A Phase I Trial of Image-Guided, Risk-Volume-Adapted Post-Prostatectomy Radiotherapy.

PURPOSE: This is a phase I trial with the primary objective of identifying the most compressed dose schedule (DS) tolerable using risk-volume-adapted, hypofractionated, post-operative radiotherapy (PORT) for biochemically recurrent prostate cancer. Secondary endpoints included biochemical progression free survival (bPFS) and quality of life (QOL). METHODS: Patients were treated with one of 3 isoeffective dose schedules (DS1: 20 fractions, DS2: 15 fractions, DS3: 10 fractions) that escalated dose to the imaging-defined local recurrence (73Gy3 EQD2) and de-escalated dose to the remainder of the prostate bed (48Gy3 EQD2). Escalation followed a standard 3+3 design with a 6-patient expansion at the maximally tolerated hypofractionated dose schedule (MTHDS). Dose limiting toxicity (DLT) was defined as CTCAE v.4.0 grade (G) 3 toxicity lasting >4 days within 21 days of PORT completion or grade 4 gastrointestinal (GI) or genitourinary (GU) toxicities thereafter. QOL was assessed longitudinally through 24 months with the EPIC-26. RESULTS: Between 01/2018 and 12/2023, 15 patients were treated (3 with DS1, 3 with DS2, and 9 with DS3). The median follow-up was 48 months. No DLTs were observed on any DS, and, thus, expansion occurred at DS3. The cumulative incidence of G3 GI and GU toxicity was 7% and 9% at 24 months, respectively, with no G4 events observed. Transient, acute G2+ GI toxicity was most common. QOL worsened transiently during study follow-up in urinary incontinence, GI, and sexual subdomains but was similar to baseline by 24 months. The bPFS was 91% at both 24- and 60-months. CONCLUSIONS: The maximally tolerated hypofractionated dose schedule for hypofractionated, risk-volume-adapted PORT was determined to be DS3 (36.4Gy to the prostate bed and 47.1Gy to the imaging-defined recurrence in 10 daily fractions). No >G3 events were observed. Transient declines in QOL did not persist through 24 months.

Location-based Radiology Report-Guided Semi-supervised Learning for Prostate Cancer Detection.

Prostate cancer is one of the most prevalent malignancies in the world. While deep learning has potential to further improve computer-aided prostate cancer detection on MRI, its efficacy hinges on the exhaustive curation of manually annotated images. We propose a novel methodology of semisupervised learning (SSL) guided by automatically extracted clinical information, specifically the lesion locations in radiology reports, allowing for use of unannotated images to reduce the annotation burden. By leveraging lesion locations, we refined pseudo labels, which were then used to train our location-based SSL model. We show that our SSL method can improve prostate lesion detection by utilizing unannotated images, with more substantial impacts being observed when larger proportions of unannotated images are used.

Editorial for Special Topics: Imaging-Based Diagnosis for Prostate Cancer-State of the Art.

This Special Topics Issue, "Imaging-based Diagnosis of Prostate Cancer-State of the Art", of Diagnostics compiles 10 select articles [...].

Glypican-3 deficiency in liver cancer upregulates MAPK/ERK pathway but decreases cell proliferation.

