Maximum Intensity Projection Images Research Articles

18F-FDG PET Maximum-Intensity Projections and Artificial Intelligence: A Win-Win Combination to Easily Measure Prognostic Biomarkers in DLBCL Patients.

Total metabolic tumor volume (TMTV) and tumor dissemination (Dmax) calculated from baseline 18F-FDG PET/CT images are prognostic biomarkers in diffuse large B-cell lymphoma (DLBCL) patients. Yet, their automated calculation remains challenging. The purpose of this study was to investigate whether TMTV and Dmax features could be replaced by surrogate features automatically calculated using an artificial intelligence (AI) algorithm from only 2 maximum-intensity projections (MIPs) of the whole-body 18F-FDG PET images. Methods: Two cohorts of DLBCL patients from the REMARC (NCT01122472) and LNH073B (NCT00498043) trials were retrospectively analyzed. Experts delineated lymphoma lesions from the baseline whole-body 18F-FDG PET/CT images, from which TMTV and Dmax were measured. Coronal and sagittal MIP images and associated 2-dimensional reference lesion masks were calculated. An AI algorithm was trained on the REMARC MIP data to segment lymphoma regions. The AI algorithm was then used to estimate surrogate TMTV (sTMTV) and surrogate Dmax (sDmax) on both datasets. The ability of the original and surrogate TMTV and Dmax to stratify patients was compared. Results: Three hundred eighty-two patients (mean age ± SD, 62.1 y ± 13.4 y; 207 men) were evaluated. sTMTV was highly correlated with TMTV for REMARC and LNH073B datasets (Spearman r = 0.878 and 0.752, respectively), and so were sDmax and Dmax (r = 0.709 and 0.714, respectively). The hazard ratios for progression free survival of volume and MIP-based features derived using AI were similar, for example, TMTV: 11.24 (95% CI: 2.10-46.20), sTMTV: 11.81 (95% CI: 3.29-31.77), and Dmax: 9.0 (95% CI: 2.53-23.63), sDmax: 12.49 (95% CI: 3.42-34.50). Conclusion: Surrogate TMTV and Dmax calculated from only 2 PET MIP images are prognostic biomarkers in DLBCL patients and can be automatically estimated using an AI algorithm.

PSMA PET tumor-to-salivary glands ratio (PSG score) to predict response to Lu-177 PSMA radioligand therapy: An international multicenter retrospective study.

5043 Background: PSMA-targeted radioligand therapy can improve the outcome of patients with advanced mCRPC. However, patients do not respond uniformly and a PSA decline of ≥50% (PSA50) was achieved only in 46% of the patients in the VISION trial. We hypothesized that using the parotid glands instead of the liver as the reference organ enables more selective stratification. The aim of this study was to test PSMA PET tumor-to-Salivary Glands ratio (PSG score) to predict outcomes after Lu-177 PSMA. Methods: This was an international multicenter retrospective study using the established dataset consisting of 270 men with mCRPC treated with Lu-177 PSMA (Gafita A, Lancet Oncol 2021). First, we assessed baseline PSMA PET quantitatively (qPSG score) to calculate the tumor-to-salivary gland ratio (qPSG = SUVmean whole-body-tumor / SUVmean parotid glands) using a semi-automatic segmentation software (qPSMA). Patients were divided into three groups: high (qPSG > 1.5), intermediate (qPSG = 0.5 - 1.5), and low (qPSG < 0.5). Second, we assessed the reproducibility and the predictive value of the PSG score visually (vPSG score) graded by ten nuclear medicine physicians. Each reader read the baseline PSMA PET 3D maximum intensity projection (MIP) images, and classified the patients into three groups: (high) most of the lesions (> 80%) show higher uptake than parotid glands; (intermediate) neither “low” nor “high”; (low) most of the lesions (> 80%) show lower uptake than parotid glands. In case of disagreement, a majority vote was used. Outcome measures were PSA-progression free-survival (PSA-PFS), overall survival (OS), and PSA50. Results: 237 men were analyzed after excluding 33 men whose parotid glands were out of the scan range. The number of the patients in the high, intermediate, and low groups were 56/237 (23.6%), 163 (68.8%), and 18 (7.6%) for qPSG score, and 106/237 (44.7%), 96 (40.5%), and 35 (14.8%) for vPSG score, respectively. The inter- and intra-readers reproducibility of the vPSG score showed substantial (Fleiss’ weighted Kappa: 0.68) and almost perfect (Cohen's weighted Kappa (mean): 0.83) agreement, respectively. The median PSA-PFS of high, intermediate, and low groups were 7.2, 4.0, and 1.9 months (p < 0.001) for qPSG score and 6.7, 3.8, and 1.9 months (p < 0.001) for vPSG score, respectively. Higher PSA50 rate was observed in the high group followed by the intermediate and low groups (high vs intermediate vs low: [qPSG] 69.6% vs 38.7% vs 16.7%; [vPSG] 63.2% vs 33.3% vs 16.1%). The median OS was longer in the high group than in the intermediate + low (i.e., non-high) group by qPSG (15.0 vs. 11.7 months (p = 0.013)) and vPSG score (14.3 vs.11.0 months (p = 0.038)). Conclusions: The PSG score is a valuable predictive biomarker for response to Lu-177 PSMA. The vPSG score yielded substantial reproducibility and comparable predictive value to the qPSG score.

