Prediction of the local treatment outcome in patients with oropharyngeal squamous cell carcinoma using deep learning analysis of pretreatment FDG-PET images.

BACKGROUND: This study aimed to assess the utility of deep learning analysis using pretreatment FDG-PET images to predict local treatment outcome in oropharyngeal squamous cell carcinoma (OPSCC) patients. METHODS: One hundred fifty-four OPSCC patients who received pretreatment FDG-PET were included and divided into training (n = 102) and test (n = 52) sets. The diagnosis of local failure and local progression-free survival (PFS) rates were obtained from patient medical records. In deep learning analyses, axial and coronal images were assessed by three different architectures (AlexNet, GoogLeNET, and ResNet). In the training set, FDG-PET images were analyzed after the data augmentation process for the diagnostic model creation. A multivariate clinical model was also created using a binomial logistic regression model from a patient's clinical characteristics. The test data set was subsequently analyzed for confirmation of diagnostic accuracy. Assessment of local PFS rates was also performed. RESULTS: Training sessions were successfully performed with an accuracy of 74-89%. ROC curve analyses revealed an AUC of 0.61-0.85 by the deep learning model in the test set, whereas it was 0.62 by T-stage, 0.59 by clinical stage, and 0.74 by a multivariate clinical model. The highest AUC (0.85) was obtained with deep learning analysis of ResNet architecture. Cox proportional hazards regression analysis revealed deep learning-based classification by a multivariate clinical model (P < .05), and ResNet (P < .001) was a significant predictor of the treatment outcome. In the Kaplan-Meier analysis, the deep learning-based classification divided the patient's local PFS rate better than the T-stage, clinical stage, and a multivariate clinical model. CONCLUSIONS: Deep learning-based diagnostic model with FDG-PET images indicated its possibility to predict local treatment outcomes in OPSCCs.

Deep learning analysis using FDG-PET to predict treatment outcome in patients with oral cavity squamous cell carcinoma.

OBJECTIVE: To assess the utility of deep learning analysis using 18F-fluorodeoxyglucose (FDG) uptake by positron emission tomography (PET/CT) to predict disease-free survival (DFS) in patients with oral cavity squamous cell carcinoma (OCSCC). METHODS: One hundred thirteen patients with OCSCC who received pretreatment FDG-PET/CT were included. They were divided into training (83 patients) and test (30 patients) sets. The diagnosis of treatment control/failure and the DFS rate were obtained from patients' medical records. In deep learning analyses, three planes of axial, coronal, and sagittal FDG-PET images were assessed by ResNet-101 architecture. In the training set, image analysis was performed for the diagnostic model creation. The test data set was subsequently analyzed for confirmation of diagnostic accuracy. T-stage, clinical stage, and conventional FDG-PET parameters (the maximum and mean standardized uptake value (SUVmax and SUVmean), heterogeneity index, metabolic tumor volume (MTV), and total lesion glycolysis (TLG) were also assessed with determining the optimal cutoff from training dataset and then validated their diagnostic ability from test dataset. RESULTS: In dividing into patients with treatment control and failure, the highest diagnostic accuracy of 0.8 was obtained using deep learning classification, with a sensitivity of 0.8, specificity of 0.8, positive predictive value of 0.89, and negative predictive value of 0.67. In the Kaplan-Meier analysis, the DFS rate was significantly different only with the analysis of deep learning-based classification (p < .01). CONCLUSIONS: Deep learning-based diagnosis with FDG-PET images may predict treatment outcome in patients with OCSCC. KEY POINTS: • Deep learning-based diagnosis of FDG-PET images showed the highest diagnostic accuracy to predict the treatment outcome in patients with oral cavity squamous cell carcinoma. • Deep learning-based diagnosis was shown to differentiate patients between good and poor disease-free survival more clearly than conventional T-stage, clinical stage, and conventional FDG-PET-based parameters.

The utility of CT for predicting bile leaks in hepatic trauma.

