Is quantitation necessary for oncological PET studies? Against

Convolutional Neural Network Detection of Axillary Lymph Node Metastasis Using Standard Clinical Breast MRI

BACKGROUND

Axillary lymph node status is important for breast cancer staging and treatment planning as the majority of breast cancer metastasis spreads through the axillary lymph nodes. There is currently no reliable noninvasive imaging method to detect nodal metastasis associated with breast cancer.

MATERIALS AND METHODS

Magnetic resonance imaging (MRI) data were those from the peak contrast dynamic image from 1.5 Tesla MRI scanners at the pre-neoadjuvant chemotherapy stage. Data consisted of 66 abnormal nodes from 38 patients and 193 normal nodes from 61 patients. Abnormal nodes were those determined by expert radiologist based on ¹⁸Fluorodeoxyglucose positron emission tomography images. Normal nodes were those with negative diagnosis of breast cancer. The convolutional neural network consisted of 5 convolutional layers with filters from 16 to 128. Receiver operating characteristic analysis was performed to evaluate prediction performance. For comparison, an expert radiologist also scored the same nodes as normal or abnormal.

RESULTS

The convolutional neural network model yielded a specificity of 79.3% ± 5.1%, sensitivity of 92.1% ± 2.9%, positive predictive value of 76.9% ± 4.0%, negative predictive value of 93.3% ± 1.9%, accuracy of 84.8% ± 2.4%, and receiver operating characteristic area under the curve of 0.91 ± 0.02 for the validation data set. These results compared favorably with scoring by radiologists (accuracy of 78%).

CONCLUSION

The results are encouraging and suggest that this approach may prove useful for classifying lymph node status on MRI in clinical settings in patients with breast cancer, although additional studies are needed before routine clinical use can be realized. This approach has the potential to ultimately be a noninvasive alternative to lymph node biopsy.

Quantifying and reducing the effect of calibration error on variability of PET/CT standardized uptake value measurements

The purpose of this study was to measure the errors introduced by regular calibration of PET/CT scanners and to minimize the effect of calibration error on standardized uptake value measurements.

METHODS

Global calibration factors from 2 PET/CT scanners were recorded for 3.5 and 1.8 y, comparing manufacturer-recommended protocols with modified protocols to evaluate error contributions due to operator-influenced procedures. Dose calibrator measurements were evaluated using National Institute of Standards and Technology-traceable sources.

RESULTS

Dose calibrator variability was less than 1%, although there was a consistent bias. Global scaling variability was reduced from 6% to 4% for scanner 1 and from 11% to 4% for scanner 2 when quality assurance and quality control procedures were applied to the calibration protocol. When calibrations were done using a (68)Ge/(68)Ga phantom, the variability for both scanners was reduced to approximately 3%.

CONCLUSION

Applying quality assurance and quality control procedures to scanner calibration reduces variability, but there is a still a residual longitudinal scanner variability of 3%-4%. The procedures proposed here reduce the impact of operator error on scanner calibration and thereby minimize longitudinal variability in standardized uptake value measurements.

The reduction of artifacts due to metal hip implants in CT-attenuation corrected PET images from hybrid PET/CT scanners

CT beam hardening artifacts near metal hip implants may erroneously enhance or diminish radiotracer uptake following CT attenuation correction (AC) of PET images. An artifact reduction algorithm (ARA) was developed to reduce metal artifacts in CT-based AC-PET. The algorithm employed a Bayes classifier to identify beam-hardening artifacts, followed by a partial correction of the attenuation map. ARA was implemented on phantom and patient 18F-FDG studies using a clinical PET/CT scanner. In phantom studies ARA successfully removed two artifacts of erroneously elevated uptake near a stainless steel hip prosthesis which were depicted in the standard CT-AC PET. ARA has also identified two targets absent on the scanner PET images. Target-to-background ratios were 1.5-3 times higher for ARA-PET than scanner images. In a patient study, metal artifacts were of lower intensity in ARA-PET as compared to standard images. Potentially, ARA may improve detectability of small lesions located near metal hip implants.

A hybrid algorithm for PET/CT image merger in hybrid scanners

PURPOSE

To improve the PET image quality of a hybrid PET/CT scanner by merging CT borders with PET texture. PET/CT scanners provide both high-resolution CT images showing anatomical details and PET images of low-resolution physiological information about radiopharmaceutical uptake. Standard smoothing of noisy PET images may further impair PET resolution, reducing small lesion detectability.

METHODS

The CT edge data and the PET texture data were merged using a modified form of an algorithm called HCT (hybrid computed tomography). In merged PET/CT images, each PET pixel value was estimated by iteratively applying a corrected 2D Taylor expansion to each of its eight neighbors. The spatial derivative term was used only near anatomical edges provided by the CT. This counts-preserving algorithm was tested on a special resolution phantom and patient data sets obtained by PET/CT acquisitions.

RESULTS

The HCT algorithm provided phantom PET images with sharp borders and improved resolution (< or = 3 mm as compared to > or = 4 mm). HCT increased the signal to background contrast ratios by an average of 61% (40-89%) while maintaining noise reduction similar to the Gaussian filtering standard in PET. In the clinical PET images, HCT allowed for an improved delineation of pulmonary and pelvic lesions and an improved visualization of the brain.