Using deep learning to safely exclude lesions with only ultrafast breast MRI to shorten acquisition and reading time

ObjectivesTo investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload.MethodsIn this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated.ResultsThe independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75–0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%.ConclusionThis deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists’ workload and scan time.Key Points• Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening.• Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.

Usefulness of maximum intensity projection images of non-enhanced CT for detection of hyperdense middle cerebral artery sign in acute thromboembolic ischemic stroke

PurposeTo compare the sensitivity of the hyperdense middle cerebral artery (MCA) sign between maximum intensity projection (MIP) and conventional averaged images in patients with acute focal neurological deficits with acute thromboembolic MCA occlusion (MCA occlusion group) and patients with acute focal neurological deficits without MCA occlusion (control group).Materials and methodsInitial computed tomography (CT) scans on admission were reconstructed with 5 mm thickness at every 3 mm interval for averaged and MIP images from 1 mm thickness non-contrast axial source images. Images were obtained from 30 cases each in the MCA occlusion and control groups. The CT values in the region of interests (ROIs) on the affected and unaffected sides of the MCA were compared. To compare CT values among subjects, the CT values were normalized by obtaining a ratio on the affected and unaffected sides, and the normalized CT values were analyzed using the receiver operating characteristic (ROC) curve.ResultsThe hyperdense MCA sign was visually detected on MIP images in 90% cases and on 5 mm averaged images in only 57% cases in the MCA occlusion group. Based on the ROC analysis of the normalized ratio on the affected and unaffected sides, area under the curve of MIP image and averaged image was 0.941 and 0.655, respectively. On MIP images, the optimal threshold of the ratio on the affected and unaffected sides was 1.152 (sensitivity: 90.0%, and specificity: 93.3%).ConclusionThe hyperdense MCA sign sensitivity on 5 mm MIP images was significantly higher than that on conventional 5 mm averaged CT images. This could be useful for the early initiation of proper therapy for patients with acute focal neurological deficits.

Diagnostic value of three-dimensional cube fluid attenuated inversion recovery imaging and its axial MIP reconstruction in multiple sclerosis

BackgroundMagnetic resonance imaging is regarded as one of the most important markers for multiple sclerosis. It can detect lesions in order to establish dissemination in time and space, which would aid in the diagnosis. Two-dimensional FLAIR is a standard sequence in MS routine imaging because it suppresses cerebrospinal fluid signal, increasing contrast between lesions and CSF and improving white matter lesion detection. Newer 3D FLAIR sequences are expected to offer even more benefits, such as improved MS lesions detection and higher resolution due to thinner slice thickness. We aimed to compare the role of 3D Cube FLAIR imaging (versus standard 2D FLAIR) in the assessment of white matter lesions in MS patients, as well as to test the convenience of using maximum intensity projection (MIP) on 3D FLAIR images for faster and easier evaluation.ResultsThis study included 160 MS patients. A 1.5 T routine brain MRI scan was performed, which included a 2D FLAIR sequence, followed by a 3D-FLAIR sequence. All images were analyzed after 3D-FLAIR images were reformatted into axial MIP images. Lesions were counted in each sequence and classified into supra-tentorial (periventricular, deep white matter, and juxta-cortical), and infra-tentorial lesions, with the relative comparison of lesions numbers on 3D-FLAIR and MIP versus 2D-FLAIR expressed as a percentage increase or decrease. 3D FLAIR can significantly improve MS lesion detection in all areas of the brain when compared with 2D FLAIR results. At 2 mm reformatting, there is no difference in MS lesion detection between sagittal 3D FLAIR and axial MIP reconstruction, implying that the MIP algorithm can be used to simplify lesion detection by reducing the number of images while maintaining the same level of reliability.Conclusion3D FLAIR sequences should be added to conventional 2D FLAIR sequences in the MRI protocol when MS is suspected.

