Manual and semi-automated delineation of locally advanced rectal cancer subvolumes with diffusion-weighted MRI

Improved deep learning for automatic localisation and segmentation of rectal cancer on T2-weighted MRI

INTRODUCTION

The automatic segmentation approaches of rectal cancer from magnetic resonance imaging (MRI) are very valuable to relieve physicians from heavy workloads and enhance working efficiency. This study aimed to compare the segmentation accuracy of a proposed model with the other three models and the inter-observer consistency.

METHODS

A total of 65 patients with rectal cancer who underwent MRI examination were enrolled in our cohort and were randomly divided into a training cohort (n = 45) and a validation cohort (n = 20). Two experienced radiologists independently segmented rectal cancer lesions. A novel segmentation model (AttSEResUNet) was trained on T2WI based on ResUNet and attention mechanisms. The segmentation performance of the AttSEResUNet, U-Net, ResUNet and U-Net with Attention Gate (AttUNet) was compared, using Dice similarity coefficient (DSC), Hausdorff distance (HD), mean distance to agreement (MDA) and Jaccard index. The segmentation variability of automatic segmentation models and inter-observer was also evaluated.

RESULTS

The AttSEResUNet with post-processing showed perfect lesion recognition rate (100%) and false recognition rate (0), and its evaluation metrics outperformed other three models for two independent readers (observer 1: DSC = 0.839 ± 0.112, HD = 9.55 ± 6.68, MDA = 0.556 ± 0.722, Jaccard index = 0.736 ± 0.150; observer 2: DSC = 0.856 ± 0.099, HD = 11.0 ± 10.1, MDA = 0.789 ± 1.07, Jaccard index = 0.673 ± 0.130). The segmentation performance of AttSEResUNet was comparable and similar to manual variability (DSC = 0.857 ± 0.115, HD = 10.0 ± 10.0, MDA = 0.704 ± 1.17, Jaccard index = 0.666 ± 0.139).

CONCLUSION

Comparing with other three models, the proposed AttSEResUNet model was demonstrated as a more accurate model for contouring the rectal tumours in axial T2WI images, whose variability was similar to that of inter-observer.

Radiotherapy dose escalation using pre-treatment diffusion-weighted imaging in locally advanced rectal cancer: a planning study

Objectives

Diffusion-weighted MRI (DWI) may provide biologically relevant target volumes for dose-escalated radiotherapy in locally advanced rectal cancer (LARC). This planning study assessed the dosimetric feasibility of delivering hypofractionated boost treatment to intra-tumoural regions of restricted diffusion prior to conventional long-course radiotherapy.

Methods

Ten patients previously treated with curative-intent standard long-course radiotherapy (50 Gy/25#) were re-planned. Boost target volumes (BTVs) were delineated semi-automatically using 40th centile intra-tumoural apparent diffusion coefficient value with expansions (anteroposterior 11 mm, transverse 7 mm, craniocaudal 13 mm). Biased-dosed combined plans consisted of a single-fraction volumetric modulated arc therapy flattening-filter-free (VMAT-FFF) boost (phase 1) of 5, 7, or 10 Gy before long-course VMAT (phase 2). Phase 1 plans were assessed with reference to stereotactic conformality and deliverability measures. Combined plans were evaluated with reference to standard long-course therapy dose constraints.

Results

Phase 1 BTV dose targets at 5/7/10 Gy were met in all instances. Conformality constraints were met with only 1 minor violation at 5 and 7 Gy. All phase 1 and combined phase 1 + 2 plans passed patient-specific quality assurance. Combined phase 1 + 2 plans generally met organ-at-risk dose constraints. Exceptions included high-dose spillage to bladder and large bowel, predominantly in cases where previously administered, clinically acceptable non-boosted plans also could not meet constraints.

Conclusions

Targeted upfront LARC radiotherapy dose escalation to DWI-defined is feasible with appropriate patient selection and preparation.

Advances in knowledge

This is the first study to evaluate the feasibility of DWI-targeted upfront radiotherapy boost in LARC. This work will inform an upcoming clinical feasibility study.

