Search

Background: Qualitative assessment using the Deauville five-point Scale (DS) is the gold standard for interim and end-of treatment positron-emission tomography (PET) interpretation in lymphoma. In the present study we assessed the reliability and the prognostic value of different semi- quantitative parameters in comparison with DS for interim PET (iPET) interpretation in Hodgkin lymphoma (HL)., Methods: A cohort of 82 out of 260 patients with advanced stage HL enrolled in the International Validation Study (IVS), scored as 3 to 5 by the expert panel was included in the present report. Two nuclear medicine physicians blinded to patient history, clinical data and treatment outcome reviewed independently the iPET using the following parameters: DS, SUVmax, SUVpeak of the most active lesion, Qmax (ratio of SUVmax of the lesion to liver SUVmax) and Qres (ratio of SUVpeak of the lesion to liver SUVmean). The optimal sensitivity, specificity, positive and negative predictive value to predict treatment outcome was calculated for all the above parameters with the receiver operator characteristics analysis., Results: The prognostic value of all parameters were similar, the best cut-off value being 4 for DS (area under the curve [AUC], 0.81 95% CI: 0.72-0.90), 3.81 for SUVmax (AUC 0.82 95% CI: 0.73-0.91), 3.20 for SUVpeak (AUC 0.86 95% CI: 0.77-0.94), 1.07 for Qmax (AUC 0.84 95% CI: 0.75-0.93) and 1.38 for Qres (AUC 0.84 95% CI: 0.75-0.93). The reproducibility of different parameters was similar as the inter-observer variability measured with Cohen's kappa were 0.93 (95% CI: 0.84-1.01) for the DS, 0.88 (0.77-0.98) for SUVmax, 0.82 (0.70-0.95) for SUVpeak, 0.85 (0.74-0.97) for Qres and 0.78 (0.65-0.92) for Qmax. Due to the high specificity of SUVpeak (0.87) and to the good sensitivity of DS (0.86), upon the use of both parameters the positive predictive value increased from 0.65 of the DS alone to 0.79. When both parameters were positive in iPET, 3-years Failure-Free Survival (FFS) was significantly lower compared to patients whose iPET was interpreted with qualitative parameters only (DS 4 or 5): 21% vs. 35%. On the other hand, the FFS of patients with negative results was not significantly different (88% vs. 86%)., Conclusions: In this study we demonstrated that, combining semi-quantitative parameters with SUVpeak to a pure qualitative interpretation key with DS, it is possible to increase the positive predictive value of iPET and to identify with higher precision the patients subset with a very dismal prognosis. However, these retrospective findings should be confirmed prospectively in a larger patient cohort.

Role of interim-PET (I-PET) in diffuse large B-cell Lymphoma (DLBCL) is controversial. To determine predictive value of I-PET on progression-free survival (PFS), we enrolled 88 first-line DLBCL patients treated with 6-8 R-CHOP courses regardless of I-PET. PET/CT were performed at diagnosis, after 2 to 4 courses and at the end of therapy with central reviewing according to visual dichotomous criteria. Results are as follows: I-PET, 72% negative, 28% positive; final-PET (F-PET), 88% negative, 12% positive; clinical complete response 90%. Concordance between clinical response and F-PET negativity was 97% because of 2 false positive. With a median follow-up of 26.2 months, 2-year overall survival and PFS were 91% and 77%, respectively. Two-year PFS for I-PET and F-PET negative versus positive were as follows: I-PET 85% versus 72% (P = .0475); F-PET 83% versus 64% (P < .001). Because of a small number of events, 2 independent bivariate Cox models were tested for PFS. In model 1, F-PET contradicted I-PET (hazard ratio [HR] = 5.03, P = .015 vs 1.27, P = 691); in model 2, F-PET (HR = 4.54) and International propnostic Index score (HR = 5.36, P = .001) remained independent prognostic factors. In conclusion, positive I-PET is not predictive of a worse outcome in DLBCL; larger prospective studies and harmonization of I-PET reading criteria are needed.

Background: Merkel cell carcinoma (MCC) is an aggressive neuroendocrine skin tumor. Currently, 18F-fluoro-deoxy-glucose (18F-FDG) PET/CT is the functional imaging modality of choice. Few data are available on the use of 68Ga-somatostatin analogs. The aim of our study was to evaluate and compare the diagnostic performance of 18F-FDG and 68Ga-somatostatin analog PET/CT in MCC patients.Results: Fifteen patients (12 males, 3 females; median age 73 years; range 41-81 years) with histologically proven MCC (4 with unknown primary lesion) who underwent both 18F-FDG and 68Ga-somatostatin analog PET/CT for staging, re-staging, or treatment response assessment were retrospectively evaluated. Results of both studies were qualitatively analyzed and compared on a patient- and lesion-based analysis, using histology or clinical/radiological follow-up as reference standard for final diagnosis. According to final diagnosis, 8/15 patients had at least one MCC lesion and 7/15 had no evidence of disease. On a patient-based analysis, 18F-FDG and 68Ga-somatostatin analogs correctly classified as positive 8/8 (100% sensitivity) patients and as negative 6/7 (85.7% specificity) and 5/7 (71.4% specificity) patients, respectively, with no significant difference. On a lesion-based analysis, 18F-FDG detected 67/75 lesions (89%) and 68Ga-somatostatin analogs 69/75 (92%), with no significant difference. In four patients with unknown primary MCC, both tracers failed to identify the primary MCC site.Conclusions: Our preliminary data suggest that 18F-FDG and 68Ga-somatostatin analog PET/CT provide good and equivalent diagnostic performance, adding interesting insights into the complex MCC biology. However, these results do not suggest that 18F-FDG PET/CT should be replaced by 68Ga-somatostatin receptor imaging, which should be performed in addition, according to clinical indication, to the perspective of “personalized medicine.” [ABSTRACT FROM AUTHOR]

• Malignant lymphomas subtype classification with machine learning. • Texture features extracted from FDG-PET volumes of interest. • Multiple-instance learning SVM for patient-level categorization. • Random forests exploited for feature selection. Malignant lymphomas are cancers of the immune system and are characterized by enlarged lymph nodes that typically spread across many different sites. Many different histological subtypes exist, whose diagnosis is typically based on sampling (biopsy) of a single tumor site, whereas total body examinations with computed tomography and positron emission tomography, though not diagnostic, are able to provide a comprehensive picture of the patient. In this work, we exploit a data-driven approach based on multiple-instance learning algorithms and texture analysis features extracted from positron emission tomography, to predict differential diagnosis of the main malignant lymphomas subtypes. We exploit a multiple-instance learning setting where support vector machines and random forests are used as classifiers both at the level of single VOIs (instances) and at the level of patients (bags). We present results on two datasets comprising patients that suffer from four different types of malignant lymphomas, namely diffuse large B cell lymphoma, follicular lymphoma, Hodgkin's lymphoma, and mantle cell lymphoma. Despite the complexity of the task, experimental results show that, with sufficient data samples, some cancer subtypes, such as the Hodgkin's lymphoma, can be identified from texture information: in particular, we achieve a 97.0% of sensitivity (recall) and a 94.1% of predictive positive value (precision) on a dataset that consists in 60 patients. The presented study indicates that texture analysis features extracted from positron emission tomography, combined with multiple-instance machine learning algorithms, can be discriminating for different malignant lymphomas subtypes. [ABSTRACT FROM AUTHOR]