A machine learning tool to improve prediction of mediastinal lymph node metastases in non-small cell lung cancer using routinely obtainable [18F]FDG-PET/CT parameters.

BACKGROUND: In patients with non-small cell lung cancer (NSCLC), accuracy of [18F]FDG-PET/CT for pretherapeutic lymph node (LN) staging is limited by false positive findings. Our aim was to evaluate machine learning with routinely obtainable variables to improve accuracy over standard visual image assessment. METHODS: Monocentric retrospective analysis of pretherapeutic [18F]FDG-PET/CT in 491 consecutive patients with NSCLC using an analog PET/CT scanner (training + test cohort, n = 385) or digital scanner (validation, n = 106). Forty clinical variables, tumor characteristics, and image variables (e.g., primary tumor and LN SUVmax and size) were collected. Different combinations of machine learning methods for feature selection and classification of N0/1 vs. N2/3 disease were compared. Ten-fold nested cross-validation was used to derive the mean area under the ROC curve of the ten test folds ("test AUC") and AUC in the validation cohort. Reference standard was the final N stage from interdisciplinary consensus (histological results for N2/3 LNs in 96%). RESULTS: N2/3 disease was present in 190 patients (39%; training + test, 37%; validation, 46%; p = 0.09). A gradient boosting classifier (GBM) with 10 features was selected as the final model based on test AUC of 0.91 (95% confidence interval, 0.87-0.94). Validation AUC was 0.94 (0.89-0.98). At a target sensitivity of approx. 90%, test/validation accuracy of the GBM was 0.78/0.87. This was significantly higher than the accuracy based on "mediastinal LN uptake > mediastinum" (0.7/0.75; each p < 0.05) or combined PET/CT criteria (PET positive and/or LN short axis diameter > 10 mm; 0.68/0.75; each p < 0.001). Harmonization of PET images between the two scanners affected SUVmax and visual assessment of the LNs but did not diminish the AUC of the GBM. CONCLUSIONS: A machine learning model based on routinely available variables from [18F]FDG-PET/CT improved accuracy in mediastinal LN staging compared to established visual assessment criteria. A web application implementing this model was made available.

A convolutional neural network with self-attention for fully automated metabolic tumor volume delineation of head and neck cancer in [Formula: see text]F]FDG PET/CT.

PURPOSE: PET-derived metabolic tumor volume (MTV) and total lesion glycolysis of the primary tumor are known to be prognostic of clinical outcome in head and neck cancer (HNC). Including evaluation of lymph node metastases can further increase the prognostic value of PET but accurate manual delineation and classification of all lesions is time-consuming and prone to interobserver variability. Our goal, therefore, was development and evaluation of an automated tool for MTV delineation/classification of primary tumor and lymph node metastases in PET/CT investigations of HNC patients. METHODS: Automated lesion delineation was performed with a residual 3D U-Net convolutional neural network (CNN) incorporating a multi-head self-attention block. 698 [Formula: see text]F]FDG PET/CT scans from 3 different sites and 5 public databases were used for network training and testing. An external dataset of 181 [Formula: see text]F]FDG PET/CT scans from 2 additional sites was employed to assess the generalizability of the network. In these data, primary tumor and lymph node (LN) metastases were interactively delineated and labeled by two experienced physicians. Performance of the trained network models was assessed by 5-fold cross-validation in the main dataset and by pooling results from the 5 developed models in the external dataset. The Dice similarity coefficient (DSC) for individual delineation tasks and the primary tumor/metastasis classification accuracy were used as evaluation metrics. Additionally, a survival analysis using univariate Cox regression was performed comparing achieved group separation for manual and automated delineation, respectively. RESULTS: In the cross-validation experiment, delineation of all malignant lesions with the trained U-Net models achieves DSC of 0.885, 0.805, and 0.870 for primary tumor, LN metastases, and the union of both, respectively. In external testing, the DSC reaches 0.850, 0.724, and 0.823 for primary tumor, LN metastases, and the union of both, respectively. The voxel classification accuracy was 98.0% and 97.9% in cross-validation and external data, respectively. Univariate Cox analysis in the cross-validation and the external testing reveals that manually and automatically derived total MTVs are both highly prognostic with respect to overall survival, yielding essentially identical hazard ratios (HR) ([Formula: see text]; [Formula: see text] vs. [Formula: see text]; [Formula: see text] in cross-validation and [Formula: see text]; [Formula: see text] vs. [Formula: see text]; [Formula: see text] in external testing). CONCLUSION: To the best of our knowledge, this work presents the first CNN model for successful MTV delineation and lesion classification in HNC. In the vast majority of patients, the network performs satisfactory delineation and classification of primary tumor and lymph node metastases and only rarely requires more than minimal manual correction. It is thus able to massively facilitate study data evaluation in large patient groups and also does have clear potential for supervised clinical application.

