Segmentation of liver and liver lesions using deep learning.

Segmentation of organs and lesions could be employed for the express purpose of dosimetry in nuclear medicine, assisted image interpretations, and mass image processing studies. Deep leaning created liver and liver lesion segmentation on clinical 3D MRI data has not been fully addressed in previous experiments. To this end, the required data were collected from 128 patients, including their T1w and T2w MRI images, and ground truth labels of the liver and liver lesions were generated. The collection of 110 T1w-T2w MRI image sets was divided, with 94 designated for training and 16 for validation. Furthermore, 18 more datasets were separately allocated for use as hold-out test datasets. The T1w and T2w MRI images were preprocessed into a two-channel format so that they were used as inputs to the deep learning model based on the Isensee 2017 network. To calculate the final Dice coefficient of the network performance on test datasets, the binary average of T1w and T2w predicted images was used. The deep learning model could segment all 18 test cases, with an average Dice coefficient of 88% for the liver and 53% for the liver tumor. Liver segmentation was carried out with rather a high accuracy; this could be achieved for liver dosimetry during systemic or selective radiation therapies as well as for attenuation correction in PET/MRI scanners. Nevertheless, the delineation of liver lesions was not optimal; therefore, tumor detection was not practical by the proposed method on clinical data.

Generalizability assessment of COVID-19 3D CT data for deep learning-based disease detection.

BACKGROUND: Artificial intelligence technologies in classification/detection of COVID-19 positive cases suffer from generalizability. Moreover, accessing and preparing another large dataset is not always feasible and time-consuming. Several studies have combined smaller COVID-19 CT datasets into "supersets" to maximize the number of training samples. This study aims to assess generalizability by splitting datasets into different portions based on 3D CT images using deep learning. METHOD: Two large datasets, including 1110 3D CT images, were split into five segments of 20% each. Each dataset's first 20% segment was separated as a holdout test set. 3D-CNN training was performed with the remaining 80% from each dataset. Two small external datasets were also used to independently evaluate the trained models. RESULTS: The total combination of 80% of each dataset has an accuracy of 91% on Iranmehr and 83% on Moscow holdout test datasets. Results indicated that 80% of the primary datasets are adequate for fully training a model. The additional fine-tuning using 40% of a secondary dataset helps the model generalize to a third, unseen dataset. The highest accuracy achieved through transfer learning was 85% on LDCT dataset and 83% on Iranmehr holdout test sets when retrained on 80% of Iranmehr dataset. CONCLUSION: While the total combination of both datasets produced the best results, different combinations and transfer learning still produced generalizable results. Adopting the proposed methodology may help to obtain satisfactory results in the case of limited external datasets.

Factors determining generalization in deep learning models for scoring COVID-CT images.

The COVID-19 pandemic has inspired unprecedented data collection and computer vision modelling efforts worldwide, focused on the diagnosis of COVID-19 from medical images. However, these models have found limited, if any, clinical application due in part to unproven generalization to data sets beyond their source training corpus. This study investigates the generalizability of deep learning models using publicly available COVID-19 Computed Tomography data through cross dataset validation. The predictive ability of these models for COVID-19 severity is assessed using an independent dataset that is stratified for COVID-19 lung involvement. Each inter-dataset study is performed using histogram equalization, and contrast limited adaptive histogram equalization with and without a learning Gabor filter. We show that under certain conditions, deep learning models can generalize well to an external dataset with F1 scores up to 86%. The best performing model shows predictive accuracy of between 75% and 96% for lung involvement scoring against an external expertly stratified dataset. From these results we identify key factors promoting deep learning generalization, being primarily the uniform acquisition of training images, and secondly diversity in CT slice position.

An Impressive Approach in Nuclear Medicine: Theranostics.

Radiometal-based theranostics or theragnostics, first used in the early 2000s, is the combined application of diagnostic and therapeutic agents that target the same molecule, and represents a considerable advancement in nuclear medicine. One of the promising fields related to theranostics is radioligand therapy. For instance, the concepts of targeting the prostate-specific membrane antigen (PSMA) for imaging and therapy in prostate cancer, or somatostatin receptor targeted imaging and therapy in neuroendocrine tumors (NETs) are part of the field of theranostics. Combining targeted imaging and therapy can improve prognostication, therapeutic decision-making, and monitoring of the therapy.

Internal dosimetry for radioembolization therapy with Yttrium-90 microspheres.