The Role of MRI and PET/CT in Radiotherapy Target Volume Determination in Gastrointestinal Cancers-Review of the Literature

Positron emission tomography with computed tomography (PET/CT) and magnetic resonance imaging (MRI) could improve accuracy in target volume determination for gastrointestinal cancers. A systematic search of the PubMed database was performed, focusing on studies published within the last 20 years. Articles were considered eligible for the review if they included patients with anal canal, esophageal, rectal or pancreatic cancer, as well as PET/CT or MRI for radiotherapy treatment planning, and if they reported interobserver variability or changes in treatment planning volume due to different imaging modalities or correlation between the imaging modality and histopathologic specimen. The search of the literature retrieved 1396 articles. We retrieved six articles from an additional search of the reference lists of related articles. Forty-one studies were included in the final review. PET/CT seems indispensable for target volume determination of pathological lymph nodes in esophageal and anal canal cancer. MRI seems appropriate for the delineation of primary tumors in the pelvis as rectal and anal canal cancer. Delineation of the target volumes for radiotherapy of pancreatic cancer remains challenging, and additional studies are needed.

DWI-MR and PET-CT Functional Imaging for Boost Tumor Volume Delineation in Neoadjuvant Rectal Cancer Treatment

BACKGROUND/AIM

T2 weighted magnetic resonance (MR) imaging is the gold standard for locally advanced rectal cancer (LARC) staging. The potential benefit of functional imaging, as diffusion-weighted MR (DWI) and positron emission tomography-computed tomography (PET-CT), could be considered for treatment intensification strategies. Dose intensification resulted in better pathological complete response (pCR) rates. This study evaluated the inter-observer agreement between two radiation oncologists, and the difference in gross tumor volume (GTV) delineation in simulation-CT, T2-MR, DWI-MR, and PET-CT in patients with LARC.

PATIENTS AND METHODS

Two radiation oncologists prospectively delineated GTVs of 24 patients on simul-CT (CT_GTV), T2-weighted MR (T2_GTV), echo planar b1000 DWI (DWI_GTV) and PET-CT (PET_GTV). Observers' agreement was assessed using Dice index. Kruskal-Wallis test assessed differences between methods.

RESULTS

Mean CT_GTV, T2_GTV, DWI_GTV, and PET_GTV were 41.3±26.9 cc, 25.9±15.2 cc, 21±14.8 cc, and 37.7±27.7 cc for the first observer, and 42.2±27.9 cc, 27.6±16.9 cc, 19.9±14.9cc, and 34.8±24.3 cc for the second observer, respectively. Mean Dice index was 0.85 for CT_GTV, 0.84 for T2_GTV, 0.82 for DWI_GTV, and 0.89 for PET_GTV, representative of almost perfect agreement. Kruskal-Wallis test showed a statistically significant difference between methods (p=0.009). Dunn test showed there were differences between DWI_GTV vs. PET_GTV (p=0.040) and DWI_GTV vs. CT_GTV (p=0.008).

CONCLUSION

DWI resulted in smaller volume delineation compared to CT, T2-MR, and PET-CT functional images. Almost perfect agreements were reported for each imaging modality between two observers. DWI-MR seems to remain the optimal strategy for boost volume delineation for dose escalation in patients with LARC.

Predicting neoadjuvant chemoradiotherapy response with functional imaging and liquid biomarkers in locally advanced rectal cancer

INTRODUCTION

Non-invasive predictive quantitative biomarkers are required to guide treatment individualization in patients with locally advanced rectal cancer (LARC) in order to maximise therapeutic outcomes and minimise treatment toxicity. Magnetic resonance imaging (MRI), positron emission tomography (PET) and blood biomarkers have the potential to predict chemoradiotherapy (CRT) response in LARC.

AREAS COVERED

This review examines the value of functional imaging (MRI and PET) and liquid biomarkers (circulating tumor cells (CTCs) and circulating tumor nucleic acid (ctNA)) in the prediction of CRT response in LARC. Selected imaging and liquid biomarker studies are presented and the current status of the most promising imaging (apparent diffusion co-efficient (ADC), K^trans, SUV_max, metabolic tumor volume (MTV) and total lesion glycolysis (TLG) and liquid biomarkers (circulating tumor cells (CTCs), circulating tumor nucleic acid (ctNA)) is discussed. The potential applications of imaging and liquid biomarkers for treatment stratification and a pathway to clinical translation are presented.