Glypican-3 (GPC3) is overexpressed in hepatocellular carcinomas and hepatoblastomas and represents an important therapeutic target but the biologic importance of GPC3 in liver cancer is unclear. To date, there are limited data characterizing the biological implications of GPC3 knockout (KO) in liver cancers that intrinsically express this target. Here, we report on the development and characterization of GPC3-KO liver cancer cell lines and compare to them to parental lines. GPC3-KO variants were established in HepG2 and Hep3B liver cancer cell lines using a lentivirus-mediated CRISPR/Cas9 system. We assessed the effects of GPC3 deficiency on oncogenic properties in vitro and in murine xenograft models. Downstream cellular signaling pathway changes induced by GPC3 deficiency were examined by RNAseq and western blot. To confirm the usefulness of the models for GPC3-targeted drug development, we evaluated the target engagement of a GPC3-selective antibody, GC33, conjugated to the positron-emitting zirconium-89 (89Zr) in subcutaneous murine xenografts of wild type (WT) and KO liver cancer cell lines. Deletion of GPC3 significantly reduced liver cancer cell proliferation, migration, and invasion compared to the parental cell lines. Additionally, the tumor growth of GPC3-KO liver cancer xenografts was significantly slower compared with control xenografts. RNA sequencing analysis also showed GPC3-KO resulted in a reduction in the expression of genes associated with cell cycle regulation, invasion, and migration. Specifically, we observed the downregulation of components in the AKT/NFκB/WNT signaling pathways and of molecules related to cell cycle regulation with GPC3-KO. In contrast, pMAPK/ERK1/2 was upregulated, suggesting an adaptive compensatory response. KO lines demonstrated increased sensitivity to ERK (GDC09994), while AKT (MK2206) inhibition was more effective in WT lines. Using antibody-based positron emission tomography (immunoPET) imaging, we confirmed that 89Zr-GC33 accumulated exclusively in GPC3-expression xenografts but not in GPC3-KO xenografts with high tumor uptake and tumor-to-liver signal ratio. We show that GPC3-KO liver cancer cell lines exhibit decreased tumorigenicity and altered signaling pathways, including upregulated pMAPK/ERK1/2, compared to parental lines. Furthermore, we successfully distinguished between GPC3+ and GPC3- tumors using the GPC3-targeted immunoPET imaging agent, demonstrating the potential utility of these cell lines in facilitating GPC3-selective drug development.

A selective CutMix approach improves generalizability of deep learning-based grading and risk assessment of prostate cancer.

The Gleason score is an important predictor of prognosis in prostate cancer. However, its subjective nature can result in over- or under-grading. Our objective was to train an artificial intelligence (AI)-based algorithm to grade prostate cancer in specimens from patients who underwent radical prostatectomy (RP) and to assess the correlation of AI-estimated proportions of different Gleason patterns with biochemical recurrence-free survival (RFS), metastasis-free survival (MFS), and overall survival (OS). Training and validation of algorithms for cancer detection and grading were completed with three large datasets containing a total of 580 whole-mount prostate slides from 191 RP patients at two centers and 6218 annotated needle biopsy slides from the publicly available Prostate Cancer Grading Assessment dataset. A cancer detection model was trained using MobileNetV3 on 0.5 mm × 0.5 mm cancer areas (tiles) captured at 10× magnification. For cancer grading, a Gleason pattern detector was trained on tiles using a ResNet50 convolutional neural network and a selective CutMix training strategy involving a mixture of real and artificial examples. This strategy resulted in improved model generalizability in the test set compared with three different control experiments when evaluated on both needle biopsy slides and whole-mount prostate slides from different centers. In an additional test cohort of RP patients who were clinically followed over 30 years, quantitative Gleason pattern AI estimates achieved concordance indexes of 0.69, 0.72, and 0.64 for predicting RFS, MFS, and OS times, outperforming the control experiments and International Society of Urological Pathology system (ISUP) grading by pathologists. Finally, unsupervised clustering of test RP patient specimens into low-, medium-, and high-risk groups based on AI-estimated proportions of each Gleason pattern resulted in significantly improved RFS and MFS stratification compared with ISUP grading. In summary, deep learning-based quantitative Gleason scoring using a selective CutMix training strategy may improve prognostication after prostate cancer surgery.

Deep learning-based image quality assessment: impact on detection accuracy of prostate cancer extraprostatic extension on MRI.

