18F-FDG whole-body PET/CT for the evaluation of suspected native valve infective endocarditis.

18F-FDG-PET/CT is indicated in the workup of patients with suspected infective endocarditis to detect intra-cardiac and disseminated infections, as well as its source. We present the case of a 66-year-old female patient known for recurrent diabetic foot infection, with equivocal TTE results and persistent MRSA bacteremia despite medical management. PET/CT revealed evidence of left foot osteomyelitis. Whole body PET/CT diagnosed native mitral valve infective endocarditis (IE) and right lower lobe segmental pulmonary artery uptake, consistent with septic pulmonary embolism (PE).

A phase 2, single-arm trial evaluating 131 I-PSMA-1095 targeted radioligand therapy for metastatic castration-resistant prostate cancer.

BACKGROUND: Metastatic castration-resistant prostate cancer (mCRPC) remains uniformly lethal. Prostate specific membrane antigen (PSMA) is a transmembrane glycoprotein overexpressed in prostate cancer. 131 I-PSMA-1095 (also known as 131 I-MIP-1095) is a PSMA-targeted radioligand which selectively delivers therapeutic radiation to cancer cells and the tumor microenvironment. METHODS: We conducted a single-arm, phase 2 trial to assess efficacy and tolerability of 131 I-PSMA-1095 in mCRPC patients who had exhausted all lines of approved therapy. All patients underwent 18 F-DCFPyL PET and 18 F-FDG PET to determine PSMA-positive tumor volume, and patients with >50% PSMA-positive tumor volume were treated with up to four doses of 131 I-PSMA-1095. The primary endpoint was the response rate of prostate specific antigen (PSA). Secondary endpoints included rates of radiographic response and adverse events. Overall and radiographic progression-free survival were also analyzed. RESULTS: Eleven patients were screened for inclusion and nine patients received 131 I-PSMA-1095. The median baseline PSA was 162 µg/l, and six patients demonstrated a >50% PSA decrease. One patient demonstrated a confirmed radiographic response. Median overall survival was 10.3 months, and median progression-free survival was 5.4 months. Four patients experienced adverse events of grade 3 or higher, the most frequent being thrombocytopenia and anemia. CONCLUSION: 131 I-PSMA-1095 is highly active against heavily-pretreated PSMA-positive mCRPC, significantly decreasing tumor burden as measured by PSA. Adverse events, mainly hematologic toxicity, were not infrequent, likely related to off-target irradiation. This hematologic toxicity, as well as a higher logistical burden associated with use, could represent relative disadvantages of 131 I-PSMA-1095 compared to 177 Lu-PSMA-617.

177Lu-PSMA-617 in Metastatic Castration-Resistant Prostate Cancer: A Review of the Evidence and Implications for Canadian Clinical Practice.

Prostate-specific membrane antigen (PSMA) is highly expressed in prostate cancer and a therapeutic target. Lutetium-177 (177Lu)-PSMA-617 is the first radioligand therapy to be approved in Canada for use in patients with metastatic castration-resistant prostate cancer (mCRPC). As this treatment represents a new therapeutic class, guidance regarding how to integrate it into clinical practice is needed. This article aims to review the evidence from prospective phase 2 and 3 clinical trials and meta-analyses of observational studies on the use of 177Lu-PSMA-617 in prostate cancer and discuss how Canadian clinicians might best apply these data in practice. The selection of appropriate patients, the practicalities of treatment administration, including necessary facilities for treatment procedures, the assessment of treatment response, and the management of adverse events are considered. Survival benefits were observed in clinical trials of 177Lu-PSMA-617 in patients with progressive, PSMA-positive mCRPC who were pretreated with androgen receptor pathway inhibitors and taxanes, as well as in taxane-naïve patients. However, the results of ongoing trials are awaited to clarify questions regarding the optimal sequencing of 177Lu-PSMA-617 with other therapies, as well as the implications of predictive biomarkers, personalized dosimetry, and combinations with other therapies.