The purpose of this study was to determine the efficacy of CT to predict the development of bile leaks in hepatic trauma. This HIPAA-compliant retrospective study was IRB approved and consent was waived. All patients who sustained hepatic trauma between January 1, 2006, and January 31, 2012, and who underwent CT and hepatobiliary scans during the same hospital admission were included. One hundred and thirty-two patients met the inclusion criteria. Comparison between the presence of biliary injury relative to American Association for the Surgery of Trauma (AAST) hepatic injury grade and mean distance of the hepatic laceration to the inferior vena cava (IVC) was made. The ability of free fluid to predict bile injury was analyzed. Forty-one (31 %) of the 132 patients had positive hepatobiliary scans. Of these 41 patients, seven (17 %) sustained low-grade and 34 (83 %) sustained high-grade hepatic injury compared with the 37 (41 %) low-grade and 54 (59 %) high-grade hepatic injuries in the negative hepatobiliary scan group. The mean distance to the IVC was 2.4 cm (SD 2.9 cm) and 3.6 cm (SD 3.3 cm) in patients with and without bile leaks, respectively. A statistically significant difference in the proportion of high-grade injuries and the mean distance from the IVC between the two groups was identified. The presence of free fluid on CT is sensitive, but not specific, for detecting a bile leak. CT findings, including AAST liver injury grade and location of the liver laceration, are able to predict which patients are at risk for developing bile leaks as seen on hepatobiliary scintigraphy, whereas the presence of free fluid is not.

Quantitative interpretation of FDG PET/CT with myocardial perfusion imaging increases diagnostic information in the evaluation of cardiac sarcoidosis.

BACKGROUND: FDG PET/CT with myocardial perfusion imaging is a useful method for evaluating cardiac sarcoidosis (CS), but interpretation is not standardized. We developed a method for quantification of cardiac FDG PET/CT and evaluated its relationship to conventional interpretation, perfusion defects, clinical events, and immunosuppressive treatment. METHODS AND RESULTS: FDG PET/CT with MPI studies performed for CS (n = 38) were retrospectively compared to negative control studies acquired for oncologic indications (n = 10). Quantitative measures of FDG volume-intensity (Cardiac Metabolic Activity, CMA) was performed using standardized uptake values (SUVs). CMA (477.7 ± 909 vs 0.55 ± 2.1 vs 0.3 ± 0.3 g glucose, P = .02) was significantly greater in visually FDG-positive studies compared to visually negative and oncologic negative studies. Among patients with CS, CMA was greater in studies with an EF < 50% (760.3 ± 1,148 vs 87.4 ± 161 g glucose, P = .03) and preceding an adverse clinical event (1,095 ± 1,253 vs 73 ± 144 g glucose, P = .006). CMA was the only independent predictor of events by multivariate analysis. In patients with repeat examinations (n = 7), CMA decreased with prednisone treatment in 5 of 6 patients. CONCLUSIONS: Quantification of FDG uptake in CS correlates with lower EFs, clinical events, and immunosuppression treatment.

Interreader agreement and variability of FDG PET volumetric parameters in human solid tumors.

OBJECTIVE: The purpose of this article is to evaluate the interreader agreement and variability of two (18)F-FDG PET parameters, metabolic tumor volume and total lesion glycolysis, in human solid tumors. MATERIALS AND METHODS: One hundred eleven patients (mean [± SD] age, 61.9 ± 12.5 years) with baseline staging FDG PET/CT scans were included. Two readers independently read the scans and segmented metabolic tumor volume and total lesion glycolysis using two fixed thresholds, 40% and 50% of the lesion's maximum standardized uptake value (SUVmax). The impact of the lesion's FDG avidity and location on reader agreement and variability was established. Intraclass correlation coefficient (ICC), precision, and Bland-Altman analysis were used to evaluate agreement and variability. RESULTS: The ICCs for 40% and 50% SUVmax segmentations of metabolic tumor volume between the readers were 0.987 and 0.995, and the corresponding values for 40% and 50% SUVmax segmentations of total lesion glycolysis were 0.987 and 0.986, respectively (p = 0.0001). The corresponding precisions were 0.5%, 0.2%, 0.5%, and 0.5%, respectively. The mean biases between the readers for 40% and 50% SUVmax segmentations of metabolic tumor volume were -1.78 ± 8.42 mL and -0.46 ± 2.1 mL and for 40% and 50% SUVmax segmentations of total lesion glycolysis were -7.3 ± 31.6 g and -2.97 ± 12.86 g, respectively. Subgroup analysis showed better precision and lesser variability for 50% SUVmax segmentations of metabolic tumor volume and total lesion glycolysis in patients with the highest and lowest FDG-avid primary tumors. The precision was highest and variability was lowest for lung tumors. CONCLUSION: There is excellent interreader agreement for measurement of metabolic tumor volume and total lesion glycolysis with 40% and 50% SUVmax threshold segmentations in human solid tumors.