Weakly Supervised Tumor Detection in PET Using Class Response for Treatment Outcome Prediction.

It is proven that radiomic characteristics extracted from the tumor region are predictive. The first step in radiomic analysis is the segmentation of the lesion. However, this task is time consuming and requires a highly trained physician. This process could be automated using computer-aided detection (CAD) tools. Current state-of-the-art methods are trained in a supervised learning setting, which requires a lot of data that are usually not available in the medical imaging field. The challenge is to train one model to segment different types of tumors with only a weak segmentation ground truth. In this work, we propose a prediction framework including a 3D tumor segmentation in positron emission tomography (PET) images, based on a weakly supervised deep learning method, and an outcome prediction based on a 3D-CNN classifier applied to the segmented tumor regions. The key step is to locate the tumor in 3D. We propose to (1) calculate two maximum intensity projection (MIP) images from 3D PET images in two directions, (2) classify the MIP images into different types of cancers, (3) generate the class activation maps through a multitask learning approach with a weak prior knowledge, and (4) segment the 3D tumor region from the two 2D activation maps with a proposed new loss function for the multitask. The proposed approach achieves state-of-the-art prediction results with a small data set and with a weak segmentation ground truth. Our model was tested and validated for treatment response and survival in lung and esophageal cancers on 195 patients, with an area under the receiver operating characteristic curve (AUC) of 67% and 59%, respectively, and a dice coefficient of 73% and 0.77% for tumor segmentation.

Breast Tumor Identification in Ultrafast MRI Using Temporal and Spatial Information.

Simple SummaryThe diagnosis of breast cancer with MRI is based on both morphological evaluation and kinetic curve assessment. Current computer-aided diagnosis methods for malignancy determination mainly focus on morphology features but ignored the temporal information in dynamic contrast-enhanced MRI sequences. Malignant and benign lesions usually have different enhancement patterns during the wash-in phase. Ultrafast breast MRI with high temporal resolution can capture the inflow of contrast in breast lesions. This advantage of ultrafast MRI enables the combination of both temporal and spatial information for automatic breast lesion analysis model development. We found that temporal information helps to significantly improve the performance of breast lesion classification. This suggests that ultrafast MRI provides useful information for malignancy identification and temporal information, which is indispensable for similar model development.Purpose: To investigate the feasibility of using deep learning methods to differentiate benign from malignant breast lesions in ultrafast MRI with both temporal and spatial information. Methods: A total of 173 single breasts of 122 women (151 examinations) with lesions above 5 mm were retrospectively included. A total of 109 out of 173 lesions were benign. Maximum intensity projection (MIP) images were generated from each of the 14 contrast-enhanced T1-weighted acquisitions in the ultrafast MRI scan. A 2D convolutional neural network (CNN) and a long short-term memory (LSTM) network were employed to extract morphological and temporal features, respectively. The 2D CNN model was trained with the MIPs from the last four acquisitions to ensure the visibility of the lesions, while the LSTM model took MIPs of an entire scan as input. The performance of each model and their combination were evaluated with 100-times repeated stratified four-fold cross-validation. Those models were then compared with models developed with standard DCE-MRI which followed the same data split. Results: In the differentiation between benign and malignant lesions, the ultrafast MRI-based 2D CNN achieved a mean AUC of 0.81 ± 0.06, and the LSTM network achieved a mean AUC of 0.78 ± 0.07; their combination showed a mean AUC of 0.83 ± 0.06 in the cross-validation. The mean AUC values were significantly higher for ultrafast MRI-based models than standard DCE-MRI-based models. Conclusion: Deep learning models developed with ultrafast breast MRI achieved higher performances than standard DCE-MRI for malignancy discrimination. The improved AUC values of the combined models indicate an added value of temporal information extracted by the LSTM model in breast lesion characterization.

Reproducibility for Hepatocellular Carcinoma CT Radiomic Features: Influence of Delineation Variability Based on 3D-CT, 4D-CT and Multiple-Parameter MR Images.