Neoadjuvant versus Concurrent Androgen Deprivation Therapy in Localized Prostate Cancer Treated with Radiotherapy: A Systematic Review of the Literature.

BACKGROUND: There is an ongoing debate on the optimal sequencing of androgen deprivation therapy (ADT) and radiotherapy (RT) in patients with localized prostate cancer (PCa). Recent data favors concurrent ADT and RT over the neoadjuvant approach. METHODS: We conducted a systematic review in PubMed, EMBASE, and Cochrane Databases assessing the combination and optimal sequencing of ADT and RT for Intermediate-Risk (IR) and High-Risk (HR) PCa. FINDINGS: Twenty randomized control trials, one abstract, one individual patient data meta-analysis, and two retrospective studies were selected. HR PCa patients had improved survival outcomes with RT and ADT, particularly when a long-course Neoadjuvant-Concurrent-Adjuvant ADT was used. This benefit was seen in IR PCa when adding short-course ADT, although less consistently. The best available evidence indicates that concurrent over neoadjuvant sequencing is associated with better metastases-free survival at 15 years. Although most patients had IR PCa, HR participants may have been undertreated with short-course ADT and the absence of pelvic RT. Conversely, retrospective data suggests a survival benefit when using the neoadjuvant approach in HR PCa patients. INTERPRETATION: The available literature supports concurrent ADT and RT initiation for IR PCa. Neoadjuvant-concurrent-adjuvant sequencing should remain the standard approach for HR PCa and is an option for IR PCa.

Advances in PARP Inhibitors for Prostate Cancer.

Poly-adenosine diphosphate-ribose polymerase plays an essential role in cell function by regulating apoptosis, genomic stability and DNA repair. PARPi is a promising drug class that has gained significant traction in the last decade with good outcomes in different cancers. Several trials have sought to test its effectiveness in metastatic castration resistant prostate cancer (mCRPC). We conducted a comprehensive literature review to evaluate the current role of PARPi in this setting. To this effect, we conducted queries in the PubMed, Embase and Cochrane databases. We reviewed and compared all major contemporary publications on the topic. In particular, recent phase II and III studies have also demonstrated the benefits of olaparib, rucaparib, niraparib, talazoparib in CRPC. Drug effectiveness has been assessed through radiological progression or overall response. Given the notion of synthetic lethality and potential synergy with other oncological therapies, several trials are looking to integrate PARPi in combined therapies. There remains ongoing controversy on the need for genetic screening prior to treatment initiation as well as the optimal patient population, which would benefit most from PARPi. PARPi is an important asset in the oncological arsenal for mCRPC. New combinations with PARPi may improve outcomes in earlier phases of prostate cancer.

Port-Site Metastasis Identified on Prostate-Specific Membrane Antigen-Targeted 18 F-DCFPyL PET/CT After Robot-Assisted Laparoscopic Radical Prostatectomy.

ABSTRACT: Port-site metastasis is an extremely rare complication following minimally invasive oncologic surgery for prostate cancer. We present the case of a 74-year-old man who underwent robot-assisted laparoscopic radical prostatectomy followed by salvage radiotherapy. Despite treatment, he developed biochemical recurrence. However, there was no evidence of disease on CT and bone scan at a prostate-specific antigen of 4.6 ng/mL. Subsequently, 18 F-DCFPyL PET/CT revealed a solitary focus of intense uptake in the right rectus abdominis muscle that was felt to represent a port-site metastasis. Histopathologic evaluation with immunostaining following ultrasound-guided needle biopsy confirmed the presence of metastatic adenocarcinoma of the prostate.

Differential Detection of Hepatic Metastases on 68 Ga-DOTATATE PET/CT and 177 Lu-DOTATATE SPECT/CT.