CONCLUSION

A new reconstruction algorithm for merging CT anatomical edge data with functional PET data has been introduced. The algorithm smooths noisy PET images while retaining sharper edges at corresponding anatomical borders, resulting in an improvement in resolution and contrast ratio.

High Incidence of Chest Malignancy Detected by FDG PET in Patients Suspected of Recurrent Squamous Cell Carcinoma of the Upper Aerodigestive Tract

OBJECTIVE

To determine the incidence of chest neoplasms detected by FDG PET in patients with previously treated squamous cell head and neck cancer (HNC), being evaluated for possible recurrent disease.

METHODS

This is a retrospective review of 41 patients (M = 29, F = 12: average age = 58 years) with previously treated HNC who underwent FDG PET of the neck and chest as part of routine evaluation for locoregional and/or distant recurrence. Thirty-four of 41 patients had advanced stage III or IV HNC. All FDG PET studies were reviewed by dedicated nuclear medicine physicians, including evaluation for abnormal uptake in the chest. The chest FDG findings were correlated with serial chest radiographs or chest CT. The occurrence rate of incidental chest malignancy was determined and based on characteristic imaging findings, biopsy, and/or clinical course.

RESULTS

Twelve of 41 patients had abnormal FDG uptake in the lungs and/or mediastinum. Ten of 12 patients were found to have neoplasms that could represent either metastases or a new lung primary. Five of these 10 were unsuspected neoplasms prior to FDG PET. The other 2/12 FDG PET scans in the chest were false positive. There was one false-negative FDG PET, with subsequent PET and CT demonstrating pulmonary metastases. Overall, there was a 27% incidence of chest malignancies in patients with advanced HNC being evaluated for possible recurrence.

CONCLUSION

Our study demonstrated a chest malignancy in 1 out of 4 patients with advanced HNC being evaluated for locoregional and/or distant spread. Fifty percent were unsuspected prior to FDG PET. This result suggests that FDG PET of the lungs should be routinely included in the evaluation of high-risk patients.

Diagnostic accuracy of 2-[fluorine-18]fluro-2-deoxy-D-glucose positron emission tomography imaging in nonsquamous tumors of the head and neck

PURPOSE

To evaluate the diagnostic accuracy of 2-[fluorine-18]fluro-2-deoxy-D-glucose positron emission tomography (FDG-PET) for detecting nonsquamous tumors of the extracranial head and neck (NSTHN).

MATERIALS AND METHODS

The records of all patients with suspected or proven NSTHN undergoing PET imaging in our institution over a 12-year period were reviewed. Forty-four patients (24 males and 20 females; age range, 6-81 years; mean age, 51.2 years) were classified into 3 main groups: thyroid tumors (n = 19), salivary gland tumors (n = 7), and miscellaneous lesions (n = 18). The PET findings for each individual group with respect to the primary site, cervical nodal and distant metastases, were correlated with histopathology or follow-up (2-year minimum)

RESULTS

The overall diagnostic accuracy of FDG-PET for all NSTHN was 86%, However, the diagnostic accuracy varied for the histologic subtype (thyroid, salivary, miscellaneous) and the area being evaluated (primary site, nodal metastases, distant metastases).

CONCLUSION

There is variable diagnostic accuracy of FDG-PET in detecting different histologic subtypes of NSTHN. This information should be taken into consideration when considering PET for evaluation of NSTHN.

[F-18-FDG positron emission tomography in the diagnosis of ovarian recurrence. Comparison with CT scan and CA 125]

OBJECTIVE

The aim of this study was to evaluate the clinical utility of FDG-PET for detecting recurrent disease in patients with ovarian cancer.

MATERIAL AND METHODS

Twenty-one FDG-PET studies performed in 19 patients who had previously undergone surgery and chemotherapy for ovarian cancer were reviewed retrospectively. In a maximum interval of one week regarding the FDG-PET study, computed tomography (CT) was performed and CA-125 levels were measured. In 16 cases the relapse suspicion was due to elevation of the tumor marker CA125 and in 5 cases it was due to CT. PET images were obtained at 45 min after the intravenous injection of 370 MBq of FDG. The results of the visual interpretation were compared with the CA125 levels and the images of the CT, and related to the definitive diagnosis. Recurrence was confirmed in 19/21 cases, by means of pathological findings (11 cases) and clinical follow-up for a median of 11 months in the others.

RESULTS

Recurrence was confirmed in 16 cases with increased CA-125 and the tumor marker was true-negative in 2 disease free cases, but there were 3 false-negative results (sensitivity of 84 % and accuracy of 86 %). CT correctly identified 9 cases with recurrence, but it was false-negative in 10 cases and false-positive in 2 disease free patients (sensitivity of 47 % and accuracy of 43 %). FDG-PET correctly detected the 19 cases with recurrence but it was false-positive in 2 cases with a sensitivity of 100 % and accuracy of 90 %. In 3 patients with CA125 false-negative and 10 patients with false-negative CT, FDG-PET was positive and recurrence was confirmed.

CONCLUSION

These preliminary results suggest that in the follow-up of patients with ovarian cancer FDG-PET could detect recurrence with higher accuracy than CT, and even with higher sensitivity than the tumor marker CA125, being useful at the same time to locate the recurrence when the tumor marker is positive.