PurposeAccurate lesion segmentation is a prerequisite for radiomic feature extraction. It helps to reduce the features variability so as to improve the reporting quality of radiomics study. In this research, we aimed to conduct a radiomic feature reproducibility test of inter-/intra-observer delineation variability in hepatocellular carcinoma using 3D-CT images, 4D-CT images and multiple-parameter MR images.Materials and MethodsFor this retrospective study, 19 HCC patients undergoing 3D-CT, 4D-CT and multiple-parameter MR scans were included in this study. The gross tumor volume (GTV) was independently delineated twice by two observers based on contrast-enhanced computed tomography (CECT), maximum intensity projection (MIP), LAVA-Flex, T2W FRFSE and DWI-EPI images. We also delineated the peritumoral region, which was defined as 0 to 5 mm radius surrounding the GTV. 107 radiomic features were automatically extracted from CECT images using 3D-Slicer software. Quartile coefficient of dispersion (QCD) and intraclass correlation coefficient (ICC) were applied to assess the variability of each radiomic feature. QCD<10% and ICC≥0.75 were considered small variations and excellent reliability. Finally, the principal component analysis (PCA) was used to test the feasibility of dimensionality reduction.ResultsFor tumor tissues, the numbers of radiomic features with QCD<10% indicated no obvious inter-/intra-observer differences or discrepancies in 3D-CT, 4D-CT and multiple-parameter MR delineation. However, the number of radiomic features (mean 89) with ICC≥0.75 was the highest in the multiple-parameter MR group, followed by the 3DCT group (mean 77) and the MIP group (mean 73). The peritumor tissues also showed similar results. A total of 15 and 7 radiomic features presented excellent reproducibility and small variation in tumor and peritumoral tissues, respectively. Two robust features showed excellent reproducibility and small variation in tumor and peritumoral tissues. In addition, the values of the two features both represented statistically significant differences among tumor and peritumoral tissues (P<0.05). The PCA results indicated that the first seven principal components could preserve at least 90% of the variance of the original set of features.ConclusionDelineation on multiple-parameter MR images could help to improve the reproducibility of the HCC CT radiomic features and weaken the inter-/intra-observer influence.

Effect of urinary glucose concentration and pH on signal intensity in magnetic resonance images.

PurposeWith advances in anti-diabetes drugs, increasing numbers of patients have high urinary glucose concentrations, which may alter magnetic resonance (MR) signal intensity. We sought to elucidate the effect of urinary glucose concentration and pH on transverse relaxation and MR signal intensity.Materials and methodsThe transverse relaxation rate (R2) was measured in samples with different glucose concentrations (in vitro) and in the urinary bladder of seven patients with diabetes and nine healthy volunteers (in vivo). The glucose concentration and pH in the in vitro samples and urine were measured. The signal intensity ratio of the bladder to adjacent tissues was obtained on T2-weighted imaging (WI), T1WI, and MR urography (in vivo). To clarify the effect of pH further, the urine of two healthy subjects was adjusted with acid and/or base to obtain various pH values (ex vivo).ResultsR2 increased significantly with high glucose concentrations in the in vitro study. In the in vivo study, high glucose concentration (p < 0.001) and low pH (p = 0.005) were significantly associated with high R2. R2 was higher (p = 0.002) and the signal in maximum-intensity projection images of MR urography was lower (p = 0.005) in patients with diabetes than in healthy subjects. Ex vivo study revealed that a decrease in pH in acid portion resulted in increased R2.ConclusionHigh concentrations of urinary glucose and low pH both enhance transverse relaxation, which, in turn, causes low signal intensity in urinary bladder on long echo time (TE) images, such as MR urography. Radiologists should be aware of this phenomenon when interpreting abnormally low-intensity bladders on long TE images.

Panorama of multidetector-row computed tomography findings of carcinoma gall bladder - A retrospective observational study.

Carcinoma of the gall bladder (Ca GB) has marked ethnic and geographical variations with a broad spectrum of imaging findings on multidetector-row computed tomography (MDCT). We aim to study the spectrum of the imaging findings of Ca GB on MDCT in an Eastern Indian hospital as these imaging findings help in accurate diagnosis and staging of this lethal disease. The MDCT images of 100 biopsy-proven cases of adenocarcinoma of GB performed from January 1, 2017, to December 31, 2018, in our hospital were retrospectively reviewed by two experienced radiologists blinded to the diagnosis. The CT protocol was a plain scan followed by triple-phase contrast-enhanced CT. Reconstructed images in the form of maximum intensity projection, volume-rendered technology, and minimum intensity projection images were also assessed. Descriptive statistics were used for data analysis. Ca GB showed a female predominance (female:male -1.9:1). The mean age of presentation was 54.7 years (females) and 58 years (males). The morphological patterns were a mass replacing the GB (46%), focal or diffuse wall thickening (26%), and an intraluminal polypoidal mass (28%). Direct extension was to liver (76%), duodenum (32%), colon (19%), pancreas (9%), and abdominal wall (1%). Biliary dilatation (53%), vascular invasion (14%), omental involvement (23%), ascites (22%), N1 (40%), and retroperitoneal lymphadenopathy (33%) were common. Distant metastasis comprised of hepatic (42%), pulmonary (7%), Krukenberg's tumor (6%), and osseous (1%) lesions. The stages at diagnosis were I or II (3%), IIIA (4%), IIIB (16%), IVA (10%), and IVB (67%). Ca GB has a broad spectrum of findings on MDCT and it mostly presents at an advanced stage. MDCT with reconstructions is beneficial in the assessment of locoregional and distant spread and cancer staging which has a direct implication on patient management, survival, and mortality.

