 68.2%) (p = 0.029), while the proportion of false negatives was lower (25.9% vs. 31.7% (p = 0.035) for T2W Dixon and STIR, respectively). There was a moderate agreement between the T2W Dixon and STIR-based extension scores on a 0-4 scale, with a kappa of 0.45 (95% CI = 0.34-0.56).

CONCLUSIONS

Most high signal bone marrow changes identified on a 1.5-T whole-body MRI were seen on both STIR and water-only T2W Dixon, underscoring the importance of using identical protocols when following bone-marrow signal changes over time.

KEY POINTS

• Whole-body MRI is increasingly being used to diagnose and monitor diseases in children, such as chronic non-bacterial osteomyelitis, malignant/metastatic disease, and histiocytosis. • Standardized and validated imaging protocols, as well as reference standards by age for the growing skeleton are lacking. • Prospective single-center study showed that 90% of high signal bone marrow areas identified on a 1.5-T whole-body MRI in healthy children is seen on both STIR and water-only T2W Dixon, while 5% is seen on STIR only and 5% on T2W Dixon only.

[Whole-body MRI in cancer predisposition syndromes]

BACKGROUND

In recent decades, whole-body magnetic resonance imaging (WB-MRI) has become established as the modality of choice for the diagnosis, staging, and follow-up of oncological diseases as well as for the screening of cancer predisposition syndromes, such as Li-Fraumeni syndrome.

METHODS

As a comprehensive imaging modality without ionizing radiation, WB-MRI can be used repetitively and because of its excellent soft tissue contrast and high resolution provides early and precise detection of pathologies. This article discusses the technical requirements, some examination strategies and the clinical significance of typical findings of WB-MRI in patients with cancer predisposition syndromes.

Whole body magnetic resonance imaging in healthy children and adolescents. Bone marrow appearances of the axial skeleton

OBJECTIVE

To describe the findings of focal high signal on T2 weighted (T2W) images of the bone marrow in the axial skeleton as assessed by whole-body MRI in healthy and asymptomatic children and adolescents.

MATERIAL AND METHODS

We assessed the bone marrow of the mandible, shoulder girdle, thorax, spine, and pelvis on water-only Dixon T2W sequences as part of a whole-body MRI protocol in 196 healthy and asymptomatic children aged 5-19 years. Intensity (0-2 scale) and extension (1-4 scale) of focal high signal areas in the bone marrow were scored and divided into minor or major findings, based on intensity and extension to identify the potentially conspicuous lesions in a clinical setting.

RESULTS

We registered 415 areas of increased signal in the axial skeleton whereof 75 (38.3%) were major findings. Fifty-eight (29.6%) individuals had at least one major finding, mainly located in the pelvis (54, 72%). We found no differences according to gender. The number of minor findings increased with age (p = 0.020), but there were no significant differences in the number of major findings. The most conspicuous findings were in the pelvis, spine and sternum.

CONCLUSION

Non-specific bone marrow T2W hyperintensities in the axial skeleton are frequently detected on whole-body MRI in healthy, asymptomatic children. Awareness of this is important as some findings may resemble clinically silent lesions in children with suspected multifocal skeletal disease.