Synthetic MRI-Assisted and Self-Supervised Adaptive Segmentation of Organs-at-Risk (OARs) in MRI-Based Radiation Therapy

PURPOSE/OBJECTIVE(S)

This study proposes a self-supervised solution for OAR segmentation, combining patch-based adaptation and unsupervised synthesis of T2-weighted MRI data to finetune the segmentation model. The aim is to improve adaptation to patient anatomy, overcome limited annotated MRI data, and enhance the generalizability of automatic segmentation models for gynecological cancers.

MATERIALS/METHODS

The study used a patch-based cycle consistent generative adversarial network (cycle-GAN) for unsupervised MRI synthesis from CT scans of 20 patients, and a residual U-Net model for OARs segmentation. The segmentation model was trained and validated on synthetic MRI (sMRI) of 103 and 25 patient scans respectively, then finetuned on 78 MRI scans from radiation therapy fractions of 15 additional patients through three-fold cross validation. Self-supervised adaptation was applied, incorporating affine and elastic deformations, intensity shifting, and scaling. The model was trained on 96 × 96 × 96 sub-volumes and validated on entire pelvic sections of the same images. A combination of Dice and weighted cross entropy (CE) losses, with weights assigned for bladder (1), small bowel (1), rectum (2), sigmoid (2), left femoral head (0), and right femoral head (0), was used for OAR segmentation. The performance was evaluated against the model trained only on a limited number of acquired MRI data, as well as sMRI pretrained models with encoder weight freezing and either equal weighting or soft-tissue adjusted weighting.

RESULTS

Our sMRI-assisted approach showed improved performance for challenging pelvic OARs compared to the method using only the acquired MRI data. The self-supervised fraction-adaptive segmentation results indicated better performance in target soft-tissues when using at least one treatment fraction for organ-specific adaptation.

CONCLUSION

Our framework leverages pre-existing CT planning data for gynecological cancers to enhance the segmentation performance of OARs during MR-guided adaptive treatments. This approach offers substantial benefits for the radiation therapy workflow, including reduced variability in per-fraction segmentation and clinical burden. Further studies that involve human expert evaluations will be conducted to assess the practicality of this approach in radiation therapy.

Whole-body MRI: a practical guide for imaging patients with malignant bone disease

Whole-body magnetic resonance imaging (MRI) is now a crucial tool for the assessment of the extent of systemic malignant bone disease and response to treatment, and forms part of national and international recommendations for imaging patients with myeloma or metastatic prostate cancer. Recent developments in scanners have enabled acquisition of good-quality whole-body MRI data within 45 minutes on modern MRI systems from all main manufacturers. This provides complimentary morphological and functional whole-body imaging; however, lack of prior experience and acquisition times required can act as a barrier to adoption in busy radiology departments. This article aims to tackle the former by reviewing the indications and providing guidance for technical delivery and clinical interpretation of whole-body MRI for patients with malignant bone disease.

Implementing diffusion-weighted MRI for body imaging in prospective multicentre trials: current considerations and future perspectives

For body imaging, diffusion-weighted MRI may be used for tumour detection, staging, prognostic information, assessing response and follow-up. Disease detection and staging involve qualitative, subjective assessment of images, whereas for prognosis, progression or response, quantitative evaluation of the apparent diffusion coefficient (ADC) is required. Validation and qualification of ADC in multicentre trials involves examination of i) technical performance to determine biomarker bias and reproducibility and ii) biological performance to interrogate a specific aspect of biology or to forecast outcome. Unfortunately, the variety of acquisition and analysis methodologies employed at different centres make ADC values non-comparable between them. This invalidates implementation in multicentre trials and limits utility of ADC as a biomarker. This article reviews the factors contributing to ADC variability in terms of data acquisition and analysis. Hardware and software considerations are discussed when implementing standardised protocols across multi-vendor platforms together with methods for quality assurance and quality control. Processes of data collection, archiving, curation, analysis, central reading and handling incidental findings are considered in the conduct of multicentre trials. Data protection and good clinical practice are essential prerequisites. Developing international consensus of procedures is critical to successful validation if ADC is to become a useful biomarker in oncology.

KEY POINTS

• Standardised acquisition/analysis allows quantification of imaging biomarkers in multicentre trials. • Establishing "precision" of the measurement in the multicentre context is essential. • A repository with traceable data of known provenance promotes further research.