An artificial intelligence system using maximum intensity projection MR images facilitates classification of non-mass enhancement breast lesions.

To build an artificial intelligence (AI) system to classify benign and malignant non-mass enhancement (NME) lesions using maximum intensity projection (MIP) of early post-contrast subtracted breast MR images. This retrospective study collected 965 pure NME lesions (539 benign and 426 malignant) confirmed by histopathology or follow-up in 903 women. The 754 NME lesions acquired by one MR scanner were randomly split into the training set, validation set, and test set A (482/121/151 lesions). The 211 NME lesions acquired by another MR scanner were used as test set B. The AI system was developed using ResNet-50 with the axial and sagittal MIP images. One senior and one junior radiologist reviewed the MIP images of each case independently and rated its Breast Imaging Reporting and Data System category. The performance of the AI system and the radiologists was evaluated using the area under the receiver operating characteristic curve (AUC). The AI system yielded AUCs of 0.859 and 0.816 in the test sets A and B, respectively. The AI system achieved comparable performance as the senior radiologist (p = 0.558, p = 0.041) and outperformed the junior radiologist (p < 0.001, p = 0.009) in both test sets A and B. After AI assistance, the AUC of the junior radiologist increased from 0.740 to 0.862 in test set A (p < 0.001) and from 0.732 to 0.843 in test set B (p < 0.001). Our MIP-based AI system yielded good applicability in classifying NME lesions in breast MRI and can assist the junior radiologist achieve better performance. • Our MIP-based AI system yielded good applicability in the dataset both from the same and a different MR scanner in predicting malignant NME lesions. • The AI system achieved comparable diagnostic performance with the senior radiologist and outperformed the junior radiologist. • This AI system can assist the junior radiologist achieve better performance in the classification of NME lesions in MRI.

Radiological evaluation of maxillary artery and descending palatine artery in the pterygopalatine fossa by 3D rotational angiography.

The aim of this study is to evaluate the branching patterns and topographical features of the third part of the maxillary artery (t-MA) and descending palatine artery (DPA) by 3-Dimensional Rotational Angiography (3DRA) images and to define the radiological classification of their variations, based upon the previous cadaveric studies and a review of the literature. This study was conducted from May 2020 through June 2021. All consecutive adult patients who were examined with 3D-RA were enrolled in the study. The morphological evaluations and measurements of t-MA, DPA and their branches were made on maximum intensity projection images with 10-20mm slice thickness. Eighty-five hemifaces, including 58 females and 45 right sides, were evaluated. The diameter of the t-MA was measured as 1.73 ± 0.30mm. The most common pattern of the t-MA according to its course was loop type (63/85, 74.1%) and according to branching pattern was Type Ib (29/85, 34.1%). The mean diameter of DPA was 1.19 ± 0.20mm. The DPA presented as a single trunk in 11/85 cases. Type II, which was defined as one lesser palatine artery originating from distal-DPA, was the most common morphological variation (51.8%). 3DRA imaging provides valuable information for vascular anatomical studies. The most common morphological variation related to t-MA, DPA is the distal branching pattern.

Predicting the outcome of mCPRC patients after Lu-177 PSMA therapy using semi-quantitative and visual criteria in baseline PSMA PET: An international multicenter retrospective study.