FDG PET metabolic tumor volume segmentation and pathologic volume of primary human solid tumors.

OBJECTIVE: The purpose of this study was to establish the correlation and reliability among the pathologic tumor volume and gradient and fixed threshold segmentations of (18)F-FDG PET metabolic tumor volume of human solid tumors. MATERIALS AND METHODS: There were 52 patients included in the study who had undergone baseline PET/CT with subsequent resection of head and neck, lung, and colorectal tumors. The pathologic volume was calculated from three dimensions of the gross tumor specimen as a reference standard. The primary tumor metabolic tumor volume was segmented using gradient and 30%, 40%, and 50% maximum standardized uptake value (SUVmax) threshold methods. Pearson correlation coefficient, intraclass correlation coefficient, and Bland-Altman analyses were performed to establish the correlation and reliability among the pathologic volume and segmented metabolic tumor volume. RESULTS: The mean pathologic volume; gradient-based metabolic tumor volume; and 30%, 40%, and 50% SUVmax threshold metabolic tumor volumes were 13.46, 13.75, 15.47, 10.63, and 7.57 mL, respectively. The intraclass correlation coefficients among the pathologic volume and the gradient-based and 30%, 40%, and 50% SUVmax threshold metabolic tumor volumes were 0.95, 0.85, 0.80, and 0.76, respectively. The Bland-Altman biases were -0.3, -2.0, 2.82, and 5.9 mL, respectively. Of the small tumors (< 10 mL), 23 of the 35 patients had PET segmented volume outside 50% of the pathologic volume, and among the large tumors (≥ 10 mL) three of the 17 patients had PET segmented volumes that were outside 50% of pathologic volume. CONCLUSION: FDG PET metabolic tumor volume estimated using gradient segmentation had superior correlation and reliability with the estimated ellipsoid pathologic volume of the tumors compared with threshold method segmentation.

Liver SULmean at FDG PET/CT: interreader agreement and impact of placement of volume of interest.

PURPOSE: To evaluate how interreader agreement and the site of the volume of interest (VOI) affect the agreement and variability of liver mean standardized uptake value normalized to lean body mass (SUL(mean)) at fluorine 18 fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT). MATERIALS AND METHODS: Institutional review board approval was obtained for this HIPAA-compliant retrospective review of PET/CT images and patient records. PET/CT images were reviewed in 116 randomly selected patients who had undergone a baseline PET/CT examination and who had normal livers according to imaging and biochemical test results. A 30-mm-diameter spherical VOI was placed within the right lobe of the liver above, below, and at the level of the main portal vein. Two readers performed all measurements independently. Analysis of variance, intraclass correlation coefficient (ICC) analysis, and Bland-Altman analysis were performed. RESULTS: The mean SUL(mean) was between 2.11 and 2.17 at the upper, portal, and lower levels of the right lobe of the liver. The coefficient of variance was between 21.0% and 23.1%, without significant differences for location, with the least variance in the upper level. The ICC of the two readers varied between 0.98 and 0.99 (95% confidence interval [CI]: 0.97, 0.99; P = .0001) at each level. The greatest precision (narrowest CI) was also in the upper level. Bias was 0.025 ± 0.10 (standard deviation) at the upper level, was 0.004 ± 0.14 at the lower level, and was 0.047 ± 0.10 at the portal vein (P = .02). For each reader, there was almost perfect reliability between the SUL(mean) measurements made at the three levels, with an ICC of 0.98 (95% CI: 0.98, 0.99; P = .0001). CONCLUSION: Liver SUL(mean) at FDG PET/CT has excellent interreader agreement, with similar values and variance whether measured at the upper, lower, or portal vein levels within the right lobe of the liver.