An optimized imaging protocol for [99mTc]Tc-DPD scintigraphy and SPECT/CT quantification in cardiac transthyretin (ATTR) amyloidosis.

BACKGROUND: In [99mTc]Tc-DPD scintigraphy for myocardial ATTR amyloidosis, planar images 3 hour p.i. and SPECT/CT acquisition in L-mode are recommended. This study investigated if earlier planar images (1 hour p.i.) are beneficial and if SPECT/CT acquisition should be preferred in H-mode (180° detector angle) or L-mode (90°). METHODS: In SPECT/CT phantom measurements (NaI cameras, N = 2; CZT, N = 1), peak contrast recovery (CRpeak) was derived from sphere inserts or myocardial insert (cardiac phantom; signal-to-background ratio [SBR], 10:1 or 5:1). In 25 positive and 38 negative patients (reference: endomyocardial biopsy or clinical diagnosis), Perugini scores and heart-to-contralateral (H/CL) count ratios were derived from planar images 1 hour and 3 hour p.i. RESULTS: In phantom measurements, accuracy of myocardial CRpeak at SBR 10:1 (H-mode, 0.95-0.99) and reproducibility at 5:1 (H-mode, 1.02-1.14) was comparable for H-mode and L-mode. However, L-mode showed higher variability of background counts and sphere CRpeak throughout the field of view than H-mode. In patients, sensitivity/specificity were ≥ 95% for H/CL ratios at both time points and visual scoring 3 hour. At 1 hour, visual scores showed specificity of 89% and reduced reader's confidence. CONCLUSIONS: Early DPD images provided no additional value for visual scoring or H/CL ratios. In SPECT/CT, H-mode is preferred over L-mode, especially if quantification is applied apart from the myocardium.

Assessment of the tibial slope is highly dependent on the type and accuracy of the preceding acquisition.

BACKGROUND: Precise measurement of the tibial slope (TS) is crucial for realignment surgery, ligament reconstruction, and arthroplasty. However, there is little consensus on the ideal assessment. It was hypothesized that the tibial slope changes according to the acquisition technique and both tibial length as well as femoral rotation serve as potential confounders. METHODS: 104 patients (37 women, 67 men; range 12-66 years) were retrospectively selected, of which all patients underwent a 1.5-Tesla MRI and either additional standard lateral radiographs (SLR, n = 52) or posterior stress radiographs (PSR, n = 52) of the index knee. Two blinded observers evaluated the medial tibial slope as the medial TS is primarily used in clinical practice. Additionally, the length of the diaphyseal axis and the extent of radiographic malrotation were measured. RESULTS: Mean TS on MRI was significantly lower compared to radiographs (4.2° ± 2.9° vs. 9.1° ± 3.6°; p < 0.0001). There was a significant correlation between MRI and PSR (p < 0.0001 with r = 0.7), but not with SLR (p = 0.93 with r = 0.24). Tibial length was a significant predictor for the difference between MRI and SLR (regression coefficient ß = - 0.03; p = 0.035), yet not between MRI and PSR (ß = - 0.003; p = 0.9). Femoral rotation proved to be a significant predictor for the agreement between both observers (PSR: ß = 0.14; p = 0.001 and SLR: ß = 0.08; p = 0.04). ICC indicated a high interrater agreement for the radiographic assessment (ICC ≥ 0.72). CONCLUSIONS: There is a substantial variance between MRI and radiographic measurement of the tibial slope. However, as MRI assessment is time-consuming and requires specialized software, instrumented radiographs might be an alternative. Due care has to be taken to ensure that radiographs contain a sufficient tibial length, and femoral rotation is avoided. STUDY DESIGN: Case series (diagnosis); Level of evidence, 4.

