No abstract

Optimization of injected gallium-68 ( 68 Ga) activity for 68 Ga-prostate-specific membrane antigen positron emission tomography/computed tomography ( 68 Ga-PSMA PET/CT) studies is relevant for image quality, radiation protection, and from an economic point of view. However, no clear guidelines are available for 68 Ga-PSMA studies. Therefore, a phantom study is performed to determine the highest coefficient of variation (COV) acceptable for reliable image interpretation and quantification. To evaluate image interpretation, the relationship of COV and contrast-to-noise ratio (CNR) was studied. The CNR should remain larger than five, according to the Rose criterion. To evaluate image quantification, the effect of COV on the percentage difference (PD) between quantification results of two studies was analyzed. Comparison was done by calculating the PD of the SUV max . The maximum allowable PD SUVmax was set at 20%. The highest COV at which both criteria are still met is defined as COV max . Of the NEMA Image Quality phantom, a 20 min/bed (2 bed positions) scan was acquired in list-mode PET (Philips Gemini TF PET/CT). The spheres to background activity ratio was approximately 9:1. To obtain images with different COV, lower activity was mimicked by reconstructions with acquisition times of 10 min/bed to 5 s/bed. Pairs of images were obtained by reconstruction of two non-overlapping parts of list-mode data. For the 10-mm diameter sphere, a COV of 25% still meets the criteria of CNR SUVmean ≥ 5 and PD SUVmax ≤ 20%. This phantom scan was acquired with an acquisition time of 116 s and a background activity concentration of 0.71 MBq/kg. Translation to a clinical protocol results in a clinical activity regimen of 3.5 MBq/kg min at injection. To verify this activity regimen, 15 patients (6 MBq/kg min) with a total of 22 lesions are included. Additional reconstructions were made to mimic the proposed activity regimen. Based on the CNR SUVmax , no lesions were missed with this proposed activity regimen. For our institution, a clinical activity regimen of 3.5 MBq/kg min at injection is acceptable, which indicates that activity can be reduced by almost 50% compared with the current code of practice. Our proposed method could be used to obtain an objective activity regimen for other PET/CT systems and tracers.

Purpose: Dopamine transporter (DAT) imaging with 123I-FP-CIT SPECT is used to support the diagnosis of Parkinson’s disease (PD) in clinically uncertain cases. Previous studies showed that automatic classification of 123I‑FP‑CIT SPECT images (marketed as DaTSCAN) is feasible by using machine learning algorithms. However, these studies lacked sizable use of data from routine clinical practice. This study aims to contribute to the discussion whether artificial intelligence (AI) can be applied in clinical practice. Moreover, we investigated the need for hospital specific training data.Methods: A convolutional neural network (CNN) named DaTNet-3 was designed and trained to classify DaTSCAN images as either normal or supportive of a dopaminergic deficit. Both a multi-site data set (n = 2412) from the Parkinson’s Progression Marker Initiative (PPMI) and an in-house data set containing clinical images (n = 932) obtained in routine practice at the St Antonius hospital (STA) were used for training and testing. STA images were labeled based on interpretation by nuclear medicine physicians. To investigate whether indeterminate scans effects classification accuracy, a threshold was applied on the output probability.Results: DaTNet-3 trained with STA data reached an accuracy of 89.0% in correctly identifying images of the clinical STA test set as either normal or with decreased striatal DAT binding (98.5% on the PPMI test set). When thresholded, accuracy increased to 95.7%. This increase was not observed when trained with PPMI data, indicating the incorrect images were confidently classified as the incorrect class.Conclusion: Based on results of DaTNet-3 we conclude that automatic interpretation of DaTSCAN images with AI is feasible and robust. Further, we conclude DaTNet-3 performs slightly better when it is trained with hospital specific data. This difference increased when output probability was thresholded. Therefore we conclude that the usability of a data set increases if it contains indeterminate images.

Purpose Dopamine transporter (DAT) imaging with 123I-FP-CIT SPECT is used to support the diagnosis of Parkinson’s disease (PD) in clinically uncertain cases. Previous studies showed that automatic classification of 123IFPCIT SPECT images (marketed as DaTSCAN) is feasible by using machine learning algorithms. However, these studies lacked sizable use of data from routine clinical practice. This study aims to contribute to the discussion whether artificial intelligence (AI) can be applied in clinical practice. Moreover, we investigated the need for hospital specific training data. Methods A convolutional neural network (CNN) named DaTNet-3 was designed and trained to classify DaTSCAN images as either normal or supportive of a dopaminergic deficit. Both a multi-site data set (n = 2412) from the Parkinson’s Progression Marker Initiative (PPMI) and an in-house data set containing clinical images (n = 932) obtained in routine practice at the St Antonius hospital (STA) were used for training and testing. STA images were labeled based on interpretation by nuclear medicine physicians. To investigate whether indeterminate scans effects classification accuracy, a threshold was applied on the output probability. Results DaTNet-3 trained with STA data reached an accuracy of 89.0% in correctly identifying images of the clinical STA test set as either normal or with decreased striatal DAT binding (98.5% on the PPMI test set). When thresholded, accuracy increased to 95.7%. This increase was not observed when trained with PPMI data, indicating the incorrect images were confidently classified as the incorrect class. Conclusion Based on results of DaTNet-3 we conclude that automatic interpretation of DaTSCAN images with AI is feasible and robust. Further, we conclude DaTNet-3 performs slightly better when it is trained with hospital specific data. This difference increased when output probability was thresholded. Therefore we conclude that the usability of a data set increases if it contains indeterminate images.

No abstract