Three-stage segmentation of lung region from CT images using deep neural networks.

BACKGROUND: Lung region segmentation is an important stage of automated image-based approaches for the diagnosis of respiratory diseases. Manual methods executed by experts are considered the gold standard, but it is time consuming and the accuracy is dependent on radiologists' experience. Automated methods are relatively fast and reproducible with potential to facilitate physician interpretation of images. However, these benefits are possible only after overcoming several challenges. The traditional methods that are formulated as a three-stage segmentation demonstrate promising results on normal CT data but perform poorly in the presence of pathological features and variations in image quality attributes. The implementation of deep learning methods that can demonstrate superior performance over traditional methods is dependent on the quantity, quality, cost and the time it takes to generate training data. Thus, efficient and clinically relevant automated segmentation method is desired for the diagnosis of respiratory diseases. METHODS: We implement each of the three stages of traditional methods using deep learning methods trained on five different configurations of training data with ground truths obtained from the 3D Image Reconstruction for Comparison of Algorithm Database (3DIRCAD) and the Interstitial Lung Diseases (ILD) database. The data was augmented with the Lung Image Database Consortium (LIDC-IDRI) image collection and a realistic phantom. A convolutional neural network (CNN) at the preprocessing stage classifies the input into lung and none lung regions. The processing stage was implemented using a CNN-based U-net while the postprocessing stage utilize another U-net and CNN for contour refinement and filtering out false positives, respectively. RESULTS: The performance of the proposed method was evaluated on 1230 and 1100 CT slices from the 3DIRCAD and ILD databases. We investigate the performance of the proposed method on five configurations of training data and three configurations of the segmentation system; three-stage segmentation and three-stage segmentation without a CNN classifier and contrast enhancement, respectively. The Dice-score recorded by the proposed method range from 0.76 to 0.95. CONCLUSION: The clinical relevance and segmentation accuracy of deep learning models can improve though deep learning-based three-stage segmentation, image quality evaluation and enhancement as well as augmenting the training data with large volume of cheap and quality training data. We propose a new and novel deep learning-based method of contour refinement.

Pre-dosing with lilotomab prior to therapy with 177Lu-lilotomab satetraxetan significantly increases the ratio of tumor to red marrow absorbed dose in non-Hodgkin lymphoma patients.

PURPOSE: 177Lu-lilotomab satetraxetan is a novel anti-CD37 antibody radionuclide conjugate for the treatment of non-Hodgkin lymphoma (NHL). Four arms with different combinations of pre-dosing and pre-treatment have been investigated in a first-in-human phase 1/2a study for relapsed CD37+ indolent NHL. The aim of this work was to determine the tumor and normal tissue absorbed doses for all four arms, and investigate possible variations in the ratios of tumor to organs-at-risk absorbed doses. METHODS: Two of the phase 1 arms included cold lilotomab pre-dosing (arm 1 and 4; 40 mg fixed and 100 mg/m2 BSA dosage, respectively) and two did not (arms 2 and 3). All patients were pre-treated with different regimens of rituximab. The patients received either 10, 15, or 20 MBq 177Lu-lilotomab satetraxetan per kg body weight. Nineteen patients were included for dosimetry, and a total of 47 lesions were included. The absorbed doses were calculated from multiple SPECT/CT-images and normalized by administered activity for each patient. Two-sided Student's t tests were used for all statistical analyses. RESULTS: Organs with distinct uptake of 177Lu-lilotomab satetraxetan, in addition to tumors, were red marrow (RM), liver, spleen, and kidneys. The mean RM absorbed doses were 0.94, 1.55, 1.44, and 0.89 mGy/MBq for arms 1-4, respectively. For the patients not pre-dosed with lilotomab (arms 2 and 3 combined) the mean RM absorbed dose was 1.48 mGy/MBq, which was significantly higher than for both arm 1 (p = 0.04) and arm 4 (p = 0.02). Of the other organs, the highest uptake was found in the spleen, and there was a significantly lower spleen absorbed dose for arm-4 patients than for the patient group without lilotomab pre-dosing (1.13 vs. 3.20 mGy/MBq; p < 0.01). Mean tumor absorbed doses were 2.15, 2.31, 1.33, and 2.67 mGy/MBq for arms 1-4, respectively. After averaging the tumor absorbed dose for each patient, the patient mean tumor absorbed dose to RM absorbed dose ratios were obtained, given mean values of 1.07 for the patient group not pre-dosed with lilotomab, of 2.16 for arm 1, and of 4.62 for arm 4. The ratios were significantly higher in both arms 1 and 4 compared to the group without pre-dosing (p = 0.05 and p = 0.02). No statistically significant difference between arms 1 and 4 was found. CONCLUSIONS: RM is the primary dose-limiting organ for 177Lu-lilotomab satetraxetan treatment, and pre-dosing with lilotomab has a mitigating effect on RM absorbed dose. Increasing the amount of lilotomab from 40 mg to 100 mg/m2 was found to slightly decrease the RM absorbed dose and increase the ratio of tumor to RM absorbed dose. Still, both pre-dosing amounts resulted in significantly higher tumor to RM absorbed dose ratios. The findings encourage continued use of pre-dosing with lilotomab.

Tumor-Absorbed Dose for Non-Hodgkin Lymphoma Patients Treated with the Anti-CD37 Antibody Radionuclide Conjugate 177Lu-Lilotomab Satetraxetan.

177Lu-lilotomab satetraxetan is a novel antibody radionuclide conjugate currently tested in a phase 1/2a first-in-human dosage escalation trial for patients with relapsed CD37+ indolent non-Hodgkin lymphoma. The aim of this work was to develop dosimetric methods and calculate tumor-absorbed radiation doses for patients treated with 177Lu-lilotomab satetraxetan. METHODS: Patients were treated at escalating injected activities (10, 15 and 20 MBq/kg) of 177Lu-lilotomab satetraxetan and with different predosing, with or without 40 mg of unlabeled lilotomab. Eight patients were included for the tumor dosimetry study. Tumor radioactivity concentrations were calculated from SPECT acquisitions at multiple time points, and tumor masses were delineated from corresponding CT scans. Tumor-absorbed doses were then calculated using the OLINDA sphere model. To perform voxel dosimetry, the SPECT/CT data and an in-house-developed MATLAB program were combined to investigate the dose rate homogeneity. RESULTS: Twenty-six tumors in 8 patients were ascribed a mean tumor-absorbed dose. Absorbed doses ranged from 75 to 794 cGy, with a median of 264 cGy across different dosage levels and different predosing. A significant correlation between the dosage level and tumor-absorbed dose was found. Twenty-one tumors were included for voxel dosimetry and parameters describing dose-volume coverage calculated. The investigation of intratumor voxel doses indicates that mean tumor dose is correlated to these parameters. CONCLUSION: Tumor-absorbed doses for patients treated with 177Lu-lilotomab satetraxetan are comparable to doses reported for other radioimmunotherapy compounds. Although the intertumor variability was considerable, a correlation between tumor dose and patient dosage level was found. Our results indicate that mean dose may be used as the sole dosimetric parameter on the lesion level.