Interobserver variability in chest CT and whole body FDG-PET screening for distant metastases in head and neck cancer patients

Evaluation of an Automatic Classification Algorithm Using Convolutional Neural Networks in Oncological Positron Emission Tomography

Introduction: Our aim was to evaluate the performance in clinical research and in clinical routine of a research prototype, called positron emission tomography (PET) Assisted Reporting System (PARS) (Siemens Healthineers) and based on a convolutional neural network (CNN), which is designed to detect suspected cancer sites in fluorine-18 fluorodeoxyglucose (¹⁸F-FDG) PET/computed tomography (CT).

Method: We retrospectively studied two cohorts of patients. The first cohort consisted of research-based patients who underwent PET scans as part of the initial workup for diffuse large B-cell lymphoma (DLBCL). The second cohort consisted of patients who underwent PET scans as part of the evaluation of miscellaneous cancers in clinical routine. In both cohorts, we assessed the correlation between manually and automatically segmented total metabolic tumor volumes (TMTVs), and the overlap between both segmentations (Dice score). For the research cohort, we also compared the prognostic value for progression-free survival (PFS) and overall survival (OS) of manually and automatically obtained TMTVs.

Results: For the first cohort (research cohort), data from 119 patients were retrospectively analyzed. The median Dice score between automatic and manual segmentations was 0.65. The intraclass correlation coefficient between automatically and manually obtained TMTVs was 0.68. Both TMTV results were predictive of PFS (hazard ratio: 2.1 and 3.3 for automatically based and manually based TMTVs, respectively) and OS (hazard ratio: 2.4 and 3.1 for automatically based and manually based TMTVs, respectively). For the second cohort (routine cohort), data from 430 patients were retrospectively analyzed. The median Dice score between automatic and manual segmentations was 0.48. The intraclass correlation coefficient between automatically and manually obtained TMTVs was 0.61.

Conclusion: The TMTVs determined for the research cohort remain predictive of total and PFS for DLBCL. However, the segmentations and TMTVs determined automatically by the algorithm need to be verified and, sometimes, corrected to be similar to the manual segmentation.

18F-FDG PET/CT Uptake Classification in Lymphoma and Lung Cancer by Using Deep Convolutional Neural Networks

Background Fluorine 18 (¹⁸F)-fluorodeoxyglucose (FDG) PET/CT is a routine tool for staging patients with lymphoma and lung cancer. Purpose To evaluate configurations of deep convolutional neural networks (CNNs) to localize and classify uptake patterns of whole-body ¹⁸F-FDG PET/CT images in patients with lung cancer and lymphoma. Materials and Methods This was a retrospective analysis of consecutive patients with lung cancer or lymphoma referred to a single center from August 2011 to August 2013. Two nuclear medicine experts manually delineated foci with increased ¹⁸F-FDG uptake, specified the anatomic location, and classified these findings as suspicious for tumor or metastasis or nonsuspicious. By using these expert readings as the reference standard, a CNN was developed to detect foci positive for ¹⁸F-FDG uptake, predict the anatomic location, and determine the expert classification. Examinations were divided into independent training (60%), validation (20%), and test (20%) subsets. Results This study included 629 patients (mean age, 52.2 years ± 20.4 [standard deviation]; 394 men). There were 302 patients with lung cancer and 327 patients with lymphoma. For the test set (123 patients; 10 782 foci), the CNN areas under the receiver operating characteristic curve (AUCs) for determining hypermetabolic ¹⁸F-FDG PET/CT foci that were suspicious for cancer versus nonsuspicious by using the five input features were as follows: CT alone, 0.78 (95% confidence interval [CI]: 0.72, 0.83); ¹⁸F-FDG PET alone, 0.97 (95% CI: 0.97, 0.98); ¹⁸F-FDG PET/CT, 0.98 (95% CI: 0.97, 0.99); ¹⁸F-FDG PET/CT maximum intensity projection (MIP), 0.98 (95% CI: 0.98, 0.99); and ¹⁸F-FDG PET/CT MIP atlas, 0.99 (95% CI: 0.98, 1.00). The combination of ¹⁸F-FDG PET and CT information improved overall classification accuracy (AUC, 0.975 vs 0.981, respectively; P < .001). Anatomic localization accuracy of the CNN was 2543 of 2639 (96.4%; 95% CI: 95.5%, 97.1%) for body part, 2292 of 2639 (86.9%; 95% CI: 85.3%, 88.5%) for region (ie, organ), and 2149 of 2639 (81.4%; 95% CI: 79.3%-83.5%) for subregion. Conclusion The fully automated anatomic localization and classification of fluorine 18-fluorodeoxyglucose PET uptake patterns in foci suspicious and nonsuspicious for cancer in patients with lung cancer and lymphoma by using a convolutional neural network is feasible and achieves high diagnostic performance when both CT and PET images are used. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Froelich and Salavati in this issue.