ABSTRACT: 68 Ga-DOTATATE PET/CT is indicated for selecting patients for peptide receptor radionuclide therapy (PRRT). Although highly sensitive, the detectability of smaller lesions, particularly in the liver, is lower. We present the case of a 58-year-old man with metastatic well-differentiated pancreatic neuroendocrine tumor whose MRI revealed progression of hepatic metastases. 68 Ga-DOTATATE PET/CT performed to determine eligibility for PRRT did not demonstrate DOTATATE-avid disease within the liver. 18 F-FDG PET/CT was also negative at the liver and the patient proceeded to 177 Lu-DOTATATE PRRT, where multi-time point posttherapy planar imaging and SPECT/CT showed intense uptake in the known liver metastases.

Comparing digital to analog prostate-specific membrane antigen-targeted piflufolastat 18 F PET/CT in prostate cancer patients in early biochemical failure.

PURPOSE: Prostate-specific membrane antigen (PSMA) positron emission tomography/computer tomography (PET/CT) in prostate cancer patients with biochemical failure(BCF) showslimited sensitivity when the prostate-specific antigen(PSA) <0.5 ng/mL. The development of digital PET/CT has greatly improved smaller lesion detection. This study's goal was to compare the performance and clinical value of PSMA-targeted piflufolastat PET/CT for prostate cancer BCF with digital versus analog PET/CT. METHODS: In this retrospective study, all piflufolastat PET/CT scans in subjects with PSA ≤ 3.0 ng/mL who were referred for prostate cancer BCF were included. The performance characteristics of 171 analog PET/CT studies in 155 subjects from May 2017 to January 2020 and 106 digital PET/CT studies in 103 subjects from February 2020 to December 2020 were compared. Lesions were considered malignant if they did not match the known physiological distribution of piflufolastat and did not represent uptake in benign lesions. PSMA PET/CT studies were considered positive if at least one malignant lesion was detected and negative if none were detected. RESULTS: Digital piflufolastat PET/CT outperformed analog piflufolastat PET/CT in subjects with PSA < 0.5 ng/mL with a positivity rate of 69% versus 37%, respectively. In patients with PSA ≥ 0.5 ng/mL, both technologies performed similarly. There was no statistically significant difference between the number or size of piflufolastat-avid lesions detected per PET/CT study. CONCLUSION: In prostate cancer patients with BCF and PSA < 0.5 ng/mL, digital piflufolastat PET/CT has a higher detection rate of malignant lesions than analog piflufolastat PET/CT.

Salvage external beam radiotherapy after HIFU failure in localized prostate cancer: A single institution experience.

Purpose/objectives: High-intensity focused ultrasound (HIFU) remains investigational as primary treatment for localized prostate cancer but is sometimes offered to select patients. At HIFU failure, data guiding salvage treatment is limited to small retrospective series with short follow-up. We evaluated our institutional experience using salvage radiation therapy (SRT) after HIFU failure. Materials/methods: We conducted a retrospective analysis of patients with local failure post-HIFU who received salvage image-guided external beam radiation therapy (EBRT) delivered via intensity-modulated radiotherapy (IMRT). Our primary endpoint was biochemical failure-free survival (bFFS) defined as prostate-specific antigen (PSA) nadir + 2 ng/mL. Secondary endpoints included metastasis-free survival (MFS) and overall survival (OS). Endpoints were evaluated using Kaplan-Meier analysis. Results: From 2013 to 2018, 12 out of 96 patients treated with primary HIFU received SRT via conventional or moderate hypofractionation. Median time from HIFU to SRT was 13.5 months. Seven patients had stage migration to high-risk disease at the time of SRT. Mean PSA prior to SRT was 8.2ug/L and mean nadir post-SRT was 1.2ug/L. Acute International Prostate Symptom Score (IPSS) as well as International Index of Erectile Dysfunction (IIEF) scores were similar to baseline (p = 0.5 and 0.1, respectively). Late toxicities were comparable to those reported after primary EBRT for localized prostate cancer. At a median follow-up of 46 months, the OS was 100%. The 5-year bFFS and MFS were both 83.3%. Conclusions: To our knowledge, we report one of the largest series on contemporary SRT post HIFU failure. We show that SRT is feasible, effective and carries no additional acute or delayed toxicity.