OBJECTIVE: To assess impact of image quality on prostate cancer extraprostatic extension (EPE) detection on MRI using a deep learning-based AI algorithm. MATERIALS AND METHODS: This retrospective, single institution study included patients who were imaged with mpMRI and subsequently underwent radical prostatectomy from June 2007 to August 2022. One genitourinary radiologist prospectively evaluated each patient using the NCI EPE grading system. Each T2WI was classified as low- or high-quality by a previously developed AI algorithm. Fisher's exact tests were performed to compare EPE detection metrics between low- and high-quality images. Univariable and multivariable analyses were conducted to assess the predictive value of image quality for pathological EPE. RESULTS: A total of 773 consecutive patients (median age 61 [IQR 56-67] years) were evaluated. At radical prostatectomy, 23% (180/773) of patients had EPE at pathology, and 41% (131/318) of positive EPE calls on mpMRI were confirmed to have EPE. The AI algorithm classified 36% (280/773) of T2WIs as low-quality and 64% (493/773) as high-quality. For EPE grade ≥ 1, high-quality T2WI significantly improved specificity for EPE detection (72% [95% CI 67-76%] vs. 63% [95% CI 56-69%], P = 0.03), but did not significantly affect sensitivity (72% [95% CI 62-80%] vs. 75% [95% CI 63-85%]), positive predictive value (44% [95% CI 39-49%] vs. 38% [95% CI 32-43%]), or negative predictive value (89% [95% CI 86-92%] vs. 89% [95% CI 85-93%]). Sensitivity, specificity, PPV, and NPV for EPE grades ≥ 2 and ≥ 3 did not show significant differences attributable to imaging quality. For NCI EPE grade 1, high-quality images (OR 3.05, 95% CI 1.54-5.86; P < 0.001) demonstrated a stronger association with pathologic EPE than low-quality images (OR 1.76, 95% CI 0.63-4.24; P = 0.24). CONCLUSION: Our study successfully employed a deep learning-based AI algorithm to classify image quality of prostate MRI and demonstrated that better quality T2WI was associated with more accurate prediction of EPE at final pathology.

In vivo tracking of ex vivo generated 89 Zr-oxine labeled plasma cells by PET in a non-human primate model.

B cells are an attractive platform for engineering to produce protein-based biologics absent in genetic disorders, and potentially for the treatment of metabolic diseases and cancer. As part of pre-clinical development of B cell medicines, we demonstrate a method to collect, ex vivo expand, differentiate, radioactively label, and track adoptively transferred non-human primate (NHP) B cells. These cells underwent 10- to 15-fold expansion, initiated IgG class switching, and differentiated into antibody secreting cells. Zirconium-89-oxine labeled cells were infused into autologous donors without any preconditioning and tracked by PET/CT imaging. Within 24 hours of infusion, 20% of the initial dose homed to the bone marrow and spleen and distributed stably and equally between the two. Interestingly, approximately half of the dose homed to the liver. Image analysis of the bone marrow demonstrated inhomogeneous distribution of the cells. The subjects experienced no clinically significant side effects or laboratory abnormalities. A second infusion of B cells into one of the subjects resulted in an almost identical distribution of cells, suggesting a non-limiting engraftment niche and feasibility of repeated infusions. This work supports the NHP as a valuable model to assess the potential of B cell medicines as potential treatment for human diseases.

Near-infrared duocarmycin photorelease from a Treg-targeted antibody-drug conjugate improves efficacy of PD-1 blockade in syngeneic murine tumor models.