The absorbed doses in the liver and adjacent viscera in Yttrium-90 radioembolization therapy for metastatic liver lesions are not well-documented. We sought for a clinically practical way to determine the dosimetry of this advent treatment. Six different female XCAT BMIs and seven different male XCAT BMIs were generated. Using Monte Carlo GATE code simulation, the total of 100MBq 90 Y was deposited uniformly in the source organ, liver. Self-irradiation and absorbed doses in lung, kidney and bone marrow were calculated. The mean energy of Yittrium-90 (i.e., 0.937 MeV) was used. The S-values and equivalent doses in target organs were estimated. The dose absorbed in the liver was between 84 and 53 Gy and below the target of 80 to 150 Gy. The absorbed dose in the bone marrow, lungs, and kidneys are very low and below 0.1 , 0.4, and 0.5 Gy respectively. Our study indicates that larger activities than the conventional dose of 3 GBq may be both required and safe. Further confirmations in clinical settings are needed.

The importance of BMI in dosimetry of 153Sm-EDTMP bone pain palliation therapy: A Monte Carlo study.

Using digital phantoms as an atlas compared to acquiring CT data for internal radionuclide dosimetry decreases patient overall radiation dose and reduces the required analysis effort and time for organ segmentation. The drawback is that the phantom may not match exactly with the patient. We assessed the effect of varying BMIs on dosimetry results for a bone pain palliation agent, 153Sm-EDTMP. The simulation was done using the GATE Monte Carlo code. Female XCAT phantoms with the following different BMIs were employed: 18.6, 20.8, 22.1, 26.8, 30.3 and 34.7kg/m2. S-factors (mGy/MBq.s) and SAFs (kg-1) were calculated for the dosimetry of the radiation from major source organs including spine, ribs, kidney and bladder into different target organs as well as whole body dosimetry from spine. The differences in dose estimates from different phantoms compared to those from the phantom with BMI of 26.8kg/m2 as the reference, were calculated for both gamma and beta radiations. The relative differences (RD) of the S-factors or SAFs from the values of reference phantom were calculated. RDs greater than 10% and 100% were frequent in radiations to organs for photon and beta particles, respectively. The relative differences in whole body SAFs from the reference phantom were 15.4%, 7%, 4.2%, -9.8% and -1.4% for BMIs of 18.6, 20.8, 22.1, 30.3 and 34.7kg/m2, respectively. The differences in whole body S-factors for the phantoms with BMIs of 18.6, 20.8, 22.1, 30.3 and 34.7kg/m2 were 39.5%, 19.4%, 8.8%, -7.9% and -4.3%, respectively. The dosimetry of the gamma photons and beta particles changes substantially with the use of phantoms with different BMIs. The change in S-factors is important for dose calculation and can change the prescribed therapeutic dose of 153Sm-EDTMP. Thus a phantom with BMI better matched to the patient is suggested for therapeutic purposes where dose estimates closer to those in the actual patient are required.

Practical Nuclear Medicine and Utility of Phantoms for Internal Dosimetry: XCAT Compared with Zubal.

PURPOSE: The absorbed doses for two radioisotopes, 99mTc and 131I, between previously validated Zubal phantom and the recently developed XCAT phantom were compared. MATERIALS AND METHODS: GATE Monte Carlo code was used to simulate the statistical process of radiation. A XCAT phantom with voxel and matrix sizes similar to a standard Zubal phantom was generated. According to Medical International Radiation Dose formalism, specific absorbed fraction (SAF) values for photons and S-factors for beta particles were tabulated. The amounts of absorbed doses were calculated and compared in different organs. RESULTS: The differences of gamma radiation doses, SAFs, between Zubal and XCAT are >50% in adrenal from adrenal, pancreas from pancreas and thyroid from thyroid, in lung from kidney, kidneys from lungs and in kidneys from thyroid and thyroid from kidneys. The beta radiation doses differences between Zubal and XCAT are >50% in thyroid from thyroid, bladder from bladder, kidney from kidney, liver from bladder, thyroid from bladder and kidney from thyroid. The size and distances of the organs were different between XCAT and Zubal phantoms. Denoted differences of SAF and S-factors correspond to the different organ geometries in phantoms. CONCLUSION: The results of absorbed doses in Zubal and XCAT phantoms are different. The variations prohibit easy comparison or interchangeability of dosimetry between these phantoms.