Development of a diffusion-weighted MRI protocol for multicentre abdominal imaging and evaluation of the effects of fasting on measurement of apparent diffusion coefficients (ADCs) in healthy liver

OBJECTIVE

To assess the effect of fasting and eating on estimates of apparent diffusion coefficient (ADC) in the livers of healthy volunteers using a diffusion-weighted MRI protocol with b-values of 100, 500 and 900 s mm(-2) in a multicentre study at 1.5 T.

METHODS

20 volunteers were scanned using 4 clinical 1.5-T MR scanners. Volunteers were scanned after fasting for at least 4 h and after eating a meal; the scans were repeated on a subsequent day. Median ADC estimates were calculated from all pixels in three slices near the centre of the liver. Analysis of variance (ANOVA) was used to assess the difference between ADC estimates in fasted and non-fasted states and between ADC estimates on different days.

RESULTS

ANOVA showed no difference between ADC estimates in fasted and non-fasted states (p = 0.8) nor between ADC estimates on different days (p = 0.8). The repeatability of the measurements was good, with coefficients of variation of 5.1% and 4.6% in fasted and non-fasted states, respectively.

CONCLUSION

There was no significant difference in ADC estimates between fasted and non-fasted measurements, indicating that the perfusion sensitivity of ADC estimates obtained from b-values of 100, 500 and 900 s mm(-2) is sufficiently low that changes in blood flow in the liver after eating are undetectable beyond the variability in the measurements.

ADVANCES IN KNOWLEDGE

Assessment of the effect of prandial state on ADC estimates is critical, in order to determine the appropriate patient preparation for biological validation in clinical trials.

Functional MRI and CT biomarkers in oncology

Imaging biomarkers derived from MRI or CT describe functional properties of tumours and normal tissues. They are finding increasing numbers of applications in diagnosis, monitoring of response to treatment and assessment of progression or recurrence. Imaging biomarkers also provide scope for assessment of heterogeneity within and between lesions. A wide variety of functional parameters have been investigated for use as biomarkers in oncology. Some imaging techniques are used routinely in clinical applications while others are currently restricted to clinical trials or preclinical studies. Apparent diffusion coefficient, magnetization transfer ratio and native T1 relaxation time provide information about structure and organization of tissues. Vascular properties may be described using parameters derived from dynamic contrast-enhanced MRI, dynamic contrast-enhanced CT, transverse relaxation rate (R2*), vessel size index and relative blood volume, while magnetic resonance spectroscopy may be used to probe the metabolic profile of tumours. This review describes the mechanisms of contrast underpinning each technique and the technical requirements for robust and reproducible imaging. The current status of each biomarker is described in terms of its validation, qualification and clinical applications, followed by a discussion of the current limitations and future perspectives.

Phantom for assessment of fat suppression in large field-of-view diffusion-weighted magnetic resonance imaging

We present the development and application of a phantom for assessment and optimization of fat suppression over a large field-of-view in diffusion-weighted magnetic resonance imaging at 1.5 T and 3 T. A Perspex cylinder (inner diameter 185 mm, height 300 mm) which contains a second cylinder (inner diameter 140 mm) was constructed. The inner cylinder was filled with water doped with copper sulphate and sodium chloride and the annulus was filled with corn oil, which closely matches the spectrum and longitudinal relaxation times of subcutaneous abdominal fat. Placement of the phantom on the couch at 45° to the z-axis presented an elliptical cross-section, which was of a similar size and shape to axial abdominal images. The use of a phantom for optimization of fat suppression allowed quantitative comparison between studies without the differences introduced by variability between human subjects. We have demonstrated that the phantom is suitable for selection of inversion delay times, spectral adiabatic inversion recovery delays and assessment of combinatorial methods of fat suppression. The phantom is valuable in protocol development and the assessment of new techniques, particularly in multi-centre trials.

Myocardial infarction in the neonate with normal heart and coronary arteries

Myocardial infarction is rarely recognized in the newborn. We report two cases in which the infant had a normal heart with normal coronary arteries. A review of previously described cases suggests that the most frequent cause of neonatal myocardial infarction is coronary artery occlusion secondary to paradoxical thromboembolization. It is speculated that infarction also can result from coronary hypoperfusion in asphyxiated infants. This report serves to remind the clinician that myocardial infarction can occur in the neonatal period and that an ECG should be obtained when evaluating a newborn with acute onset of shock.

A Brief Account of the Library of the Medical Society of the County of Kings

Images