The Role of 18-Fluoro-2-Deoxy-Glucose Positron Emission Tomography/Computed Tomography as Response and Prognosis Predictive Factor of Concurrent Chemoradiotherapy after Induction Chemotherapy in Head and Neck Squamous Cell Carcinoma: A Prospective Study

Background and objectives: The importance of induction chemotherapy (ICT) followed by concurrent chemoradiotherapy (CCRT) has been re-established in recent years aiming at fewer metastatic sites and better control of the disease. We prospectively studied the possibility of early prediction of overall survival (OS) and progression-free survival (PFS) after 3 cycles of chemotherapy with doxetacel, cisplatin and 5-fluorouracil using 18-fluoro-2-deoxy-glucose positron emission tomography computed tomography (¹⁸F-FDG PET/CT) in patients with head and neck squamous cell cancer. To our knowledge, this is the first such study. Materials and Methods: Thirty-five patients were studied. They underwent an ¹⁸F-FDG PET/CT examination twice: a day before ICT and 10–14 days after the last cycle of ICT. Tumor-standardized uptake value (SUVmax) and hypermetabolic tumor volume were measured on both scans. The mean age of patients was 56.5 years. Complete responses to CCRT PFS and OS were calculated. Results: Our results showed that a decrease of ≥30% in the SUVmax value after ICT was a prognostic factor of tumor response to PFS and OS (p = 0.026 and p = 0.021). The groups of patients with a SUVmax between 10 and 14.5 in the primary tumor on a pre-ICT ¹⁸F-FDG PET/CT scan had statistically shorter PFS and OS (p = 0.001, p = 0.006) when compared with other groups of patients with SUVmax less than 10 or SUVmax more than 14.5. A decrease of less than 55% of hypermetabolic tumor volume of the primary tumor was significantly related to poor prognosis in PFS and OS (p = 0.033, p = 0.017). Conclusions: SUVmax and hypermetabolic tumor volume measured on ¹⁸F-FDG PET/CT after ICT might be valuable prognostic tools for predicting OS and PFS and, thus, for the selection of patients with head and neck cancer who will benefit from CCRT.

18F-FDG-PET/CT based total metabolic tumor volume change during neoadjuvant chemotherapy predicts outcome in advanced epithelial ovarian cancer

OBJECTIVE

To evaluate the predictive potential of total metabolic tumor volume (MTV) reduction during neoadjuvant chemotherapy (NACT) with ¹⁸F-FDG-PET/CT in an advanced FIGO stage III/IV epithelial ovarian cancer (EOC) patient cohort.

METHODS

Twenty-nine primarily inoperable EOC patients underwent ¹⁸F-FDG-PET/CT before and after NACT. The pre- and post-NACT total MTV, in addition to the percentage MTV reduction during NACT, were compared with primary therapy outcome and progression-free survival (PFS). ROC-analysis determined an optimal threshold for MTV reduction identifying patients with progressive or stable disease (PD/SD) at the end of primary therapy. A multivariate analysis with residual tumor (0/>0), FIGO stage (III/IV) and MTV reduction compared to PFS was performed. The association between MTV reduction and overall survival (OS) was evaluated.

RESULTS

The median pre- and post-NACT total MTV were 352 cm³ (range 150 to 1322 cm³) and 51 cm³ (range 0 to 417 cm³), respectively. The median MTV reduction during NACT was 89% (range 24% to 100%). Post-NACT MTV and MTV reduction associated with primary therapy outcome (MTV post-NACT p = 0.007, MTV reduction p = 0.001) and PFS (MTV post-NACT p = 0.005, MTV reduction p = 0.005). MTV reduction <85% identified the PD/SD patients (sensitivity 70%, specificity 78%, AUC 0.79). In a multivariate analysis, MTV reduction (p = 0.002) and FIGO stage (p = 0.003) were statistically significant variables associated with PFS. MTV reduction during NACT corresponded to OS (p = 0.05).