Ga-68-PSMA PET/CT in treatment-naïve patients with prostate cancer: Which clinical parameters and risk stratification systems best predict PSMA-positive metastases?

PURPOSE: To evaluate the accuracy of clinical parameters and established pre-treatment risk stratification systems for prostate cancer (PCa) in predicting PSMA-positive metastases in men undergoing Ga-68-PSMA PET/CT as initial staging examination. MATERIALS AND METHODS: A retrospective analysis in 108 consecutive treatment-naïve patients with biopsy-proven PCa undergoing Ga-68-PSMA PET/CT (median age, 72 years [range, 49-82 years]) was performed. Prediction of PSMA-positive metastases by serum PSA, clinical T stage (cT), ISUP group, percentage of positive biopsy cores, and derived risk scores (D'Amico risk classification system, Roach [RF], Yale formula [YF], and Briganti nomogram [BN]) was examined with ROC analysis. RESULTS: Any PSMA-positive metastases were found in 36 of 108 patients, including LN metastases in 28 patients, extrapelvic LN metastases in 15 patients, and organ metastases in 19 patients (bone, 19; lung, 1). AUCs for PSA, cT, ISUP, and percentage of positive biopsy cores regarding PSMA-positive metastases did not differ significantly (range, 0.6-0.8; each P > 0.05). D'Amico (AUC, 0.61-0.64) was inferior to RF (0.76-0.83), YF (0.81-0.86), and BN (0.73 to 0.88; each P < 0.05). Among the 89 high-risk patients (D'Amico), decision for or against PET imaging based on RF (cut-off, >18.0), YF (>10.8), or BN (>8.0) would have prevented PSMA PET/CT in 4 (5%), 15 (17%), or 18 patients (20%), respectively, while preserving a sensitivity ≥95% for PSMA-positive metastases. CONCLUSIONS: Clinical parameters and established risk stratification systems for PCa can predict Ga-68-PSMA PET-positive metastases in treatment-naïve patients. Especially YF and BN may improve identification of patients with the highest probability of metastatic disease detected by Ga-68-PSMA PET/CT.

Pretherapeutic FDG-PET total metabolic tumor volume predicts response to induction therapy in pediatric Hodgkin's lymphoma.

BACKGROUND: Standardized treatment in pediatric patients with Hodgkin's lymphoma (HL) follows risk stratification by tumor stage, erythrocyte sedimentation rate and tumor bulk. We aimed to identify quantitative parameters from pretherapeutic FDG-PET to assist prediction of response to induction chemotherapy. METHODS: Retrospective analysis in 50 children with HL (f:18; m:32; median age, 14.8 [4-18] a) consecutively treated according to EuroNet-PHL-C1 (n = 42) or -C2 treatment protocol (n = 8). Total metabolic tumor volume (MTV) in pretherapeutic FDG-PET was defined using a semi-automated, background-adapted threshold. Metabolic (SUVmax, SUVmean, SUVpeak, total lesion glycolysis [MTV*SUVmean]) and heterogeneity parameters (asphericity [ASP], entropy, contrast, local homogeneity, energy, and cumulative SUV-volume histograms) were derived. Early response assessment (ERA) was performed after 2 cycles of induction chemotherapy according to treatment protocol and verified by reference rating. Prediction of inadequate response (IR) in ERA was based on ROC analysis separated by stage I/II (1 and 26 patients) and stage III/IV disease (7 and 16 patients) or treatment group/level (TG/TL) 1 to 3. RESULTS: IR was seen in 28/50 patients (TG/TL 1, 6/12 patients; TG/TL 2, 10/17; TG/TL 3, 12/21). Among all PET parameters, MTV best predicted IR; ASP was the best heterogeneity parameter. AUC of MTV was 0.84 (95%-confidence interval, 0.69-0.99) in stage I/II and 0.86 (0.7-1.0) in stage III/IV. In patients of TG/TL 1, AUC of MTV was 0.92 (0.74-1.0); in TG/TL 2 0.71 (0.44-0.99), and in TG/TL 3 0.85 (0.69-1.0). Patients with high vs. low MTV had IR in 86 vs. 0% in TG/TL 1, 80 vs. 29% in TG/TL 2, and 90 vs. 27% in TG/TL 3 (cut-off, > 80 ml, > 160 ml, > 410 ml). CONCLUSIONS: In this explorative study, high total MTV best predicted inadequate response to induction therapy in pediatric HL of all pretherapeutic FDG-PET parameters - in both low and high stages as well as the 3 different TG/TL. TRIAL REGISTRATION: Ethics committee number: EA2/151/16 (retrospectively registered).