Regulatory T cells (Tregs) play a crucial role in mediating immunosuppression in the tumor microenvironment. Furthermore, Tregs contribute to the lack of efficacy and hyperprogressive disease upon Programmed cell death protein 1 (PD-1) blockade immunotherapy. Thus, Tregs are considered a promising therapeutic target, especially when combined with PD-1 blockade. However, systemic depletion of Tregs causes severe autoimmune adverse events, which poses a serious challenge to Treg-directed therapy. Here, we developed a novel treatment to locally and predominantly damage Tregs by near-infrared duocarmycin photorelease (NIR-DPR). In this technology, we prepared anti-CD25 F(ab')2 conjugates, which site-specifically uncage duocarmycin in CD25-expressing cells upon exposure to NIR light. In vitro, CD25-targeted NIR-DPR significantly increased apoptosis of CD25-expressing HT2-A5E cells. When tumors were irradiated with NIR light in vivo, intratumoral CD25+ Treg populations decreased and Ki-67 and Interleukin-10 expression was suppressed, indicating impaired functioning of intratumoral CD25+ Tregs. CD25-targeted NIR-DPR suppressed tumor growth and improved survival in syngeneic murine tumor models. Of note, CD25-targeted NIR-DPR synergistically enhanced the efficacy of PD-1 blockade, especially in tumors with higher CD8+/Treg PD-1 ratios. Furthermore, the combination therapy induced significant anti-cancer immunity including maturation of dendritic cells, extensive intratumoral infiltration of cytotoxic CD8+ T cells, and increased differentiation into CD8+ memory T cells. Altogether, CD25-targeted NIR-DPR locally and predominantly targets Tregs in the tumor microenvironment and synergistically improves the efficacy of PD-1 blockade, suggesting that this combination therapy can be a rational anti-cancer combination immunotherapy.

Long-term engraftment and maturation of autologous iPSC-derived cardiomyocytes in two rhesus macaques.

Cellular therapies with cardiomyocytes produced from induced pluripotent stem cells (iPSC-CMs) offer a potential route to cardiac regeneration as a treatment for chronic ischemic heart disease. Here, we report successful long-term engraftment and in vivo maturation of autologous iPSC-CMs in two rhesus macaques with small, subclinical chronic myocardial infarctions, all without immunosuppression. Longitudinal positron emission tomography imaging using the sodium/iodide symporter (NIS) reporter gene revealed stable grafts for over 6 and 12 months, with no teratoma formation. Histological analyses suggested capability of the transplanted iPSC-CMs to mature and integrate with endogenous myocardium, with no sign of immune cell infiltration or rejection. By contrast, allogeneic iPSC-CMs were rejected within 8 weeks of transplantation. This study provides the longest-term safety and maturation data to date in any large animal model, addresses concerns regarding neoantigen immunoreactivity of autologous iPSC therapies, and suggests that autologous iPSC-CMs would similarly engraft and mature in human hearts.

Evaluation of a Cascaded Deep Learning-based Algorithm for Prostate Lesion Detection at Biparametric MRI.

Background Multiparametric MRI (mpMRI) improves prostate cancer (PCa) detection compared with systematic biopsy, but its interpretation is prone to interreader variation, which results in performance inconsistency. Artificial intelligence (AI) models can assist in mpMRI interpretation, but large training data sets and extensive model testing are required. Purpose To evaluate a biparametric MRI AI algorithm for intraprostatic lesion detection and segmentation and to compare its performance with radiologist readings and biopsy results. Materials and Methods This secondary analysis of a prospective registry included consecutive patients with suspected or known PCa who underwent mpMRI, US-guided systematic biopsy, or combined systematic and MRI/US fusion-guided biopsy between April 2019 and September 2022. All lesions were prospectively evaluated using Prostate Imaging Reporting and Data System version 2.1. The lesion- and participant-level performance of a previously developed cascaded deep learning algorithm was compared with histopathologic outcomes and radiologist readings using sensitivity, positive predictive value (PPV), and Dice similarity coefficient (DSC). Results A total of 658 male participants (median age, 67 years [IQR, 61-71 years]) with 1029 MRI-visible lesions were included. At histopathologic analysis, 45% (294 of 658) of participants had lesions of International Society of Urological Pathology (ISUP) grade group (GG) 2 or higher. The algorithm identified 96% (282 of 294; 95% CI: 94%, 98%) of all participants with clinically significant PCa, whereas the radiologist identified 98% (287 of 294; 95% CI: 96%, 99%; P = .23). The algorithm identified 84% (103 of 122), 96% (152 of 159), 96% (47 of 49), 95% (38 of 40), and 98% (45 of 46) of participants with ISUP GG 1, 2, 3, 4, and 5 lesions, respectively. In the lesion-level analysis using radiologist ground truth, the detection sensitivity was 55% (569 of 1029; 95% CI: 52%, 58%), and the PPV was 57% (535 of 934; 95% CI: 54%, 61%). The mean number of false-positive lesions per participant was 0.61 (range, 0-3). The lesion segmentation DSC was 0.29. Conclusion The AI algorithm detected cancer-suspicious lesions on biparametric MRI scans with a performance comparable to that of an experienced radiologist. Moreover, the algorithm reliably predicted clinically significant lesions at histopathologic examination. ClinicalTrials.gov Identifier: NCT03354416 © RSNA, 2024 Supplemental material is available for this article.