CONCLUSION

¹⁸F-FDG-PET/CT is helpful in NACT response evaluation. Patients with total MTV reduction <85% during NACT might be candidates for second-line chemotherapy and clinical trials, instead of interval debulking surgery.

Prospective head-to-head comparison of 11C-choline-PET/MR and 11C-choline-PET/CT for restaging of biochemical recurrent prostate cancer

PURPOSE

Whole-body integrated ¹¹C-choline PET/MR might provide advantages compared to ¹¹C-choline PET/CT for restaging of prostate cancer (PC) due to the high soft-tissue contrast and the use of multiparametric MRI, especially for detection of local recurrence and bone metastases.

MATERIALS AND METHODS

Ninety-four patients with recurrent PC underwent a single-injection/dual-imaging protocol with contrast-enhanced PET/CT followed by fully diagnostic PET/MR. Imaging datasets were read separately by two reader teams (team 1 and 2) assessing the presence of local recurrence, lymph node and bone metastases in predefined regions using a five-point scale. Detection rates were calculated. The diagnostic performance of PET/CT vs. PET/MR was compared using ROC analysis. Inter-observer and inter-modality variability, radiation exposure, and mean imaging time were evaluated. Clinical follow-up, imaging, and/or histopathology served as standard of reference (SOR).

RESULTS

Seventy-five patients qualified for the final image analysis. A total of 188 regions were regarded as positive: local recurrence in 37 patients, 87 regions with lymph node metastases, and 64 regions with bone metastases. Mean detection rate between both readers teams for PET/MR was 84.7% compared to 77.3% for PET/CT (p > 0.05). Local recurrence was identified significantly more often in PET/MR compared to PET/CT by team 1. Lymph node and bone metastases were identified significantly more often in PET/CT compared to PET/MR by both teams. However, this difference was not present in the subgroup of patients with PSA values ≤2 ng/ml. Inter-modality and inter-observer agreement (K > 0.6) was moderate to substantial for nearly all categories. Mean reduction of radiation exposure for PET/MR compared to PET/CT was 79.7% (range, 72.6-86.2%). Mean imaging time for PET/CT was substantially lower (18.4 ± 0.7 min) compared to PET/MR (50.4 ± 7.9 min).

CONCLUSIONS

¹¹C-choline PET/MR is a robust imaging modality for restaging biochemical recurrent PC and interpretations between different readers are consistent. It provides a higher diagnostic value for detecting local recurrence compared to PET/CT with the advantage of substantial dose reduction. Drawbacks of PET/MR are a substantially longer imaging time and a slight inferiority in detecting bone and lymph node metastases in patients with PSA values >2 ng/ml. Thus, we suggest the use of ¹¹C-choline PET/MR especially for patients with low (≤2 ng/ml) PSA values, whereas PET/CT is preferable in the subgroup with higher PSA values.

Preoperative 18F-FDG-PET/CT vs Contrast-Enhanced CT to Identify Regional Nodal Metastasis among Patients with Head and Neck Squamous Cell Carcinoma

Objective Our objective was to compare the accuracy of preoperative positron emission tomography (PET)/computed tomography (CT) and contrast-enhanced CT (CECT) in detecting cervical nodal metastases in patients treated with neck dissection and to scrutinize the ability of each modality to determine nodal stage. Study Design Case series with chart review. Setting Montefiore Medical Center, Bronx, New York. Subjects and Methods Patients who underwent neck dissection at our institution for primary treatment of head and neck squamous cell carcinoma (HNSCC) and had received preoperative PET/CT and CECT were included in this study. Imaging studies were reinterpreted by 3 specialists within the field and compared for interreader agreement. Concordance between radiology and histopathology was measured using neck levels and sides, along with patient nodal stage. Sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and agreement coefficients were calculated. Results Seventy-three patients were included in the study. Sensitivity was 0.69 and 0.94 (level and side) for PET/CT vs 0.53 and 0.66 for CECT ( P = .056, P = .001). Specificity was 0.86 and 0.56 for PET/CT vs 0.91 and 0.76 for CECT ( P = .014, P = .024). No significant difference was found in overall accuracy ( P = .33, P = .88). The overall agreement percentages between N stage called by imaging modality and pathology were 52% and 55% for PET/CT and CECT, respectively. Conclusion No significant difference in sensitivity was found between PET/CT and CECT. CECT was found to have superior specificity compared with PET/CT. The information gleaned from each modality in the pretreatment evaluation of HNSCC appears to be complementary.