A direct comparison of contrast-enhanced ultrasound and dynamic contrast-enhanced magnetic resonance imaging for prostate cancer detection and prediction of aggressiveness.

INTRODUCTION: Dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) and contrast-enhanced ultrasound (CEUS) analyse tissue vascularization. We evaluated if CEUS can provide comparable information as DCE-MRI for the detection of prostate cancer (PCa) and prediction of its aggressiveness. MATERIAL AND METHODS: A post-hoc evaluation of 92 patients was performed. In each patient CEUS and DCE-MRI parameters of the most suspicious lesion identified on MRI were analysed. The predictive values for discrimination between benign lesions, low-/intermediate- and high-grade PCa were evaluated. Results of targeted biopsy served as reference standard (benign lesions, n=51; low- and intermediate-grade PCa [Gleason grade group 1 and 2], n=22; high-grade PCa [≥ Gleason grade group 3], n=19). RESULTS: In peripheral zone lesions of all tested CEUS parameters only time to peak (TTPCEUS) showed significant differences between benign lesions and PCa (AUC 0.65). Of all tested DCE-MRI parameters, rate constant (Kep) was the best discriminator of high-grade PCa in the whole prostate (AUC 0.83) and in peripheral zone lesions (AUC 0.89). CONCLUSION: DCE-MRI showed a superior performance for detection of PCa and prediction of its aggressiveness. CEUS and DCE-MRI performed better in peripheral zone lesions than in transition zone lesions. KEY POINTS: • DCE-MRI gathers information about vascularization and capillary permeability characteristics of tissues. • DCE-MRI can detect PCa and predict its aggressiveness. • CEUS also gathers information about vascularization of tissues. • For detection of PCa and prediction of aggressiveness DCE-MRI performed superiorly. • Both imaging techniques performed better in peripheral zone lesions.

Degenerative changes after posterior cruciate ligament reconstruction are irrespective of posterior knee stability: MRI-based long-term results.

INTRODUCTION: Posterior cruciate ligament reconstruction (PCLR) is advocated to prevent an early onset of osteoarthritis. We hypothesized that posterior instability after PCLR correlates with degenerative changes. MATERIALS AND METHODS: MRIs of 42 (12 female/30 male; 39 ± 9 years) patients were enrolled with a minimum 5-year follow-up (FFU) after PCLR. In addition, 25 contralateral and 15 follow-up MRIs (12 months after baseline) were performed. Degenerative changes were graded using WORMS. Posterior tibial translation (PTT) was measured using posterior stress radiographs. Outcome parameters included WORMS/cartilage subscore for the whole joint, patellofemoral (PFJ), medial (MFTJ), and lateral femorotibial joint (LFTJ). RESULTS: Final follow-up was 101 (range 68-168) months. WORMS reached 41.5 [18.5-56.8]. Regional WORMS for PFJ was significantly higher than MFTJ and LFTJ. Cartilage subscore yielded 7 [2.8-15]. MFTJ and PFJ were significantly higher than LFTJ. Primary outcome parameters were significantly higher than the contralateral knee (P < 0.0001) and significantly increased within 12 months (P = 0.0002). There was a significant correlation between the intraoperative degree of cartilage injury and WORMS (P < 0.0001 with r = 0.64) and between the number of previous surgery and the cartilage subscore (P = 0.03 with r = 0.32). Meniscal surgery led to a significantly higher WORMS (P = 0.035). Combined risk models revealed that women below the mean age had significantly lower WORMS (P = 0.001) and cartilage subscores (P = 0.003). CONCLUSIONS: Patients undergo degenerative changes after PCLR, which are significantly higher compared to the contralateral knee. These occur predominantly at PFJ/MFTJ and are irrespective of posterior stability. Concomitant meniscus/cartilage injuries and a high number of previous surgeries are further risk factors.