Deep learning-based whole-body PSMA PET/CT attenuation correction utilizing Pix-2-Pix GAN.

PURPOSE: Sequential PET/CT studies oncology patients can undergo during their treatment follow-up course is limited by radiation dosage. We propose an artificial intelligence (AI) tool to produce attenuation-corrected PET (AC-PET) images from non-attenuation-corrected PET (NAC-PET) images to reduce need for low-dose CT scans. METHODS: A deep learning algorithm based on 2D Pix-2-Pix generative adversarial network (GAN) architecture was developed from paired AC-PET and NAC-PET images. 18F-DCFPyL PSMA PET-CT studies from 302 prostate cancer patients, split into training, validation, and testing cohorts (n = 183, 60, 59, respectively). Models were trained with two normalization strategies: Standard Uptake Value (SUV)-based and SUV-Nyul-based. Scan-level performance was evaluated by normalized mean square error (NMSE), mean absolute error (MAE), structural similarity index (SSIM), and peak signal-to-noise ratio (PSNR). Lesion-level analysis was performed in regions-of-interest prospectively from nuclear medicine physicians. SUV metrics were evaluated using intraclass correlation coefficient (ICC), repeatability coefficient (RC), and linear mixed-effects modeling. RESULTS: Median NMSE, MAE, SSIM, and PSNR were 13.26%, 3.59%, 0.891, and 26.82, respectively, in the independent test cohort. ICC for SUVmax and SUVmean were 0.88 and 0.89, which indicated a high correlation between original and AI-generated quantitative imaging markers. Lesion location, density (Hounsfield units), and lesion uptake were all shown to impact relative error in generated SUV metrics (all p < 0.05). CONCLUSION: The Pix-2-Pix GAN model for generating AC-PET demonstrates SUV metrics that highly correlate with original images. AI-generated PET images show clinical potential for reducing the need for CT scans for attenuation correction while preserving quantitative markers and image quality.

Salivary excretion of systemically injected [18F]DCFPyL in prostate cancer patients undergoing PSMA scans.

Introduction: Prostate-specific membrane antigen (PSMA) is present in high amounts in salivary glands, but it is unclear whether labeled binders of PSMA are excreted in the saliva. Methods: Ten patients with prostate cancer underwent whole-body [18F]DCFPyL PET/CT (NCT03181867), and saliva samples were collected between 0-120 minutes post-injection. [18F]DCFPyL salivary excretion was measured over 120 minutes and expressed as %ID/g. Protein-associated binding was estimated by the percentage of [18F]DCFPyL versus parent radiotracer. Results: All PET scans of 10 patients (69 ± 8 years) with histologically confirmed prostate cancer (PSA= 2.4 ± 2.4, and Gleason Grade = 6-9) showed high uptake of [18F]-DCFPyL in salivary glands while 8 patients demonstrated high uptake in the saliva at 45 minutes. The intact [18F]-DCFPyL (98%) was also confirmed in the saliva samples at 120 min with increasing salivary radioactivity between 30-120 min. Conclusion: Systemically injected [18F]DCFPyL shows salivary gland uptake, an increasing amount of which is secreted in saliva over time and is not maximized by 120 minutes post-injection. Although probably insignificant for diagnostic studies, patients undergoing PSMA-targeted therapies should be aware of radioactivity in saliva.

