Automated and Semiautomated Segmentation of Rectal Tumor Volumes on Diffusion-Weighted MRI: Can It Replace Manual Volumetry?

Heeswijk, Miriam M. van; Lambregts, Doenja M. J.; Griethuysen, Joost J.M. van; Oei, Stanley; Rao, Sheng-Xiang; Graaff, Carla A.M. de; Vliegen, Roy F. A.; Beets, Geerard L.; Papanikolaou, Nikolaos; Beets‐Tan, Regina G. H.

doi:10.1016/j.ijrobp.2015.12.017

Cited by 55 publications

(67 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It would seem logical to assume that such an approach (given the excellent agreement when comparing the VOIs) would also result in a good agreement in ADC measurements if these were to be derived from these (semi-)automatically generated VOIs. It was reported that the median delineation time decreased from 180–296 s for precise manual delineation to 41–69 s for semi-automated segmentation, which entails a considerable decrease in time and input required from readers [22]. In our study, delineation time also significantly reduced with the non-precise method to a median measurement time ranging between 21 and 43 s per tumor/patient depending on the tumor volume.…”

Section: Discussionmentioning

confidence: 46%

“…Similar to the current study, the results of manual delineation were used as the reference standard. Although the main outcome was the VOI itself (and not the ADC as in our study), high ICCs of 0.91–0.97 for semi-automated tumor segmentation were reported [22]. It would seem logical to assume that such an approach (given the excellent agreement when comparing the VOIs) would also result in a good agreement in ADC measurements if these were to be derived from these (semi-)automatically generated VOIs.…”

Section: Discussionmentioning

confidence: 74%

See 1 more Smart Citation

Measuring the apparent diffusion coefficient in primary rectal tumors: is there a benefit in performing histogram analyses?

et al. 2017

Self Cite

View full text Add to dashboard Cite

PurposeThe apparent diffusion coefficient (ADC) is a potential prognostic imaging marker in rectal cancer. Typically, mean ADC values are used, derived from precise manual whole-volume tumor delineations by experts. The aim was first to explore whether non-precise circular delineation combined with histogram analysis can be a less cumbersome alternative to acquire similar ADC measurements and second to explore whether histogram analyses provide additional prognostic information.MethodsThirty-seven patients who underwent a primary staging MRI including diffusion-weighted imaging (DWI; b0, 25, 50, 100, 500, 1000; 1.5 T) were included. Volumes-of-interest (VOIs) were drawn on b1000-DWI: (a) precise delineation, manually tracing tumor boundaries (2 expert readers), and (b) non-precise delineation, drawing circular VOIs with a wide margin around the tumor (2 non-experts). Mean ADC and histogram metrics (mean, min, max, median, SD, skewness, kurtosis, 5th–95th percentiles) were derived from the VOIs and delineation time was recorded. Measurements were compared between the two methods and correlated with prognostic outcome parameters.ResultsMedian delineation time reduced from 47–165 s (precise) to 21–43 s (non-precise). The 45th percentile of the non-precise delineation showed the best correlation with the mean ADC from the precise delineation as the reference standard (ICC 0.71–0.75). None of the mean ADC or histogram parameters showed significant prognostic value; only the total tumor volume (VOI) was significantly larger in patients with positive clinical N stage and mesorectal fascia involvement.ConclusionWhen performing non-precise tumor delineation, histogram analysis (in specific 45th ADC percentile) may be used as an alternative to obtain similar ADC values as with precise whole tumor delineation. Histogram analyses are not beneficial to obtain additional prognostic information.

show abstract

Section: Discussionmentioning

confidence: 46%

Section: Discussionmentioning

confidence: 74%

Measuring the apparent diffusion coefficient in primary rectal tumors: is there a benefit in performing histogram analyses?

et al. 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…Studies have indicated that whole-volume tumour segmentations, as opposed to single slice or sample measurements, provide the most reproducible and accurate estimates of the true tumour volumes 11,12 . The main problem with manual segmentation approaches, is that these are highly time consuming, up to 18 minutes per tumour 13 , and as such unlikely to be implemented into daily clinical practice. Previous studies have explored ways to automatically perform segmentations using software algorithms 6,13 .…”

Section: Introductionmentioning

confidence: 99%

“…The main problem with manual segmentation approaches, is that these are highly time consuming, up to 18 minutes per tumour 13 , and as such unlikely to be implemented into daily clinical practice. Previous studies have explored ways to automatically perform segmentations using software algorithms 6,13 . These approaches work best on diffusion-weighted images, as these highlight tumour and suppress background tissues, thereby providing a high tumour-to-background ratio.…”