Individualized risk assessment in neuroblastoma: does the tumoral metabolic activity on 123I-MIBG SPECT predict the outcome?

PURPOSE: Risk-adapted treatment in children with neuroblastoma (NB) is based on clinical and genetic factors. This study evaluated the metabolic tumour volume (MTV) and its asphericity (ASP) in pretherapeutic 123I-MIBG SPECT for individualized image-based prediction of outcome. METHODS: This retrospective study included 23 children (11 girls, 12 boys; median age 1.8 years, range 0.3-6.8 years) with newly diagnosed NB consecutively examined with pretherapeutic 123I-MIBG SPECT. Primary tumour MTV and ASP were defined using semiautomatic thresholds. Cox regression analysis, receiver operating characteristic analysis (cut-off determination) and Kaplan-Meier analysis with the log-rank test for event-free survival (EFS) were performed for ASP, MTV, laboratory parameters (including urinary homovanillic acid-to-creatinine ratio, HVA/C), and clinical (age, stage) and genetic factors. Predictive accuracy of the optimal multifactorial model was determined in terms of Harrell's C and likelihood ratio χ 2. RESULTS: Median follow-up was 36 months (range 7-107 months; eight patients showed disease progression/relapse, four patients died). The only significant predictors of EFS in the univariate Cox regression analysis were ASP (p = 0.029; hazard ratio, HR, 1.032 for a one unit increase), MTV (p = 0.038; HR 1.012) and MYCN amplification status (p = 0.047; HR 4.67). The mean EFS in patients with high ASP (>32.0%) and low ASP were 21 and 88 months, respectively (p = 0.013), and in those with high MTV (>46.7 ml) and low MTV were 22 and 87 months, respectively (p = 0.023). A combined risk model of either high ASP and high HVA/C or high MTV and high HVA/C best predicted EFS. CONCLUSIONS: In this exploratory study, pretherapeutic image-derived and laboratory markers of tumoral metabolic activity in NB (ASP, MTV, urinary HVA/C) allowed the identification of children with a high and low risk of progression/relapse under current therapy.

Dual time point imaging for F18-FDG-PET/CT does not improve the accuracy of nodal staging in non-small cell lung cancer patients.

OBJECTIVES: To analyze the diagnostic performance of dual time point imaging (DTPI) for pre-therapeutic lymph node (LN) staging in non-small cell lung cancer (NSCLC). METHODS: This was a retrospective analysis of 47 patients with NSCLC who had undergone DTPI by PET (early + delayed) using F18-fluorodeoxyglucose (FDG). PET raw data were reconstructed iteratively (point spread function + time-of-flight). LN uptake in PET was assessed visually (four-step score) and semi-quantitatively (SUVmax, SUVmean, ratios LN/primary, LN/liver, and LN/mediastinal blood pool). DTPI analyses included retention indices (RIs), Δ-ratios and changes in visual score. Histology or cytology served as standards of reference. Accuracy was determined based on ROC analyses. RESULTS: Thirty-six of 155 LNs were malignant. DTPI accuracy was low for all measures (visual assessment, 24.5%; RI SUVmax, 68.4%; RI SUVmean, 65.8%; Δ-ratios, 63.9-76.1%) and significantly inferior to early PET. Accuracies of early (range, 86.5-92.9%) and delayed PET (range, 85.2-92.9%) were comparable. At early PET, accuracy of the visual score (92.9%) was similar or superior to semi-quantitative analyses (range, 86.5-92.3%). CONCLUSIONS: Using a modern PET/CT device and novel image reconstruction, neither additional delayed PET nor DTPI analyses improved the accuracy of PET-based LN staging. Dedicated visual assessment criteria performed very well. KEY POINTS: • DTPI did not improve accuracy of PET-based LN staging in NSCLC. • Analyzed SUV ratios were not superior to LN SUVmax or SUVmean. • A four-step visual score may allow highly accurate, standardized LN assessment.