32 Background: The aim of this study was to assess the use of the baseline PSMA parotids uptake as a reference to determine eligibility for Lu-177 PSMA radioligand therapy (RLT) by using semi-quantitative and standardized visual criteria. Methods: We conducted a retrospective cohort study using a multicenter dataset 270 men with mCRPC treated with Lu-177 PSMA (Gafita A, Lancet Oncol 2021). For quantitative analysis, semi-automatic segmentation software (qPSMA) divided men into three groups according to the SUVmean ratio of whole-body-tumor to parotid glands: (high) &gt; 1.5; (intermediate) 0.5 - 1.5; (low) &lt; 0.5. For visual analysis, ten nuclear medicine physicians with (n = 5) and without (n = 5) clinical experience in Lu-177 PSMA RLT ( &gt; 50 cases) read each baseline PSMA PET 3D maximum intensity projection (MIP) images, and classified the patients into three groups: (high) most ( &gt; 80%) of the lesions show higher uptake than parotid glands; (intermediate) neither “low” nor “high”; (low) most ( &gt; 80%) of the lesions show lower uptake than parotid glands. In case of inter-reader disagreement, a majority vote was used. Outcome measures included PSA-progression free-survival (PSA-PFS), overall survival (OS), and PSA decline of ≥50% (PSA50). Fisher’s exact test and Kaplan–Meier analysis with the log-rank test was performed for PSA50 and survival analysis, respectively. Results: 237 men were analyzed after excluding 33 men with more than half of the parotid glands out of the PET scan field-of-view. The number of the patients in the high, intermediate, and low groups were 106/237 (44.7%), 96 (40.5%), and 35 (14.8%) for visual criteria, and 56 (23.6%), 163 (68.8%), and 18 (7.6%) for quantitative analysis, respectively. The inter- and intra-readers reproducibility of the visual scoring showed substantial (Fleiss’ weighted Kappa: 0.68) and almost perfect (Cohen's weighted Kappa (mean): 0.83) agreement, respectively. The median PSA-PFS was 6.7, 3.8, and 1.9 months (p &lt; 0.001); and 7.2, 4.0, and 1.9 months (p &lt; 0.001) in the high, intermediate, and low expression groups by visual and quantitative criteria, respectively. The PSA50 was 63.2%, 33.3%, and 17.1% (p &lt; 0.001), and 69.6%, 38.7%, and 16.7% (p &lt; 0.001) in the high, intermediate, and low expression groups by visual and quantitative criteria, respectively. OS was longer in the high PSMA expression group than in the intermediate + low (i.e., non-high) group by visual (14.3 vs.11.0 months (p = 0.038)) and quantitative criteria (15.0 vs. 11.7 months (p = 0.013)). Conclusions: Tumor-to-parotid uptake using a simple visual or semi-quantitative measure is a valuable prognostic biomarker for response in men with mCRPC treated with Lu-177 PSMA RLT.

Deep learning for image classification in dedicated breast positron emission tomography (dbPET).

This study aimed to investigate and determine the best deep learning (DL) model to predict breast cancer (BC) with dedicated breast positron emission tomography (dbPET) images. Of the 1598 women who underwent dbPET examination between April 2015 and August 2020, a total of 618 breasts on 309 examinations for 284 women who were diagnosed with BC or non-BC were analyzed in this retrospective study. The Xception-based DL model was trained to predict BC or non-BC using dbPET images from 458 breasts of 109 BCs and 349 non-BCs, which consisted of mediallateral and craniocaudal maximum intensity projection images, respectively. It was tested using dbPET images from 160 breasts of 43 BC and 117 non-BC. Two expert radiologists and two radiology residents also interpreted them. Sensitivity, specificity, and area under the receiver operating characteristic curves (AUCs) were calculated. Our DL model had a sensitivity and specificity of 93% and 93%, respectively, while radiologists had a sensitivity and specificity of 77-89% and 79-100%, respectively. Diagnostic performance of our model (AUC = 0.937) tended to be superior to that of residents (AUC = 0.876 and 0.868, p = 0.073 and 0.073), although not significantly different. Moreover, no significant differences were found between the model and experts (AUC = 0.983 and 0.941, p = 0.095 and 0.907). Our DL model could be applied to dbPET and achieve the same diagnostic ability as that of experts.

Evaluation of Lung Tumor Target Volume in a Large Sample: Target and Clinical Factors Influencing the Volume Derived From Four-Dimensional CT and Cone Beam CT.

Background and PurposeThis study aimed to systematically evaluate the influence of target-related and clinical factors on volume differences and the similarity of targets derived from four-dimensional computed tomography (4DCT) and cone beam computed tomography (CBCT) images in lung stereotactic body radiation therapy (SBRT).Materials and Methods4DCT and CBCT image data of 210 tumors from 195 patients were analyzed. The internal gross target volume (IGTV) derived from the maximum intensity projection (MIP) of 4DCT (IGTV-MIP) and the IGTV from CBCT (IGTV-CBCT) were compared with the reference IGTV from 10 phases of 4DCT (IGTV-10). The target size, tumor motion, and the similarity between IGTVs were measured. The influence of target-related and clinical factors on the adequacy of IGTVs derived from 4DCT MIP and CBCT images was evaluated.ResultsThe mean tumor motion amplitude in the 3D direction was 6.5 ± 5 mm. The mean size ratio of IGTV-CBCT and IGTV-MIP compared to IGTV-10 in all patients was 0.71 ± 0.21 and 0.8 ± 0.14, respectively. Female sex, greater BSA, and larger target size were protective factors, while the Karnofsky Performance Status, body mass index, and motion were risk factors for the similarity between IGTV-MIP and IGTV-10. Older age and larger target size were protective factors, while adhesion to the heart, coexistence with cardiopathy, and tumor motion were risk factors for the similarity between IGTV-CBCT and IGTV-10.ConclusionClinical factors should be considered when using MIP images for defining ITV, and when using CBCT images for verifying treatment targets.