PET-based primary tumor volumetric parameters and survival of patients with non-small cell lung carcinoma.

OBJECTIVE: The purpose of the study was to assess metabolic tumor volume and total glycolytic activity of the primary tumor as prognostic parameters for outcome in patients with non-small cell lung carcinoma (NSCLC). MATERIALS AND METHODS: Thirty-nine patients who had undergone a baseline staging PET/CT examination at our institution for the diagnosis of NSCLC were retrospectively identified. The maximum standardized uptake value (SUV(max)), metabolic tumor volume, and total glycolytic activity were segmented from PET using the gradient method; 12-month survival and overall survival at the end of follow-up were used as outcome measures. Multivariate logistic regression, receiver operating characteristic curve analysis, and Kaplan-Meier curves for survival analysis were generated and compared using the Mantel-Cox log-rank test. RESULTS: The mean gradient-based metabolic tumor volume and gradient-based total glycolytic activity were significantly greater in the patients who died (93.3 mL and 597.5 g) than in those who survived (19.3 mL and 193.9 g, respectively) (p < 0.003 and p < 0.031). There was no statistically significant difference in the mean SUV(max) between the patients who survived (12.7) at 12 months and those who had died (13.1) (p = 0.85). On multivariate analysis, gradient-based metabolic tumor volume was the only variable associated with 12-month mortality when adjusted for all other factors.(.) The area under the curve (AUC) for gradient-based metabolic tumor volume was 0.77 (p < 0.006). A significant difference in the time to survival was observed between high and low gradient-based metabolic tumor volume (log-rank p < 0.05) cohorts using the median gradient-based metabolic tumor volume (9.7 mL) as the cut point. CONCLUSION: PET-based volumetric imaging parameters are potential prognostic markers of outcome in patients with NSCLC.

FDG PET/CT in the management of primary pleural tumors and pleural metastases.

OBJECTIVE. FDG PET/CT is emerging as an important modality in the evaluation of pleural tumors. PET/CT has an established role in the diagnosis and staging and shows promise in therapy planning, therapy response assessment, and providing prognostic information in patients with malignant pleural mesothelioma. This modality has distinct advantages in characterizing other primary pleural tumors and pleural metastases. CONCLUSION. FDG PET/CT is a useful imaging modality in the management of patients with primary pleural tumors and pleural metastases.

The role of PET/CT in the management of cervical cancer.

OBJECTIVE. Cervical cancer is the second most common malignancy in women worldwide and the third most common cause of cancer mortality in the United States. The aim of this article is to describe cervical cancer and outline the value of (18)F-FDG PET/CT in the management of cervical malignancy. CONCLUSION. The value of PET/CT has been found in staging and treatment strategy for cervical cancer. FDG PET/CT facilitates decision-making and radiation treatment planning and provides important information about treatment response, disease recurrence, and long-term survival.

Hematocrit and lactate trends help predict outcomes in trauma independent of CT and other clinical parameters.