Quantitative assessment of the asphericity of pretherapeutic FDG uptake as an independent predictor of outcome in NSCLC.

BACKGROUND: The aim of the present study was to evaluate the predictive value of a novel quantitative measure for the spatial heterogeneity of FDG uptake, the asphericity (ASP) in patients with non-small cell lung cancer (NSCLC). METHODS: FDG-PET/CT had been performed in 60 patients (15 women, 45 men; median age, 65.5 years) with newly diagnosed NSCLC prior to therapy. The FDG-PET image of the primary tumor was segmented using the ROVER 3D segmentation tool based on thresholding at the volume-reproducing intensity threshold after subtraction of local background. ASP was defined as the relative deviation of the tumor's shape from a sphere. Univariate and multivariate Cox regression as well as Kaplan-Meier (KM) analysis and log-rank test with respect to overall (OAS) and progression-free survival (PFS) were performed for clinical variables, SUVmax/mean, metabolically active tumor volume (MTV), total lesion glycolysis (TLG), ASP and "solidity", another measure of shape irregularity. RESULTS: ASP, solidity and "primary surgical treatment" were significant independent predictors of PFS in multivariate Cox regression with binarized parameters (HR, 3.66; p<0.001, HR, 2.11; p=0.05 and HR, 2.09; p=0.05), ASP and "primary surgical treatment" of OAS (HR, 3.19; p=0.02 and HR, 3.78; p=0.01, respectively). None of the other semi-quantitative PET parameters showed significant predictive value with respect to OAS or PFS. Kaplan-Meier analysis revealed a probability of 2-year PFS of 52% in patients with low ASP compared to 12% in patients with high ASP (p<0.001). Furthermore, it showed a higher OAS rate in the case of low versus high ASP (1-year-OAS, 91% vs. 67%: p=0.02). CONCLUSIONS: The novel parameter asphericity of pretherapeutic FDG uptake seems to provide better prognostic value for PFS and OAS in NCSLC compared to SUV, metabolic tumor volume, total lesion glycolysis and solidity.

Keeping Up With ChatGPT: Evaluating Its Recognition and Interpretation of Nuclear Medicine Images.

PURPOSE: The latest iteration of GPT4 (generative pretrained transformer) is a large multimodal model that can integrate both text and image input, but its performance with medical images has not been systematically evaluated. We studied whether ChatGPT with GPT-4V(ision) can recognize images from common nuclear medicine examinations and interpret them. PATIENTS AND METHODS: Fifteen representative images (scintigraphy, 11; PET, 4) were submitted to ChatGPT with GPT-4V(ision), both in its Default and "Advanced Data Analysis (beta)" version. ChatGPT was asked to name the type of examination and tracer, explain the findings and whether there are abnormalities. ChatGPT should also mark anatomical structures or pathological findings. The appropriateness of the responses was rated by 3 nuclear medicine physicians. RESULTS: The Default version identified the examination and the tracer correctly in the majority of the 15 cases (60% or 53%) and gave an "appropriate" description of the findings or abnormalities in 47% or 33% of cases, respectively. The Default version cannot manipulate images. "Advanced Data Analysis (beta)" failed in all tasks in >90% of cases. A "major" or "incompatible" inconsistency between 3 trials of the same prompt was observed in 73% (Default version) or 87% of cases ("Advanced Data Analysis (beta)" version). CONCLUSIONS: Although GPT-4V(ision) demonstrates preliminary capabilities in analyzing nuclear medicine images, it exhibits significant limitations, particularly in its reliability (ie, correctness, predictability, and consistency).

Methodological evaluation of original articles on radiomics and machine learning for outcome prediction based on positron emission tomography (PET).