Screening for distant metastases in head and neck cancer patients using FDG-PET and chest CT: validation of an algorithm

In patients with head and neck squamous cell carcinoma and high-risk factors, the combination of whole body FDG-PET and contrast-enhanced chest CT has the highest sensitivity and accuracy when screening for distant metastases. The aim of the present study was to retrospectively validate an earlier developed algorithm for interpreting the combination of screening PET and CT. The test cohort consisted of 47 consecutive HNSCC patients with high-risk factors for distant metastases, who had previously undergone FDG-PET and CT and had a minimum 12 months of follow-up. In 12 (26 %) patients, distant metastases were detected during screening or within 12-month follow-up. In patients with locoregional control during follow-up, the sensitivity and specificity were 55 % (95 % CI 23-83 %) and 97 % (95 % CI 82-99 %), respectively, for chest CT, 55 % (95 % CI 23-83 %) and 100 % (95 % CI 88-100 %), respectively, for PET and 73 % (95 % CI 39-94 %) and 100 % (95 % CI 88-100 %), respectively, for the combination of PET and CT. The proposed algorithm was considered to have been validated. In this algorithm, all FDG-PET positive scans for distant metastases (regardless of interpretation of a solid lung lesion on CT) and CT scans with suspicious pulmonary lesions of less than 5-mm diameter (regardless of FDG-PET findings) are considered positive for distant metastases.

[18F]-FDG PET-CT prediction of response to induction chemotherapy in head and neck squamous cell carcinoma: preliminary findings

OBJECTIVES

The present study assessed the predictive value of [18F]-FDG PET-CT (positron emission tomography with 18-fluoro-eoxyglucose radiotracer, coupled to computerized tomography) for response to induction chemotherapy in head and neck squamous cell carcinoma (HNSCC).

MATERIAL AND METHODS

[18F]-FDG PET-CT was systematically performed before treatment initiation and after the first cycle of chemotherapy. Results were compared with those of endoscopy and pathologic analysis of biopsy and surgical specimens.

RESULTS

This preliminary study included 21 previously untreated HNSCC patients. A decrease of more than 30% in SUVmax (maximum standard uptake value) during induction was predictive of tumor response to chemotherapy (P=0.04). PET-CT measurement of hypermetabolic volume based on a predetermined SUV threshold (SUV=2.5), on the other hand, proved non-predictive.

CONCLUSION

These preliminary findings are promising. A larger sample, however, would be required in order to determine a more precise SUVmax reduction cut-off threshold during induction. Other methods for determining metabolic volume thresholds will be investigated. If functional imaging proves contributive, it could enable early screening of non-responders, avoiding unnecessary intoxication.

Contemporary management of lymph node metastases from an unknown primary to the neck: I. A review of diagnostic approaches

In an era of advanced diagnostics, metastasis to cervical lymph nodes from an occult primary tumor is a rare clinical entity and accounts for approximately 3% of head and neck malignancies. Histologically, two thirds of cases are squamous cell carcinomas (SCCs), with other tissue types less common in the neck. With modern imaging and tissue examinations, a primary tumor initially undetected on physical examination is revealed in >50% of patients and the site of the index primary can be predicted with a high level of probability. In the present review, the range and limitations of diagnostic procedures are summarized and the optimal diagnostic workup is proposed. Initial preferred diagnostic procedures are a fine-needle aspiration biopsy (FNAB) and imaging. This allows directed surgical biopsy (such as tonsillectomy), based on the preliminary findings, and prevents misinterpretation of postsurgical images. When no primary lesion is suggested after imaging and panendoscopy, and for patients without a history of smoking and alcohol abuse, molecular profiling of an FNAB sample for human papillomavirus (HPV) and/or Epstein-Barr virus (EBV) is important.