Outcomes and Factors Associated with Completion of Radium-223 Therapy.

Purpose: Radium-223 has been demonstrated in clinical trials to improve survival in castration-resistant prostate cancer (CRPC) patients with bone metastases. However, its performance in routine use remains to be fully characterized. This study aims to describe patient outcomes in the real world as well as identify factors associated with completion of the 6-dose regimen and alkaline phosphatase (ALP) response. Methods: Thirty-six patients who received at least one dose of radium-223 at the Jewish General Hospital in Montréal, Canada, were analysed in a retrospective manner. Using logistic regression, the primary analysis aimed to identify factors associated with treatment completion, and the secondary analysis aimed to identify factors associated with ALP response. Results: Twenty-one out of 36 patients received all 6 doses of radium-223. Fifteen patients had an ALP response, defined as a 30% decrease in ALP from baseline values. On primary analysis, baseline ALP > 120 U/L and prostate-specific antigen (PSA) > 50 µg/L were significantly associated with lower therapy completion rates (OR = 0.10, p = 0.004; OR = 0.18, p = 0.022 respectively). On adjustment for confounders, only ALP remained significant (OR = 0.14, p = 0.021). Clinical disease progression was the most common reason for treatment non-completion, and it was also associated with elevated baseline ALP (OR = 6.00, p = 0.044). On secondary analysis, previous chemotherapy for CRPC was a negative predictor of ALP response (OR = 0.15, p = 0.034). Conclusion: Elevated baseline ALP and PSA were associated with a lower rate of radium-223 regimen completion; receiving chemotherapy for CRPC prior to radium-223 was associated with a lower rate of ALP response. Supplementary Information: The online version contains supplementary material available at 10.1007/s13139-022-00760-8.

Interval Changes in PSMA PET/CT During Radium-223 Therapy for Metastatic Bone Disease from Castration-Resistant Prostate Cancer.

Background: Radium-223, an alpha-emitting therapeutic radiopharmaceutical, prolongs overall survival (OS) in patients with symptomatic bone-predominant metastatic castration-resistant prostate cancer (mCRPC). PSMA PET/CT is a molecular imaging tool for whole-body imaging of prostate cancer and may inform on the mechanisms of radium-223 activity and treatment resistance in mCRPC patients. Methods: In an open-label, single-arm, prospective trial, we enrolled patients with bone-predominant mCRPC to undergo baseline PSMA PET/CT, 6 cycles of radium-223, and post-therapy PSMA PET/CT. We assessed the relationship between multiple parameters of interval change on PSMA PET/CT on aPROMISE PSMA automated analysis and a human reader, and laboratory measurements. Results: Fourteen patients were enrolled and 9 patients completed both protocol-defined PSMA PET/CT. Of the 9 evaluable patients, 1 (11%) had a complete response and 8 (89%) had PSMA PET progressive disease. All patients showed decreases in PSMA uptake in some disease sites evident on the baseline scan. The change in overall burden of disease on PSMA PET was more strongly correlated with changes in PSA (ρ = 0.95) than ALP (ρ = 0.62). Progression in bone was a common finding on post-treatment PSMA PET/CT. Conclusion: PSMA PET was able to assess response in individual lesions during radium-223 therapy in mCRPC patients. PSMA PET responses in previously established disease sites were universal, but most patients also showed overall PSMA PET progression during 6 cycles of radium-223. Given high correlation with changes in PSA, PSMA PET may be of limited value in follow-up during or after radium-223 in bone-predominant mCRPC.

Quality and Quantity: Evaluating Tumor Biology Alongside Novel Imaging on Diagnosis of Metastatic Hormone-sensitive Prostate Cancer.

Tumor biology may play an important role as an effective predictive biomarker that is complementary to functional imaging for metastatic hormone-sensitive prostate cancer.