AIM: Despite a vast number of articles on radiomics and machine learning in positron emission tomography (PET) imaging, clinical applicability remains limited, partly owing to poor methodological quality. We therefore systematically investigated the methodology described in publications on radiomics and machine learning for PET-based outcome prediction. METHODS: A systematic search for original articles was run on PubMed. All articles were rated according to 17 criteria proposed by the authors. Criteria with >2 rating categories were binarized into "adequate" or "inadequate". The association between the number of "adequate" criteria per article and the date of publication was examined. RESULTS: One hundred articles were identified (published between 07/2017 and 09/2023). The median proportion of articles per criterion that were rated "adequate" was 65% (range: 23-98%). Nineteen articles (19%) mentioned neither a test cohort nor cross-validation to separate training from testing. The median number of criteria with an "adequate" rating per article was 12.5 out of 17 (range, 4-17), and this did not increase with later dates of publication (Spearman's rho, 0.094; p = 0.35). In 22 articles (22%), less than half of the items were rated "adequate". Only 8% of articles published the source code, and 10% made the dataset openly available. CONCLUSION: Among the articles investigated, methodological weaknesses have been identified, and the degree of compliance with recommendations on methodological quality and reporting shows potential for improvement. Better adherence to established guidelines could increase the clinical significance of radiomics and machine learning for PET-based outcome prediction and finally lead to the widespread use in routine clinical practice.

Artificial Intelligence-powered automatic volume calculation in medical images - available tools, performance and challenges for nuclear medicine.

Volumetry is crucial in oncology and endocrinology, for diagnosis, treatment planning, and evaluating response to therapy for several diseases. The integration of Artificial Intelligence (AI) and Deep Learning (DL) has significantly accelerated the automatization of volumetric calculations, enhancing accuracy and reducing variability and labor. In this review, we show that a high correlation has been observed between Machine Learning (ML) methods and expert assessments in tumor volumetry; Yet, it is recognized as more challenging than organ volumetry. Liver volumetry has shown progression in accuracy with a decrease in error. If a relative error below 10 % is acceptable, ML-based liver volumetry can be considered reliable for standardized imaging protocols if used in patients without major anomalies. Similarly, ML-supported automatic kidney volumetry has also shown consistency and reliability in volumetric calculations. In contrast, AI-supported thyroid volumetry has not been extensively developed, despite initial works in 3D ultrasound showing promising results in terms of accuracy and reproducibility. Despite the advancements presented in the reviewed literature, the lack of standardization limits the generalizability of ML methods across diverse scenarios. The domain gap, i. e., the difference in probability distribution of training and inference data, is of paramount importance before clinical deployment of AI, to maintain accuracy and reliability in patient care. The increasing availability of improved segmentation tools is expected to further incorporate AI methods into routine workflows where volumetry will play a more prominent role in radionuclide therapy planning and quantitative follow-up of disease evolution.

Performance of Deauville Criteria in [18F]FDG-PET/CT Diagnostics of Giant Cell Arteritis.

In this retrospective study, PET/CT data from 59 patients with suspected giant cell arteritis (GCA) were reviewed using the Deauville criteria to determine an optimal cut-off between PET positivity and negativity. Seventeen standardised vascular regions were analysed per patient by three investigators blinded to clinical information. Statistical analysis included ROC curves with areas under the curve (AUC), Cohen's and Fleiss' kappa (κ) to calculate sensitivity, specificity, accuracy, and agreement. According to final clinician's diagnosis and the revised 2017 ACR criteria GCA was confirmed in 29 of 59 (49.2 %) patients. With a diagnostic cut-off ≥ 4 (highest tracer uptake of a vessel wall exceeds liver uptake) for PET positivity, all investigators achieved high accuracy (range, 89.8-93.2%) and AUC (range, 0.94-0.97). Sensitivity and specificity ranged from 89.7-96.6% and 83.3-96.7%, respectively. Agreement between the three investigators suggested 'almost perfect agreement' (Fleiss' κ = 0.84) A Deauville score of ≥4 as threshold for PET positivity yielded excellent results with high accuracy and almost perfect inter-rater agreement, suggesting a standardized, reproducible, and reliable score in diagnosing GCA. However, the small sample size and reference standard could lead to biases. Therefore, verification in a multicentre study with a larger patient cohort and prospective setting is needed.

ChatGPT: Can You Prepare My Patients for [18F]FDG PET/CT and Explain My Reports?