Background: Hematocrit and lactate have an established role in trauma as indicators of bleeding and cell death, respectively. The wide availability of CT imaging and clinical data poses the question of how these can be used in combination to predict outcomes. Purpose: To assess the utility of hematocrit or lactate trends in predicting intensive care unit (ICU) admission and hospital length of stay (LOS) in patients with torso trauma combined with clinical parameters and injury findings on CT. Materials and Methods: This was a single-center retrospective study of adults with torso trauma in one year. Trends were defined as a unit change per hour. CT findings and clinical parameters were explanatory variables. Outcomes were ICU admission and hospital LOS. Multivariate logistic and negative binomial regression models were used to calculate the odds ratio (OR) and incident rate ratio (IRR). Results: Among 840 patients, 561 (72% males, age 39 ± 18) were included, and 168 patients (30%) were admitted to the ICU. Decreasing hematocrit trend [OR 2.54 (1.41-4.58), p = 0.002] and increasing lactate trend [OR 3.85 (1.35-11.01), p = 0.012] were associated with increased odds of ICU admission. LOS median was 2 (IQR: 1-5) days. Decreasing hematocrit trend [IRR 1.37 (1.13-1.66), p = 0.002] and increasing lactate trend [2.02 (1.43-2.85), p < 0.001] were associated with longer hospital LOS. Conclusion: Hematocrit and lactate trends may be helpful in predicting ICU admission and LOS in torso trauma independent of organ injuries on CT, age, or admission clinical parameters.

Identification of an APOE ε4-specific blood-based molecular pathway for Alzheimer's disease risk.

INTRODUCTION: The precise apolipoprotein E (APOE) ε4-specific molecular pathway(s) for Alzheimer's disease (AD) risk are unclear. METHODS: Plasma protein modules/cascades were analyzed using weighted gene co-expression network analysis (WGCNA) in the Alzheimer's Disease Neuroimaging Initiative study. Multivariable regression analyses were used to examine the associations among protein modules, AD diagnoses, cerebrospinal fluid (CSF) phosphorylated tau (p-tau), and brain glucose metabolism, stratified by APOE genotype. RESULTS: The Green Module was associated with AD diagnosis in APOE ε4 homozygotes. Three proteins from this module, C-reactive protein (CRP), complement C3, and complement factor H (CFH), had dose-dependent associations with CSF p-tau and cognitive impairment only in APOE ε4 homozygotes. The link among these three proteins and glucose hypometabolism was observed in brain regions of the default mode network (DMN) in APOE ε4 homozygotes. A Framingham Heart Study validation study supported the findings for AD. DISCUSSION: The study identifies the APOE ε4-specific CRP-C3-CFH inflammation pathway for AD, suggesting potential drug targets for the disease.Highlights: Identification of an APOE ε4 specific molecular pathway involving blood CRP, C3, and CFH for the risk of AD.CRP, C3, and CFH had dose-dependent associations with CSF p-Tau and brain glucose hypometabolism as well as with cognitive impairment only in APOE ε4 homozygotes.Targeting CRP, C3, and CFH may be protective and therapeutic for AD onset in APOE ε4 carriers.

Annualized changes in rate of amyloid deposition and neurodegeneration are greater in participants who become amyloid positive than those who remain amyloid negative.

This study longitudinally examined participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) who underwent a conversion in amyloid-beta (Aß) status in comparison to a group of ADNI participants who did not show a change in amyloid status over the same follow-up period. Participants included 136 ADNI dementia-free participants with 2 florbetapir positron emission tomography (PET) scans. Of these participants, 68 showed amyloid conversion as measured on florbetapir PET, and the other 68 did not. Amyloid converters and non-converters were chosen to have representative demographic data (age, education, sex, diagnostic status, and race). The amyloid converter group showed increased prevalence of APOE ε4 (p < 0.001), greater annualized percent volume loss in selected magnetic resonance imaging (MRI) regions (p < 0.05), lower cerebrospinal fluid Aß1-42 (p < 0.001), and greater amyloid retention (as measured by standard uptake value ratios) on florbetapir PET scans (p < 0.001) in comparison to the non-converter group. These results provide compelling evidence that important neuropathological changes are occurring alongside amyloid conversion.

PET/CT of cancer patients: part 1, pancreatic neoplasms.