Deep Learning Using Multiple Degrees of Maximum-Intensity Projection for PET/CT Image Classification in Breast Cancer.

Deep learning (DL) has become a remarkably powerful tool for image processing recently. However, the usefulness of DL in positron emission tomography (PET)/computed tomography (CT) for breast cancer (BC) has been insufficiently studied. This study investigated whether a DL model using images with multiple degrees of PET maximum-intensity projection (MIP) images contributes to increase diagnostic accuracy for PET/CT image classification in BC. We retrospectively gathered 400 images of 200 BC and 200 non-BC patients for training data. For each image, we obtained PET MIP images with four different degrees (0°, 30°, 60°, 90°) and made two DL models using Xception. One DL model diagnosed BC with only 0-degree MIP and the other used four different degrees. After training phases, our DL models analyzed test data including 50 BC and 50 non-BC patients. Five radiologists interpreted these test data. Sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were calculated. Our 4-degree model, 0-degree model, and radiologists had a sensitivity of 96%, 82%, and 80–98% and a specificity of 80%, 88%, and 76–92%, respectively. Our 4-degree model had equal or better diagnostic performance compared with that of the radiologists (AUC = 0.936 and 0.872–0.967, p = 0.036–0.405). A DL model similar to our 4-degree model may lead to help radiologists in their diagnostic work in the future.

Utility of Maximum Intensity Projections in Volumetric High Resolution Computed Tomography Lung in the Evaluation of Diffuse Lung Diseases- A Retrospective Analysis

Introduction: Though volumetric High Resolution Computed Tomography (HRCT) has increased the sensitivity of diagnosing diffuse lung diseases, there are some drawbacks related with lesion detection. Post processing techniques like Maximum Intensity Projection (MIP) can help in increasing the sensitivity of HRCT further. Aim: To study various possible utilities of MIP in HRCT lung. Materials and Methods: A retrospective cross-sectional study was done in a tertiary care hospital, including all Computed Tomography (CT) scans performed in July 2020 to August 2020, with multiple lung findings and diffuse involvement of lungs. Two radiologists reviewed the base mean axial images of 1 mm and MIP images in sliding scale from 2 to 10 mm in lung window. They recorded the findings in every scan based on widely accepted basic pattern of interpretation of HRCT thorax. The slice/slab thickness that best depicted each finding was noted. Inter reader agreement was calculated for each finding. Results: Fifty nine CT cases, which included scans of 34 males and 25 female patients with a mean age of 37.4±12.5 years were reviewed. There was near perfect to substantial agreement between the two radiologists on various findings (kappa &gt; 0.75). MIP images of 6-10 mm slab thickness were best suited to detect smaller nodules, assess vessel size in areas of mosaic attenuation and detect dense lesions in images with poor breath hold. The MIP images of 2-5 mm thickness were useful in detecting the location in relation to secondary pulmonary lobule, reticular densities and crazy-paving densities. Conclusion: The MIP being an easy to use and readily available post processing technique, when used with volumetric HRCT dataset of thorax enhanced the value of HRCT in detecting various lesions.

Improving the accuracy of motion quantification using area detector computed tomography for real-time tumor-tracking irradiation in stereotactic ablative radiotherapy.

CyberKnife radiotherapy enables tumor-tracking irradiation using positional information regarding the tumor and a fiducial marker in a patient's body. This positional information acts as a surrogate of tumor motion. Therefore, deviations in these movements should be quantitatively estimated and included as an internal margin for radiation treatment planning. This study aimed to investigate variations between the positions of fiducial markers and tumor regions using 320-row area detector computed tomography and to analyze the effectiveness of our proposed method in contouring tumor regions based on the fiducial marker position. To determine the moving tumor volume, a typical single-phase image was selected, and pixel values in other phase images were accumulated. Moreover, a maximum-intensity projection image was created to clarify motion deviations in the tumor volume. To evaluate the delineation accuracy, the dice similarity coefficient and mean distance to agreement were calculated in phase-selected and breath-holding computed tomography. Moving chest phantom images were acquired using helical scanning 4-dimensional computed tomography (H-4DCT) and volumetric scanning 4-dimensional computed tomography (V-4DCT), and the delineation accuracies were compared for each scanning type. The average dice similarity coefficient and mean distance to agreement were degraded in limited-phase images, which cannot represent the hysteretic motion of a tumor. Moreover, deviations in tumor volume with unstable motion reached 71.6% in H-4DCT but only 1.6% in V-4DCT. Our proposed method with V-4DCT using area detector computed tomography can achieve accurate moving tumor delineation and can clarify positional associations between the fiducial marker and tumor under respiratory motion.

