Construction of a nomogram for predicting COVID-19 in-hospital mortality: A machine learning analysis.

Background and objectives: We aim to verify the use of ML algorithms to predict patient outcome using a relatively small dataset and to create a nomogram to assess in-hospital mortality of patients with COVID-19. Methods: A database of 200 COVID-19 patients admitted to the Clinical Hospital of State University of Campinas (UNICAMP) was used in this analysis. Patient features were divided into three categories: clinical, chest abnormalities, and body composition characteristics acquired by computerized tomography. These features were evaluated independently and combined to predict patient outcomes. To minimize performance fluctuations due to low sample number, reduce possible bias related to outliers, and evaluate the uncertainties generated by the small dataset, we developed a shuffling technique, a modified version of the Monte Carlo Cross Validation, creating several subgroups for training the algorithm and complementary testing subgroups. The following ML algorithms were tested: random forest, boosted decision trees, logistic regression, support vector machines, and neural networks. Performance was evaluated by analyzing Receiver operating characteristic (ROC) curves. The importance of each feature in the determination of the outcome predictability was also studied and a nomogram was created based on the most important features selected by the exclusion test. Results: Among the different sets of features, clinical variables age, lymphocyte number and weight were the most valuable features for prognosis prediction. However, we observed that skeletal muscle radiodensity and presence of pleural effusion were also important for outcome determination. Integrating these independent predictors was successfully developed to accurately predict mortality in COVID-19 in hospital patients. A nomogram based on these five features was created to predict COVID-19 mortality in hospitalized patients. The area under the ROC curve was 0.86 ± 0.04. Conclusion: ML algorithms can be reliable for the prediction of COVID-19-related in-hospital mortality, even when using a relatively small dataset. The success of ML techniques in smaller datasets broadens the applicability of these methods in several problems in the medical area. In addition, feature importance analysis allowed us to determine the most important variables for the prediction tasks resulting in a nomogram with good accuracy and clinical utility in predicting COVID-19 in-hospital mortality.

Computed tomography-based skeletal segmentation for quantitative PET metrics of bone involvement in multiple myeloma.

PURPOSE: Quantifications in nuclear medicine are occasionally limited by the lack of standardization for defining volumes of interest (VOIs) on functional images. In the present article, we propose the use of computed tomography (CT)-based skeletal segmentation to determine anatomically the VOI in order to calculate quantitative parameters of fluorine 18 fluorodeoxyglucose (F-FDG) PET/CT images from patients with multiple myeloma. METHODS: We evaluated 101 whole-body F-FDG PET/CTs of 58 patients with multiple myeloma. An initial subjective visual analysis of the PET images was used to classify the bone involvement as negative/mild, moderate, or marked. Then, a fully automated CT-based segmentation of the skeleton was performed on PET images. The maximum, mean, and SD of the standardized uptake values (SUVmax, SUVmean, and SDSUV) were calculated for bone tissue and compared with the visual analysis. RESULTS: Forty-five (44.5%), 32 (31.7%), and 24 (23.8%) PET images were, respectively, classified as negative/mild, moderate, or marked bone involvement. All quantitative parameters were significantly related to the visual assessment of bone involvement. This association was stronger for the SUVmean [odds ratio (OR): 10.52 (95% confidence interval (CI), 5.68-19.48); P < 0.0001] and for the SDSUV [OR: 5.58 (95% CI, 3.31-9.42); P < 0.001) than for the SUVmax [OR: 1.01 (95% CI, 1.003-1.022); P = 0.003]. CONCLUSION: CT-based skeletal segmentation allows for automated and therefore reproducible calculation of PET quantitative parameters of bone involvement in patients with multiple myeloma. Using this method, the SUVmean and its respective SD correlated better with the visual analysis of F-FDG PET images than SUVmax. Its value in staging and evaluating therapy response needs to be evaluated.

Validation of a Multifocal Segmentation Method for Measuring Metabolic Tumor Volume in Hodgkin Lymphoma.