We evaluated whether the artificial intelligence chatbot ChatGPT can adequately answer patient questions related to [18F]FDG PET/CT in common clinical indications before and after scanning. Methods: Thirteen questions regarding [18F]FDG PET/CT were submitted to ChatGPT. ChatGPT was also asked to explain 6 PET/CT reports (lung cancer, Hodgkin lymphoma) and answer 6 follow-up questions (e.g., on tumor stage or recommended treatment). To be rated "useful" or "appropriate," a response had to be adequate by the standards of the nuclear medicine staff. Inconsistency was assessed by regenerating responses. Results: Responses were rated "appropriate" for 92% of 25 tasks and "useful" for 96%. Considerable inconsistencies were found between regenerated responses for 16% of tasks. Responses to 83% of sensitive questions (e.g., staging/treatment options) were rated "empathetic." Conclusion: ChatGPT might adequately substitute for advice given to patients by nuclear medicine staff in the investigated settings. Improving the consistency of ChatGPT would further increase reliability.

Same same but different: dopamine transporter SPECT on scanners with CZT vs. NaI detectors.

BACKGROUND: The aims of this study were to establish a normal database (NDB) for semiquantification of dopamine transporter (DAT) single-photon emission computed tomography (SPECT) with [123I]FP-CIT on a cadmium zinc telluride (CZT) camera, test the preexisting NaI-derived NDB for use in CZT scans, and compare the diagnostic findings in subjects imaged with a CZT scanner with either the preexisting NaI-based NDB or our newly defined CZT NDB. METHODS: The sample comprised 73 subjects with clinically uncertain parkinsonian syndrome (PS) who prospectively underwent [123I]FP-CIT SPECT on a CZT camera according to standard guidelines with identical acquisition and reconstruction protocols (DaTQUANT). Two experienced readers visually assessed the images and binarized the subjects into "non-neurodegenerative PS" and "neurodegenerative PS". Twenty-five subjects from the "non-neurodegenerative PS" subgroup were randomly selected to establish a CZT NDB. The remaining 48 subjects were defined as "test group". DaTQUANT was used to determine the specific binding ratio (SBR). For the test group, SBR values were transformed to z-scores for the putamen utilizing both the CZT NDB and the manufacturer-provided NaI-based NDB (GE NDB). A predefined fixed cut-off of -2 was used for dichotomization of z-scores to classify neurodegenerative and non-neurodegenerative PS. Performance of semiquantification using the two NDB to identify subjects with neurodegenerative PS was assessed in comparison with the visual rating. Furthermore, a randomized head-to-head comparison of both detector systems was performed semiquantitatively in a subset of 32 out of all 73 subjects. RESULTS: Compared to the visual rating as reference, semiquantification based on the dedicated CZT NDB led to fewer discordant ratings than the GE NDB in CZT scans (3 vs. 8 out of 48 subjects). This can be attributed to the putaminal z-scores being consistently higher with the GE NDB on a CZT camera (median absolute difference of 1.68), suggesting an optimal cut-off of -0.5 for the GE NDB instead of -2.0. Average binding ratios and z-scores were significantly lower in CZT compared to NaI data. CONCLUSIONS: Use of a dedicated, CZT-derived NDB is recommended in [123I]FP-CIT SPECT with a CZT camera since it improves agreement between semiquantification and visual assessment.

The Medical Informatics Initiative and the Network University Medicine - Perspectives for Nuclear Medicine.

Digitization in the healthcare sector and the support of clinical workflows with artificial intelligence (AI), including AI-supported image analysis, represent a great challenge and equally a promising perspective for preclinical and clinical nuclear medicine. In Germany, the Medical Informatics Initiative (MII) and the Network University Medicine (NUM) are of central importance for this transformation. This review article outlines these structures and highlights their future role in enabling privacy-preserving federated multi-center analyses with interoperable data structures harmonized between site-specific IT infrastructures. The newly founded working group "Digitization and AI" in the German Society of Nuclear Medicine (DGN) as well as the Fach- und Organspezifische Arbeitsgruppe (FOSA, specialty- and organ-specific working group) founded for the field of nuclear medicine (FOSA Nuklearmedizin) within the NUM aim to initiate and coordinate measures in the context of digital medicine and (image-)data-driven analyses for the DGN.