OBJECTIVE: Pancreatic cancer continues to have a poor prognosis despite impressive improvements in the outcomes of many other types of cancer, often because most pancreatic neoplasms are found to be unresectable at diagnosis. The purpose of this review is to provide an overview of pancreatic cancer and the role of modern imaging in its diagnosis and management with an emphasis on (18)F-FDG PET/CT fusion imaging. CONCLUSION: Multimodality imaging is critical in the diagnosis and management of pancreatic cancer. PET/CT is increasingly viewed as a useful, accurate, and cost-effective modality in diagnosing and managing pancreatic cancer, but further studies are warranted. Early data suggest that contrast-enhanced PET/CT performed with modern PET/CT scanners yields high-resolution anatomic information for surgical and radiotherapeutic planning and functional information for whole-body staging in the care of patients with this disease.

Parotid gland tumors: preliminary data for the value of FDG PET/CT diagnostic parameters.

OBJECTIVE: The purpose of this article is to establish (18)F-FDG metabolic imaging parameters to differentiate benign and malignant tumors of the parotid gland. MATERIALS AND METHODS: Forty-nine patients with increased FDG uptake in the parotid gland were selected for the study group (29 men and 20 women; mean age, 63.14 ± 12.32 years). Another 49 patients without head and neck malignancies were selected as the control group (24 men and 25 women; mean age, 65.80 ± 11.51 years); they did not have a parotid lesion or increased FDG uptake in the parotid gland. Maximum standardized uptake value (SUV(max)) was obtained for all patients. Metabolic tumor volume and total glycolytic activity were measured in patients with a discrete parotid lesion (n = 24). Nonparameteric Student t test (Mann-Whitney U test) was performed for between-group analysis. RESULTS: The median SUV(max) of the increased diffuse uptake (2.55; interquartile range [IQR], 1.03-4.07) was significantly lower than the median SUV(max) of tumors (8.48; IQR, 1.46-15.5) (p < 0.01). The median SUV(max) of malignant tumors (11.8; IQR, 4.45-19.15) was significantly (p < 0.05) higher than that of benign tumors (6.4; IQR, 3.4-9.0). There was significant difference (p = 0.003) in the median metabolic tumor volume for malignant tumors (8.9; IQR, 5.1-25.5) and benign tumors or lesions (1.4; IQR, 1.00-2.9). Similar results were found for total glycolytic activity for malignant tumors (67.9; IQR, 24.2-137.6) and benign tumors or lesions (8.4; IQR, 3.9-13.6) (p = 0.002). CONCLUSION: FDG PET/CT SUV(max), metabolic tumor volume, and total glycolytic activity are imaging parameters to differentiate benign and malignant tumors of the parotid gland.

Value of PET/CT in the management of liver metastases, part 1.

OBJECTIVE: Metastasis is the most common (95%) of liver lesions. Early diagnosis and staging are the keys to treatment planning and prognosis. There is a consistent benefit to the use of PET/CT for detecting hepatic, local, and distant metastases from a variety of primary malignancies, which can contribute to staging and ultimately helps to establish the best course of treatment and to determine prognosis. CONCLUSION: For colorectal cancer, FDG PET and FDG PET/CT are particularly effective for identification of additional hepatic and extrahepatic metastases, frequently upstaging the tumor stage and affecting management. In addition, PET/CT is very useful in local ablative and systemic therapy assessment and surveillance for liver metastases.

Value of PET/CT in the management of primary hepatobiliary tumors, part 2.

OBJECTIVE: Primary hepatobiliary malignancies consist of hepatocellular carcinoma, cholangiocarcinoma, and gallbladder cancer. Benign hepatic lesions include hepatic cysts, hemagiomas, adenomas, and focal nodular hyperplasias. The utility of PET/CT in imaging primary hepatobiliary lesions varies according to the type and location of the lesion. CONCLUSION: There is a consistent benefit to the use of PET/CT for detection and staging, and it ultimately helps to establish the best course of treatment and to determine prognosis. In addition, PET/CT is very useful in local ablative and systemic therapy assessment and surveillance for hepatobiliary malignancies.

FDG PET/CT in patients with HIV.