Quantification of metabolic tumor volume (MTV) and total lesion glycolysis (TLG) can be time-consuming. We evaluated the performance of an automatic multifocal segmentation (MFS) method of quantification in patients with different stages of Hodgkin lymphoma, using the multiple VOI (MV) method as reference. Methods: This prospective bicentric study included 50 patients with Hodgkin lymphoma who underwent staging 18F-FGD PET/CT. The examinations were centrally reviewed and processed with commercial MFS software to obtain MTV and TLG using 2 fixed relative thresholds (40% and 20% of SUVmax) for each lesion. All PET/CT scans were processed using the MV and MFS methods. Interclass correlation coefficients and Bland-Altman plots were used for statistical analysis. Repeated calculations of MTV and TLG values by 2 observers with different degrees of PET/CT imaging experience were used to ascertain interobserver agreement on the MFS method. Results: The means and SDs obtained for the MTV with MV and MFS were, respectively, 736 ± 856 mL and 660 ± 699 mL for the 20% threshold and 313 ± 359 mL and 372 ± 434 mL for the 40% threshold. The time spent calculating the MTV was much shorter with the MFS method than with the MV method (median time, 11.6 min [range, 1-30 min] and 64.4 min [range, 1-240 min], respectively), especially in patients with advanced disease. Time spent was similar in patients with localized disease. There were no statistical differences between the MFS values obtained by the 2 different observers. Conclusion: MTV and TLG calculations using MFS are reproducible, generate similar results to those obtained with MV, and are much less timing-consuming. Main differences between the 2 methods were related to difficulties in avoiding overlay of VOIs in the MV technique. MV and MFS perform equally well in patients with a small number of lesions.

99mTc-sestamibi SPECT/CT and 18F-FDG-PET/CT have similar performance but different imaging patterns in newly diagnosed multiple myeloma.

PURPOSE: F-fluorodeoxiglucose (F-FDG)-PET/CT has been widely used to evaluate multiple myeloma. Tc-sestamibi (MIBI) scintigraphy has also been proposed for assessing multiple myeloma, but its use with state-of-the-art single-photon emission computed tomography/computed tomography (SPECT/CT) technology has not been fully evaluated.This study aimed to compare these two imaging modalities in multiple myeloma staging. MATERIALS AND METHODS: Sixty-two patients with recently diagnosed multiple myeloma were submitted to whole-body F-FDG-PET/CT and whole-body MIBI scans plus SPECT/CT of the chest and abdomen/pelvis. Number of focal lesions, contiguous soft tissue involvement (CSTI), extramedullary lesions (EMLs) and diffuse bone marrow (BM) involvement were recorded. RESULTS: PET/CT was positive in 59 patients (95%) and MIBI SPECT/CT in 58 (93%) (P = 0.69). MIBI detected more diffuse bone marrow involvement than PET/CT (respectively 78 vs. 58% of the patients), while PET/CT demonstrated more focal lesions than MIBI SPECT/CT (81 vs. 54% of the patients) (P = 0.002). PET/CT detected EMLs in four subjects and MIBI in one subject. CSTI was found in 28 (45%) and 23 (37%) patients on PET/CT and MIBI images, respectively (P = 0.36). Three patients with lytic lesions and no FDG uptake were MIBI positive, and two subjects with lytic lesions without MIBI uptake were FDG positive. CONCLUSION: MIBI SPECT/CT performs similarly to F-FDG-PET/CT in identifying sites of active disease in multiple myeloma staging. MIBI is more efficient than FDG for detecting the diffuse involvement of bone marrow but less efficient for detecting focal lesions. Some patients presented a 'mismatch' pattern of FDG/MIBI uptake.

Proposal for a Quantitative 18F-FDG PET/CT Metabolic Parameter to Assess the Intensity of Bone Involvement in Multiple Myeloma.

Many efforts have been made to standardize the interpretation of 18F-FDG PET/CT in multiple myeloma (MM) with qualitative visual analysis or with quantitative metabolic parameters using various methods for lesion segmentation of PET images. The aim of this study was to propose a quantitative method for bone and bone marrow evaluation of 18F-FDG PET/CT considering the extent and intensity of bone 18F-FDG uptake: Intensity of Bone Involvement (IBI). Whole body 18F-FDG PET/CT of 59 consecutive MM patients were evaluated. Compact bone tissue was segmented in PET images using a global threshold for HU of the registered CT image. A whole skeleton mask was created and the percentage of its volume with 18F-FDG uptake above hepatic uptake was calculated (Percentage of Bone Involvement - PBI). IBI was defined by multiplying PBI by mean SUV above hepatic uptake. IBI was compared with visual analysis performed by two experienced nuclear medicine physicians. IBI calculation was feasible in all images (range:0.00-1.35). Visual analysis categorized PET exams into three groups (negative/mild, moderate and marked bone involvement), that had different ranges of IBI (multi comparison analysis, p < 0.0001). There was an inverse correlation between the patients' hemoglobin values and IBI (r = -0.248;p = 0.02). IBI score is an objective measure of bone and bone marrow involvement in MM, allowing the categorization of patients in different degrees of aggressiveness of the bone disease. The next step is to validate IBI in a larger group of patients, before and after treatment and in a multicentre setting.