Analytical performance of aPROMISE: automated anatomic contextualization, detection, and quantification of [18F]DCFPyL (PSMA) imaging for standardized reporting.

PURPOSE: The application of automated image analyses could improve and facilitate standardization and consistency of quantification in [18F]DCFPyL (PSMA) PET/CT scans. In the current study, we analytically validated aPROMISE, a software as a medical device that segments organs in low-dose CT images with deep learning, and subsequently detects and quantifies potential pathological lesions in PSMA PET/CT. METHODS: To evaluate the deep learning algorithm, the automated segmentations of the low-dose CT component of PSMA PET/CT scans from 20 patients were compared to manual segmentations. Dice scores were used to quantify the similarities between the automated and manual segmentations. Next, the automated quantification of tracer uptake in the reference organs and detection and pre-segmentation of potential lesions were evaluated in 339 patients with prostate cancer, who were all enrolled in the phase II/III OSPREY study. Three nuclear medicine physicians performed the retrospective independent reads of OSPREY images with aPROMISE. Quantitative consistency was assessed by the pairwise Pearson correlations and standard deviation between the readers and aPROMISE. The sensitivity of detection and pre-segmentation of potential lesions was evaluated by determining the percent of manually selected abnormal lesions that were automatically detected by aPROMISE. RESULTS: The Dice scores for bone segmentations ranged from 0.88 to 0.95. The Dice scores of the PSMA PET/CT reference organs, thoracic aorta and liver, were 0.89 and 0.97, respectively. Dice scores of other visceral organs, including prostate, were observed to be above 0.79. The Pearson correlation for blood pool reference was higher between any manual reader and aPROMISE, than between any pair of manual readers. The standard deviations of reference organ uptake across all patients as determined by aPROMISE (SD = 0.21 blood pool and SD = 1.16 liver) were lower compared to those of the manual readers. Finally, the sensitivity of aPROMISE detection and pre-segmentation was 91.5% for regional lymph nodes, 90.6% for all lymph nodes, and 86.7% for bone in metastatic patients. CONCLUSION: In this analytical study, we demonstrated the segmentation accuracy of the deep learning algorithm, the consistency in quantitative assessment across multiple readers, and the high sensitivity in detecting potential lesions. The study provides a foundational framework for clinical evaluation of aPROMISE in standardized reporting of PSMA PET/CT.

The Triple-Tracer strategy against Metastatic PrOstate cancer (3TMPO) study protocol.

OBJECTIVE: To determine the prevalence of intra-patient inter-metastatic heterogeneity based on positron emission tomography (PET)/computed tomography (CT) in patients with metastatic castration-resistant prostate cancer (mCRPC) and to determine the prevalence of neuroendocrine disease in these patients and their eligibility for radioligand therapies (RLTs). PATIENTS AND METHODS: This multicentre observational prospective clinical study will include 100 patients with mCRPC from five Canadian academic centres. Patients with radiological or biochemical progression and harbouring at least three metastases by conventional imaging will be accrued. Intra-patient inter-metastatic heterogeneity will be determined with triple-tracer imaging using fluorine-18 fluorodeoxyglucose (18 F-FDG), gallium-68-(68 Ga)-prostate-specific membrane antigen (PSMA)-617 and 68 Ga-DOTATATE, which are a glucose analogue, a PSMA receptor ligand and a somatostatin receptor ligand, respectively. The 68 Ga-PSMA-617 and 18 F-FDG PET/CT scans will be performed first. If at least one PSMA-negative/FDG-positive lesion is observed, an additional PET/CT scan with 68 Ga-DOTATATE will be performed. The tracer uptake of individual lesions will be assessed for each PET tracer and patients with lesions presenting discordant uptake profiles will be considered as having inter-metastatic heterogeneous disease and may be offered a biopsy. EXPECTED RESULTS: The proposed triple-tracer approach will allow whole-body mCRPC characterisation, investigating the inter-metastatic heterogeneity in order to better understand the phenotypic plasticity of prostate cancer, including the neuroendocrine transdifferentiation that occurs during mCRPC progression. Based on 68 Ga-PSMA-617 or 68 Ga-DOTATATE PET positivity, the potential eligibility of patients for PSMA and DOTATATE-based RLT will be assessed. Non-invasive whole-body determination of mCRPC heterogeneity and transdifferentiation is highly innovative and might establish the basis for new therapeutic strategies. Comparison of molecular imaging findings with biopsies will also link metastasis biology to radiomic features. CONCLUSION: This study will add novel, biologically relevant dimensions to molecular imaging: the non-invasive detection of inter-metastatic heterogeneity and transdifferentiation to neuroendocrine prostate cancer by using a multi-tracer PET/CT strategy to further personalise the care of patients with mCRPC.