Reducing False-Positives Due to Urinary Stagnation in the Prostatic Urethra on 18 F-DCFPyL PSMA PET/CT With MRI.

PURPOSE: Prostate-specific membrane antigen (PSMA)-targeting PET radiotracers reveal physiologic uptake in the urinary system, potentially misrepresenting activity in the prostatic urethra as an intraprostatic lesion. This study examined the correlation between midline 18 F-DCFPyL activity in the prostate and hyperintensity on T2-weighted (T2W) MRI as an indication of retained urine in the prostatic urethra. PATIENTS AND METHODS: Eighty-five patients who underwent both 18 F-DCFPyL PSMA PET/CT and prostate MRI between July 2017 and September 2023 were retrospectively analyzed for midline radiotracer activity and retained urine on postvoid T2W MRIs. Fisher's exact tests and unpaired t tests were used to compare residual urine presence and prostatic urethra measurements between patients with and without midline radiotracer activity. The influence of anatomical factors including prostate volume and urethral curvature on urinary stagnation was also explored. RESULTS: Midline activity on PSMA PET imaging was seen in 14 patients included in the case group, whereas the remaining 71 with no midline activity constituted the control group. A total of 71.4% (10/14) and 29.6% (21/71) of patients in the case and control groups had urethral hyperintensity on T2W MRI, respectively ( P < 0.01). Patients in the case group had significantly larger mean urethral dimensions, larger prostate volumes, and higher incidence of severe urethral curvature compared with the controls. CONCLUSIONS: Stagnated urine within the prostatic urethra is a potential confounding factor on PSMA PET scans. Integrating PET imaging with T2W MRI can mitigate false-positive calls, especially as PSMA PET/CT continues to gain traction in diagnosing localized prostate cancer.

The role of interventional radiology and molecular imaging for near infrared photoimmunotherapy.

Near infrared photoimmunotherapy (NIR-PIT) is a recently approved cancer therapy for recurrent head and neck cancer. It involves the intravenous administration of an antibody-photoabsorber (IRDye700DX: IR700) conjugate (APC) to target cancer cells, followed 24 h later by exposure to near infrared light to activate cell-specific cytotoxicity. NIR-PIT selectively targets cancer cells for destruction and activates a strong anticancer host immunity. The fluorescent signal emitted by IR700 enables the visualization of the APC in vivo using fluorescence imaging. Similarly, the activation of IR700 during therapy can be monitored by loss of fluorescence. NIR-PIT can be used with a variety of antibodies and therefore, a variety of cancer types. However, in most cases, NIR-PIT requires direct light exposure only achieved with interstitial diffuser light fibers that are placed with image-guided interventional needle insertion. In addition, the unique nature of NIR-PIT cell death, means that metabolic molecular imaging techniques such as PET and diffusion MRI can be used to assess therapeutic outcomes. This mini-review focuses on the potential implications of NIR-PIT for interventional radiology and therapeutic monitoring.

Near-infrared photoimmunotherapy targeting PD-L1: Improved efficacy by preconditioning the tumor microenvironment.

Near-infrared photoimmunotherapy (NIR-PIT) is a new type of cancer therapy that employs antibody-IRDye700DX conjugates (AbPCs) and near-infrared (NIR) light at a wavelength of 689 nm, the excitation wavelength of IR700. Administered intravenously, injected AbPCs bind specifically to cells expressing the target antigen, whereupon NIR light exposure causes rapid, selective killing. This process induces an anticancer T cell response, leading to sustained anticancer host immune response. Programmed cell death ligand-1 (PD-L1) is a major inhibitory immune checkpoint molecule expressed in various cancers. In this study, we first assessed the efficacy of PD-L1-targeted NIR-PIT (αPD-L1-PIT) in immune-competent tumor mouse models. αPD-L1-PIT showed a significant therapeutic effect on the tumor models with high PD-L1 expression. Furthermore, αPD-L1-PIT induced an abscopal effect on distant tumors and long-term immunological memory. In contrast, αPD-L1-PIT was not as effective for tumor models with low PD-L1 expression. To improve the efficacy of PD-L1-targeted NIR-PIT, PEGylated interferon-gamma (IFNγ) was administered with αPD-L1-PIT. The combination therapy improved the treatment efficacy by increasing PD-L1 expression leading to more efficient cell killing by αPD-L1-PIT. Furthermore, the PEGylated IFNγ led to a CD8+ T cell-dominant tumor microenvironment (TME) with an enhanced anticancer T cell response after αPD-L1-PIT. As a result, even so-called cold tumors exhibited complete responses after αPD-L1-PIT. Thus, combination therapy of PEGylated IFNγ and PD-L1-targeted NIR-PIT has the potential to be an important future strategy for cancer immunotherapy.