Section: Introductionmentioning

confidence: 99%

“…perirectal lymph nodes, prostate and ovaries) as well as artefacts may also show similar hyper-intensity and may not be recognized as such by typical simple segmentation algorithms causing these algorithms to fail to produce sufficiently accurate results 13 . In such cases, the manual input required from an experienced reader will not be limited to a threshold value or a seed point (in case of region growing), but will include manual corrections to adjust the segmentations for these effects 13 . Thus, there is an obvious need for smarter algorithms that can automatically localize and perform accurate segmentations of rectal tumours, which can reduce the need of expert input (Fig.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning for Fully-Automated Localization and Segmentation of Rectal Cancer on Multiparametric MR

Trebeschi

Griethuysen

Lambregts

et al. 2017

Sci Rep

238

149

View full text Add to dashboard Cite

Multiparametric Magnetic Resonance Imaging (MRI) can provide detailed information of the physical characteristics of rectum tumours. Several investigations suggest that volumetric analyses on anatomical and functional MRI contain clinically valuable information. However, manual delineation of tumours is a time consuming procedure, as it requires a high level of expertise. Here, we evaluate deep learning methods for automatic localization and segmentation of rectal cancers on multiparametric MR imaging. MRI scans (1.5T, T2-weighted, and DWI) of 140 patients with locally advanced rectal cancer were included in our analysis, equally divided between discovery and validation datasets. Two expert radiologists segmented each tumor. A convolutional neural network (CNN) was trained on the multiparametric MRIs of the discovery set to classify each voxel into tumour or non-tumour. On the independent validation dataset, the CNN showed high segmentation accuracy for reader1 (Dice Similarity Coefficient (DSC = 0.68) and reader2 (DSC = 0.70). The area under the curve (AUC) of the resulting probability maps was very high for both readers, AUC = 0.99 (SD = 0.05). Our results demonstrate that deep learning can perform accurate localization and segmentation of rectal cancer in MR imaging in the majority of patients. Deep learning technologies have the potential to improve the speed and accuracy of MRI-based rectum segmentations.

show abstract

Automatic segmentation of rectal tumor on diffusion‐weighted images by deep learning with U‐Net

Zhu

Zhang

Shi

et al. 2021

J Applied Clin Med Phys

View full text Add to dashboard Cite

Purpose Manual delineation of a rectal tumor on a volumetric image is time‐consuming and subjective. Deep learning has been used to segment rectal tumors automatically on T2‐weighted images, but automatic segmentation on diffusion‐weighted imaging is challenged by noise, artifact, and low resolution. In this study, a volumetric U‐shaped neural network (U‐Net) is proposed to automatically segment rectal tumors on diffusion‐weighted images. Methods Three hundred patients of locally advanced rectal cancer were enrolled in this study and divided into a training group, a validation group, and a test group. The region of rectal tumor was delineated on the diffusion‐weighted images by experienced radiologists as the ground truth. A U‐Net was designed with a volumetric input of the diffusion‐weighted images and an output of segmentation with the same size. A semi‐automatic segmentation method was used for comparison by manually choosing a threshold of gray level and automatically selecting the largest connected region. Dice similarity coefficient (DSC) was calculated to evaluate the methods. Results On the test group, deep learning method (DSC = 0.675 ± 0.144, median DSC is 0.702, maximum DSC is 0.893, and minimum DSC is 0.297) showed higher segmentation accuracy than the semi‐automatic method (DSC = 0.614 ± 0.225, median DSC is 0.685, maximum DSC is 0.869, and minimum DSC is 0.047). Paired t‐test shows significant difference (T = 2.160, p = 0.035) in DSC between the deep learning method and the semi‐automatic method in the test group. Conclusion Volumetric U‐Net can automatically segment rectal tumor region on DWI images of locally advanced rectal cancer.

show abstract

Automated and Semiautomated Segmentation of Rectal Tumor Volumes on Diffusion-Weighted MRI: Can It Replace Manual Volumetry?

Cited by 55 publications

References 21 publications

Measuring the apparent diffusion coefficient in primary rectal tumors: is there a benefit in performing histogram analyses?

Measuring the apparent diffusion coefficient in primary rectal tumors: is there a benefit in performing histogram analyses?

Deep Learning for Fully-Automated Localization and Segmentation of Rectal Cancer on Multiparametric MR

Automatic segmentation of rectal tumor on diffusion‐weighted images by deep learning with U‐Net

Contact Info

Product

Resources

About