Ensemble classification and segmentation for intracranial metastatic tumors on MRI images based on 2D U-nets

RECIST 1.1 Target Lesion Categorical Response in Metastatic Renal Cell Carcinoma: A Comparison of Conventional versus Volumetric Assessment

Purpose To investigate Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1) approximations of target lesion tumor burden by comparing categorical treatment response according to conventional RECIST versus actual tumor volume measurements of RECIST target lesions. Materials and Methods This is a retrospective cohort study of individuals with metastatic renal cell carcinoma enrolled in a clinical trial (from 2003 to 2017) and includes individuals who underwent baseline and at least one follow-up chest, abdominal, and pelvic CT study and with at least one target lesion. Target lesion volume was assessed by (a) Vmodel, a spherical model of conventional RECIST 1.1, which was extrapolated from RECIST diameter, and (b) Vactual, manually contoured volume. Volumetric responses were determined by the sum of target lesion volumes (Vmodel-sum TL and Vactual-sum TL, respectively). Categorical volumetric thresholds were extrapolated from RECIST. McNemar tests were used to compare categorical volume responses. Results Target lesions were assessed at baseline (638 participants), week 9 (593 participants), and week 17 (508 participants). Vmodel-sum TL classified more participants as having progressive disease (PD), compared with Vactual-sum TL at week 9 (52 vs 31 participants) and week 17 (57 vs 39 participants), with significant overall response discordance (P < .001). At week 9, 25 (48%) of 52 participants labeled with PD by Vmodel-sum TL were classified as having stable disease by Vactual-sum TL. Conclusion A model of RECIST 1.1 based on a single diameter measurement more frequently classified PD compared with response assessment by actual measured tumor volume. Keywords: Urinary, Kidney, Metastases, Oncology, Tumor Response, Volume Analysis, Outcomes Analysis ClinicalTrials.gov registration no. NCT01865747 © RSNA, 2023 Supplemental material is available for this article.

Automated Detection of Brain Metastases on T1-Weighted MRI Using a Convolutional Neural Network: Impact of Volume Aware Loss and Sampling Strategy

BACKGROUND

Detection of brain metastases (BM) and segmentation for treatment planning could be optimized with machine learning methods. Convolutional neural networks (CNNs) are promising, but their trade-offs between sensitivity and precision frequently lead to missing small lesions.

HYPOTHESIS

Combining volume aware (VA) loss function and sampling strategy could improve BM detection sensitivity.

STUDY TYPE

Retrospective.

POPULATION

A total of 530 radiation oncology patients (55% women) were split into a training/validation set (433 patients/1460 BM) and an independent test set (97 patients/296 BM).

FIELD STRENGTH/SEQUENCE

1.5 T and 3 T, contrast-enhanced three-dimensional (3D) T1-weighted fast gradient echo sequences.

ASSESSMENT

Ground truth masks were based on radiotherapy treatment planning contours reviewed by experts. A U-Net inspired model was trained. Three loss functions (Dice, Dice + boundary, and VA) and two sampling methods (label and VA) were compared. Results were reported with Dice scores, volumetric error, lesion detection sensitivity, and precision. A detected voxel within the ground truth constituted a true positive.

STATISTICAL TESTS

McNemar's exact test to compare detected lesions between models. Pearson's correlation coefficient and Bland-Altman analysis to compare volume agreement between predicted and ground truth volumes. Statistical significance was set at P ≤ 0.05.

RESULTS

Combining VA loss and VA sampling performed best with an overall sensitivity of 91% and precision of 81%. For BM in the 2.5-6 mm estimated sphere diameter range, VA loss reduced false negatives by 58% and VA sampling reduced it further by 30%. In the same range, the boundary loss achieved the highest precision at 81%, but a low sensitivity (24%) and a 31% Dice loss.

DATA CONCLUSION

Considering BM size in the loss and sampling function of CNN may increase the detection sensitivity regarding small BM. Our pipeline relying on a single contrast-enhanced T1-weighted MRI sequence could reach a detection sensitivity of 91%, with an average of only 0.66 false positives per scan.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

Enhanced segmentation of gastrointestinal polyps from capsule endoscopy images with artifacts using ensemble learning

BACKGROUND

Endoscopy artifacts are widespread in real capsule endoscopy (CE) images but not in high-quality standard datasets.

AIM

To improve the segmentation performance of polyps from CE images with artifacts based on ensemble learning.

METHODS

We collected 277 polyp images with CE artifacts from 5760 h of videos from 480 patients at Guangzhou First People’s Hospital from January 2016 to December 2019. Two public high-quality standard external datasets were retrieved and used for the comparison experiments. For each dataset, we randomly segmented the data into training, validation, and testing sets for model training, selection, and testing. We compared the performance of the base models and the ensemble model in segmenting polyps from images with artifacts.

RESULTS

The performance of the semantic segmentation model was affected by artifacts in the sample images, which also affected the results of polyp detection by CE using a single model. The evaluation based on real datasets with artifacts and standard datasets showed that the ensemble model of all state-of-the-art models performed better than the best corresponding base learner on the real dataset with artifacts. Compared with the corresponding optimal base learners, the intersection over union (IoU) and dice of the ensemble learning model increased to different degrees, ranging from 0.08% to 7.01% and 0.61% to 4.93%, respectively. Moreover, in the standard datasets without artifacts, most of the ensemble models were slightly better than the base learner, as demonstrated by the IoU and dice increases ranging from -0.28% to 1.20% and -0.61% to 0.76%, respectively.

CONCLUSION

Ensemble learning can improve the segmentation accuracy of polyps from CE images with artifacts. Our results demonstrated an improvement in the detection rate of polyps with interference from artifacts.

Efficacy Evaluation of Ultrasound with Active Contour Model for Hemodialysis in Children with Renal Failure

This study was aimed to explore the efficacy of ultrasound with active contour model (ACM) for hemodialysis in children with renal failure. The pulse coupled neural network (PCNN) was used to extract the initial contour of the ultrasound images, and the cloud model-based ACM was used to accurately segment the images, whose effect was compared with the classic Snake model. 84 children with chronic renal failure who received hemodialysis treatment in hospital were selected as research objects. There were 42 cases in the control group who were diagnosed by conventional ultrasound and 42 cases in the observation group who were diagnosed by ultrasound with the algorithm. Then, 42 children who underwent healthy physical examination (health group) were selected for comparison of related analysis indicators. The error rates of different algorithms were compared to analyze the levels of inflammatory factors in different groups of patients after hemodialysis. The results showed that the error rate of classical Snake model was 18.87% and that of ACM algorithm model was 11.01%, and the error rate of ACM algorithm model was significantly lower ( $P<0.05$ ). After hemodialysis, the level of tumor necrosis factor (TNF)-α was 38.76 pg/mL in the observation group and 40.05 pg/mL in the control group, which was notably decreased in both groups, especially in the observation group ( $P<0.05$ ). After hemodialysis, transforming growth factor (TGF)-β1 was 7.76 ng/mL in the observation group and 7.60 ng/mL in the control group, which was significantly reduced in both groups. After treatment, UA and Scr in both groups were significantly reduced, and the reduction was more significant in the observation group ( $P<0.05$ ). HGB and RBC were significantly increased in both groups, and the increase was more significant in the observation group ( $P<0.05$ ). In summary, ACM algorithm had a good segmentation effect on the ultrasonic images of children with renal failure. This study provided guidance for clinicians to choose the algorithm for the application of ultrasonic imaging diagnosis.