Quantitative Analysis of Actin Cable Length in Yeast.

Polarized actin cables in S. cerevisiae are linear bundles of crosslinked actin filaments that are assembled by two formins, Bnr1 (localized to the bud neck), and Bni1 (localized to the bud tip). Actin is polymerized at these two sites, which results in cables extending along the cell cortex toward the back of the mother cell. These cables serve as polarized tracks for myosin-based transport of secretory vesicles and other cargo, from the mother cell to the growing daughter cell. Until recently, descriptions of actin cable morphology and architecture have largely been qualitative or descriptive in nature. Here, we introduce a new quantitative method that enables more precise characterization of actin cable length. This technological advance generates quantitative datasets that can be used to determine the contributions of different actin regulatory proteins to the maintenance of cable architecture, and to assess how different pharmacological agents affect cable arrays. Additionally, these datasets can be used to test theoretical models, and be compared to results from computational simulations of actin assembly. Graphical abstract: Illustration of actin cable length and morphology analysis. (A) Representative maximum intensity projection image of S. cerevisiae fixed and stained with fluorescently-conjugated phalloidin to label F-actin (displayed in color), and fluorescently-conjugated Concanavalin A to label the cell wall (displayed in grey scale). Lengths of actin cables traced from the bud neck to their ends are indicated (dashed lines). (B) Inverted grey scale image of F-actin labelled with fluorescently-conjugated phalloidin and the cell wall traced in black. The length (purple) and end-to-end distance (green) of a single actin cable is indicated. Scale bar, 2 µm. (C-E) Actin cable length (C), end-to-end distance (D), and tortuosity (E) from hypothetical datasets, where each data point represents an individual cable and larger symbols represent the mean from each hypothetical experiment. Error bars, 95% confidence intervals.

Feasibility of Single-Shot Whole Thoracic Time-Resolved MR Angiography to Evaluate Patients with Multiple Pulmonary Arteriovenous Malformations.

ObjectiveTo evaluate the feasibility of single-shot whole thoracic time-resolved MR angiography (TR-MRA) to identify the feeding arteries of pulmonary arteriovenous malformations (PAVMs) and reperfusion of the lesion after embolization in patients with multiple PAVMs.Materials and MethodsNine patients (8 females and 1 male; age range, 23–65 years) with a total of 62 PAVMs who underwent percutaneous embolization for multiple PAVMs and were subsequently followed up using TR-MRA and CT obtained within 6 months from each other were retrospectively reviewed. All imaging analyses were performed by two independent readers blinded to clinical information. The visibility of the feeding arteries on maximum intensity projection (MIP) reconstruction and multiplanar reconstruction (MPR) TR-MRA images was evaluated by comparing them to CT as a reference. The accuracy of TR-MRA for diagnosing reperfusion of the PAVM after embolization was assessed in a subgroup with angiographic confirmation. The reliability between the readers in interpreting the TR-MRA results was analyzed using kappa (κ) statistics.ResultsFeeding arteries were visible on the original MIP images of TR-MRA in 82.3% (51/62) and 85.5% (53/62) of readers 1 and 2, respectively. Using the MPR, the rates increased to 93.5% (58/62) and 95.2% (59/62), respectively (κ = 0.760 and 0.792, respectively). Factors for invisibility were the course of feeding arteries in the anteroposterior plane, proximity to large enhancing vessels, adjacency to the chest wall, pulsation of the heart, and small feeding arteries. Thirty-seven PAVMs in five patients had angiographic confirmation of reperfusion status after embolization (32 occlusions and 5 reperfusions). TR-MRA showed 100% (5/5) sensitivity and 100% (32/32, including three cases in which the feeding arteries were not visible on TR-MRA) specificity for both readers.ConclusionSingle-shot whole thoracic TR-MRA with MPR showed good visibility of the feeding arteries of PAVMs and high accuracy in diagnosing reperfusion after embolization. Single-shot whole thoracic TR-MRA may be a feasible method for the follow-up of patients with multiple PAVMs.