EXPERT OPINION

Functional imaging and liquid biomarkers provide novel ways of predicting CRT response. The clinical and technical validation of the most promising imaging and liquid biopsy biomarkers in multi-centre studies with harmonised acquisition techniques is required. This will enable clinical trials to investigate treatment escalation or de-escalation pathways in rectal cancer.

Semi-automatic tumor segmentation of rectal cancer based on functional magnetic resonance imaging

•

Machine learning on magnetic resonance images (MRI) was used for tumor segmentation.

•

Voxelwise machine learning with morphological post-processing achieved good segmentation results.

•

Combining T2-weighted with functional MRI improved semi-automatic tumor segmentation.

•

Dynamic contrast enhanced MRI was the most valuable functional MRI information.

•

Tumor volume and interobserver variation were linked to measured segmentation quality.

Background and purpose

Tumor delineation is required both for radiotherapy planning and quantitative imaging biomarker purposes. It is a manual, time- and labor-intensive process prone to inter- and intraobserver variations. Semi or fully automatic segmentation could provide better efficiency and consistency. This study aimed to investigate the influence of including and combining functional with anatomical magnetic resonance imaging (MRI) sequences on the quality of automatic segmentations.

Materials and methods

T2-weighted (T2w), diffusion weighted, multi-echo T2*-weighted, and contrast enhanced dynamic multi-echo (DME) MR images of eighty-one patients with rectal cancer were used in the analysis. Four classical machine learning algorithms; adaptive boosting (ADA), linear and quadratic discriminant analysis and support vector machines, were trained for automatic segmentation of tumor and normal tissue using different combinations of the MR images as input, followed by semi-automatic morphological post-processing. Manual delineations from two experts served as ground truth. The Sørensen-Dice similarity coefficient (DICE) and mean symmetric surface distance (MSD) were used as performance metric in leave-one-out cross validation.

Results

Using T2w images alone, ADA outperformed the other algorithms, yielding a median per patient DICE of 0.67 and MSD of 3.6 mm. The performance improved when functional images were added and was highest for models based on either T2w and DME images (DICE: 0.72, MSD: 2.7 mm) or all four MRI sequences (DICE: 0.72, MSD: 2.5 mm).

Conclusion

Machine learning models using functional MRI, in particular DME, have the potential to improve automatic segmentation of rectal cancer relative to models using T2w MRI alone.

MRI-based automatic segmentation of rectal cancer using 2D U-Net on two independent cohorts

BACKGROUND

Tumor delineation is time- and labor-intensive and prone to inter- and intraobserver variations. Magnetic resonance imaging (MRI) provides good soft tissue contrast, and functional MRI captures tissue properties that may be valuable for tumor delineation. We explored MRI-based automatic segmentation of rectal cancer using a deep learning (DL) approach. We first investigated potential improvements when including both anatomical T2-weighted (T2w) MRI and diffusion-weighted MR images (DWI). Secondly, we investigated generalizability by including a second, independent cohort.

MATERIAL AND METHODS

Two cohorts of rectal cancer patients (C1 and C2) from different hospitals with 109 and 83 patients, respectively, were subject to 1.5 T MRI at baseline. T2w images were acquired for both cohorts and DWI (b-value of 500 s/mm²) for patients in C1. Tumors were manually delineated by three radiologists (two in C1, one in C2). A 2D U-Net was trained on T2w and T2w + DWI. Optimal parameters for image pre-processing and training were identified on C1 using five-fold cross-validation and patient Dice similarity coefficient (DSC_p) as performance measure. The optimized models were evaluated on a C1 hold-out test set and the generalizability was investigated using C2.

RESULTS

For cohort C1, the T2w model resulted in a median DSC_p of 0.77 on the test set. Inclusion of DWI did not further improve the performance (DSC_p 0.76). The T2w-based model trained on C1 and applied to C2 achieved a DSC_p of 0.59.

CONCLUSION

T2w MR-based DL models demonstrated high performance for automatic tumor segmentation, at the same level as published data on interobserver variation. DWI did not improve results further. Using DL models on unseen cohorts requires caution, and one cannot expect the same performance.