Sarcopenia in cardiac surgery: Dual X-ray absorptiometry study from the McGill frailty registry.

BACKGROUND: To determine the prevalence and prognostic value of sarcopenia measured by dual x-ray absorptiometry (DXA) and physical performance tests in patients undergoing coronary artery bypass surgery or heart valve procedures. METHODS: Adults undergoing cardiac surgery were prospectively enrolled and completed a questionnaire, physical performance battery, and a DXA scan (GE Lunar) to measure appendicular muscle mass indexed to height2 (AMMI). Patients were categorized as sarcopenic based on European Working Group 2 guidelines if they had low AMMI defined as <7 kg/m2 for men or <5.5 kg/m2 for women, and low muscle strength defined as 5 chair rise time ≥15 seconds. Cox proportional hazards regression was used to test the association between sarcopenia and all-cause mortality over a median follow-up of 4.3 years. RESULTS: The cohort consisted of 141 patients with a mean age of 69.7 ± 10.0 years and 21% females. The prevalence rates of low AMMI, slow chair rise time, and sarcopenia (low AMMI and slow chair rise time) were 24%, 57%, 13%, respectively. The 4-year survival rate was 79% in the non-sarcopenic group as compared to 56% in the sarcopenic group (Log-rank P = 0.01). In the multivariable model, each standard deviation of decreasing AMMI and increasing chair rise time was associated with a hazard ratio for all-cause mortality of 1.84 (95% CI 1.18, 2.86) and 1.79 (95% CI 1.26, 2.54), respectively. CONCLUSION: Lower-extremity muscle strength and DXA-based muscle mass are objective indicators of sarcopenia that are independently predictive of all-cause mortality in older cardiac surgery patients.

Magnetic Resonance Imaging Based Radiomic Models of Prostate Cancer: A Narrative Review.

The management of prostate cancer (PCa) is dependent on biomarkers of biological aggression. This includes an invasive biopsy to facilitate a histopathological assessment of the tumor's grade. This review explores the technical processes of applying magnetic resonance imaging based radiomic models to the evaluation of PCa. By exploring how a deep radiomics approach further optimizes the prediction of a PCa's grade group, it will be clear how this integration of artificial intelligence mitigates existing major technological challenges faced by a traditional radiomic model: image acquisition, small data sets, image processing, labeling/segmentation, informative features, predicting molecular features and incorporating predictive models. Other potential impacts of artificial intelligence on the personalized treatment of PCa will also be discussed. The role of deep radiomics analysis-a deep texture analysis, which extracts features from convolutional neural networks layers, will be highlighted. Existing clinical work and upcoming clinical trials will be reviewed, directing investigators to pertinent future directions in the field. For future progress to result in clinical translation, the field will likely require multi-institutional collaboration in producing prospectively populated and expertly labeled imaging libraries.

A Phase 2/3 Prospective Multicenter Study of the Diagnostic Accuracy of Prostate Specific Membrane Antigen PET/CT with 18F-DCFPyL in Prostate Cancer Patients (OSPREY).