OBJECTIVE: This article will discuss the (18)F-FDG normal variant uptake and the role of FDG PET/CT in malignancies in HIV-infected patients, CNS manifestations of HIV, assessing fever of unknown origin in HIV patients, assessing response to highly active antiretroviral therapy and assessing complications. CONCLUSION: FDG PET/CT is a valuable imaging study in the management of HIV-infected patients.

Use of 18F-FDG PET/CT as a predictive biomarker of outcome in patients with head-and-neck non-squamous cell carcinoma.

OBJECTIVE: The purpose of this article is to establish whether pretreatment (18)F-FDG uptake predicts disease-free survival (DFS) and overall survival in patients with head-and-neck non-squamous cell carcinoma (SCC). MATERIALS AND METHODS: Eighteen patients (six women and 12 men; mean [± SD] age at diagnosis, 57.89 ± 13.54 years) with head-and-neck non-SCC were included. Tumor FDG uptake was measured by the maximum standardized uptake value (SUV(max)) and was corrected for background liver FDG uptake to derive the corrected SUV(max). Receiver operating characteristic analyses were used to predict the optimal corrected SUV(max) cutoffs for respective outcomes of DFS (i.e., absence of recurrence) and death. RESULTS: The mean corrected SUV(max) of the 18 head-and-neck tumors was 5.63 ± 3.94 (range, 1.14-14.29). The optimal corrected SUV(max) cutoff for predicting DFS and overall survival was 5.79. DFS and overall survival were significantly higher among patients with corrected SUV(max) < 6 than among patients with corrected SUV(max) ≥ 6. The mean DFS for patients with corrected SUV(max) < 6 was 25.7 ± 11.14 months, and the mean DFS for patients with corrected SUV(max) ≥ 6 was 7.88 ± 7.1 months (p < 0.018). Among patients with corrected SUV(max) < 6, none died, and the mean length of follow-up for this group was 35.2 ± 9.96 months. All of the patients who died had corrected SUV(max) ≥ 6, and the overall survival for this group was 13.28 ± 12.89 months (p < 0.001). CONCLUSION: FDG uptake, as measured by corrected SUV(max), may be a predictive imaging biomarker for DFS and overall survival in patients with head-and-neck non-SCC.

Prediction of the human papillomavirus status in patients with oropharyngeal squamous cell carcinoma by FDG-PET imaging dataset using deep learning analysis: A hypothesis-generating study.

PURPOSE: To assess the diagnostic accuracy of imaging-based deep learning analysis to differentiate between human papillomavirus (HPV) positive and negative oropharyngeal squamous cell carcinomas (OPSCCs) using FDG-PET images. METHODS: One hundred and twenty patients with OPSCC who underwent pretreatment FDG-PET/CT were included and divided into the training 90 patients and validation 30 patients cohorts. In the training session, 2160 FDG-PET images were analyzed after data augmentation process by a deep learning technique to create a diagnostic model to discriminate between HPV-positive and HPV-negative OPSCCs. Validation cohort data were subsequently analyzed for confirmation of diagnostic accuracy in determining HPV status by the deep learning-based diagnosis model. In addition, two radiologists evaluated the validation cohort image-data to determine the HPV status based on each tumor's imaging findings. RESULTS: In deep learning analysis with training session, the diagnostic model using training dataset was successfully created. In the validation session, the deep learning diagnostic model revealed sensitivity of 0.83, specificity of 0.83, positive predictive value of 0.88, negative predictive value of 0.77, and diagnostic accuracy of 0.83, while the visual assessment by two radiologists revealed 0.78, 0.5, 0.7, 0.6, and 0.67 (reader 1), and 0.56, 0.67, 0.71, 0.5, and 0.6 (reader 2), respectively. Chi square test showed a significant difference between deep learning- and radiologist-based diagnostic accuracy (reader 1: P = 0.016, reader 2: P = 0.008). CONCLUSIONS: Deep learning diagnostic model with FDG-PET imaging data can be useful as one of supportive tools to determine the HPV status in patients with OPSCC.