A Mathematical Model of TCR-T Cell Therapy for Cervical Cancer.

Engineered T cell receptor (TCR)-expressing T (TCR-T) cells are intended to drive strong anti-tumor responses upon recognition of the specific cancer antigen, resulting in rapid expansion in the number of TCR-T cells and enhanced cytotoxic functions, causing cancer cell death. However, although TCR-T cell therapy against cancers has shown promising results, it remains difficult to predict which patients will benefit from such therapy. We develop a mathematical model to identify mechanisms associated with an insufficient response in a mouse cancer model. We consider a dynamical system that follows the population of cancer cells, effector TCR-T cells, regulatory T cells (Tregs), and "non-cancer-killing" TCR-T cells. We demonstrate that the majority of TCR-T cells within the tumor are "non-cancer-killing" TCR-T cells, such as exhausted cells, which contribute little or no direct cytotoxicity in the tumor microenvironment (TME). We also establish two important factors influencing tumor regression: the reversal of the immunosuppressive TME following depletion of Tregs, and the increased number of effector TCR-T cells with antitumor activity. Using mathematical modeling, we show that certain parameters, such as increasing the cytotoxicity of effector TCR-T cells and modifying the number of TCR-T cells, play important roles in determining outcomes.

Automated Detection and Grading of Extraprostatic Extension of Prostate Cancer at MRI via Cascaded Deep Learning and Random Forest Classification.

RATIONALE AND OBJECTIVES: Extraprostatic extension (EPE) is well established as a significant predictor of prostate cancer aggression and recurrence. Accurate EPE assessment prior to radical prostatectomy can impact surgical approach. We aimed to utilize a deep learning-based AI workflow for automated EPE grading from prostate T2W MRI, ADC map, and High B DWI. MATERIAL AND METHODS: An expert genitourinary radiologist conducted prospective clinical assessments of MRI scans for 634 patients and assigned risk for EPE using a grading technique. The training set and held-out independent test set consisted of 507 patients and 127 patients, respectively. Existing deep-learning AI models for prostate organ and lesion segmentation were leveraged to extract area and distance features for random forest classification models. Model performance was evaluated using balanced accuracy, ROC AUCs for each EPE grade, as well as sensitivity, specificity, and accuracy compared to EPE on histopathology. RESULTS: A balanced accuracy score of .390 ± 0.078 was achieved using a lesion detection probability threshold of 0.45 and distance features. Using the test set, ROC AUCs for AI-assigned EPE grades 0-3 were 0.70, 0.65, 0.68, and 0.55 respectively. When using EPE≥ 1 as the threshold for positive EPE, the model achieved a sensitivity of 0.67, specificity of 0.73, and accuracy of 0.72 compared to radiologist sensitivity of 0.81, specificity of 0.62, and accuracy of 0.66 using histopathology as the ground truth. CONCLUSION: Our AI workflow for assigning imaging-based EPE grades achieves an accuracy for predicting histologic EPE approaching that of physicians. This automated workflow has the potential to enhance physician decision-making for assessing the risk of EPE in patients undergoing treatment for prostate cancer due to its consistency and automation.