PURPOSE: Prostate specific membrane antigen-targeted positron emission tomography/computerized tomography has the potential to improve the detection and localization of prostate cancer. OSPREY was a prospective trial designed to determine the diagnostic performance of 18F-DCFPyL-positron emission tomography/computerized tomography for detecting sites of metastatic prostate cancer. MATERIALS AND METHODS: Two patient populations underwent 18F-DCFPyL-positron emission tomography/computerized tomography. Cohort A enrolled men with high-risk prostate cancer undergoing radical prostatectomy with pelvic lymphadenectomy. Cohort B enrolled patients with suspected recurrent/metastatic prostate cancer on conventional imaging. Three blinded central readers evaluated the 18F-DCFPyL-positron emission tomography/computerized tomography. Diagnostic performance of 18F-DCFPyL-positron emission tomography/computerized tomography was based on imaging results compared to histopathology. In cohort A, detection of pelvic nodal disease (with specificity and sensitivity as co-primary end points) and of extrapelvic metastases were evaluated. In cohort B, sensitivity and positive predictive value for prostate cancer within biopsied lesions were evaluated. RESULTS: A total of 385 patients were enrolled. In cohort A (252 evaluable patients), 18F-DCFPyL-positron emission tomography/computerized tomography had median specificity of 97.9% (95% CI: 94.5%-99.4%) and median sensitivity of 40.3% (28.1%-52.5%, not meeting prespecified end point) among 3 readers for pelvic nodal involvement; median positive predictive value and negative predictive value were 86.7% (69.7%-95.3%) and 83.2% (78.2%-88.1%), respectively. In cohort B (93 evaluable patients, median prostate specific antigen 11.3 ng/ml), median sensitivity and positive predictive value for extraprostatic lesions were 95.8% (87.8%-99.0%) and 81.9% (73.7%-90.2%), respectively. CONCLUSIONS: The primary end point for specificity was met while the primary end point for sensitivity was not. The high positive predictive value observed in both cohorts indicates that 18F-DCFPyL-positive lesions are likely to represent disease, supporting the potential utility of 18F-DCFPyL-positron emission tomography/computerized tomography to stage men with high-risk prostate cancer for nodal or distant metastases, and reliably detect sites of disease in men with suspected metastatic prostate cancer.

Blinded Clinical Evaluation for Dementia of Alzheimer's Type Classification Using FDG-PET: A Comparison Between Feature-Engineered and Non-Feature-Engineered Machine Learning Methods.

BACKGROUND: Advanced machine learning methods can aid in the identification of dementia risk using neuroimaging-derived features including FDG-PET. However, to enable the translation of these methods and test their usefulness in clinical practice, it is crucial to conduct independent validation on real clinical samples, which has yet to be properly delineated in the current literature. OBJECTIVE: In this paper, we present our efforts to enable such clinical translational through the evaluation and comparison of two machine-learning methods for discrimination between dementia of Alzheimer's type (DAT) and Non-DAT controls. METHODS: FDG-PET-based dementia scores were generated on an independent clinical sample whose clinical diagnosis was blinded to the algorithm designers. A feature-engineered approach (multi-kernel probability classifier) and a non-feature-engineered approach (3D convolutional neural network) were analyzed. Both classifiers were pre-trained on cognitively normal subjects as well as subjects with DAT. These two methods provided a probabilistic dementia score for this previously unseen clinical data. Performance of the algorithms were compared against ground-truth dementia rating assessed by experienced nuclear physicians. RESULTS: Blinded clinical evaluation on both classifiers showed good separation between the cognitively normal subjects and the patients diagnosed with DAT. The non-feature-engineered dementia score showed higher sensitivity among subjects whose diagnosis was in agreement between the machine-learning models, while the feature-engineered approach showed higher specificity in non-consensus cases. CONCLUSION: In this study, we demonstrated blinded evaluation using data from an independent clinical sample for assessing the performance in DAT classification models in a clinical setting. Our results showed good generalizability for two machine-learning approaches, marking an important step for the translation of pre-trained machine-learning models into clinical practice.