The impact of the co-registration technique and analysis methodology in comparison studies between advanced imaging modalities and whole-mount-histology reference in primary prostate cancer

RAPHIA: A deep learning pipeline for the registration of MRI and whole-mount histopathology images of the prostate

Image registration can map the ground truth extent of prostate cancer from histopathology images onto MRI, facilitating the development of machine learning methods for early prostate cancer detection. Here, we present RAdiology PatHology Image Alignment (RAPHIA), an end-to-end pipeline for efficient and accurate registration of MRI and histopathology images. RAPHIA automates several time-consuming manual steps in existing approaches including prostate segmentation, estimation of the rotation angle and horizontal flipping in histopathology images, and estimation of MRI-histopathology slice correspondences. By utilizing deep learning registration networks, RAPHIA substantially reduces computational time. Furthermore, RAPHIA obviates the need for a multimodal image similarity metric by transferring histopathology image representations to MRI image representations and vice versa. With the assistance of RAPHIA, novice users achieved expert-level performance, and their mean error in estimating histopathology rotation angle was reduced by 51% (12 degrees vs 8 degrees), their mean accuracy of estimating histopathology flipping was increased by 5% (95.3% vs 100%), and their mean error in estimating MRI-histopathology slice correspondences was reduced by 45% (1.12 slices vs 0.62 slices). When compared to a recent conventional registration approach and a deep learning registration approach, RAPHIA achieved better mapping of histopathology cancer labels, with an improved mean Dice coefficient of cancer regions outlined on MRI and the deformed histopathology (0.44 vs 0.48 vs 0.50), and a reduced mean per-case processing time (51 vs 11 vs 4.5 min). The improved performance by RAPHIA allows efficient processing of large datasets for the development of machine learning models for prostate cancer detection on MRI. Our code is publicly available at: https://github.com/pimed/RAPHIA.

Prostate-specific membrane antigen positron emission tomography in addition to multiparametric magnetic resonance imaging and biopsies to select prostate cancer patients for focal therapy

OBJECTIVE

To evaluate the additional value of prostate-specific membrane antigen positron emission tomography (PSMA-PET) to conventional diagnostic tools to select patients for hemi-ablative focal therapy (FT).

PATIENTS AND METHODS

We performed a retrospective analysis on a multicentre cohort (private and institutional) of 138 patients who underwent multiparametric magnetic resonance imaging (mpMRI), PSMA-PET, and systematic biopsies prior to radical prostatectomy between January 2011 and July 2021. Patients were eligible when they met the consensus criteria for FT: PSA <15 ng/mL, clinical/radiological T stage ≤T2b, and International Society of Urological Pathology (ISUP) grade 2-3. Clinically significant prostate cancer (csPCa) was defined as ISUP grade ≥2, extracapsular extension >0.5 mm or seminal vesicle involvement at final histopathology. The diagnostic accuracy of mpMRI, systematic biopsies and PSMA-PET for csPCa (separate and combined) was calculated within a four-quadrant prostate model by receiver-operating characteristic and 2 × 2 contingency analysis. Additionally, we assessed whether the diagnostic tools correctly identified patients suitable for hemi-ablative FT.

RESULTS

In total 552 prostate quadrants were analysed and 272 (49%) contained csPCa on final histopathology. The area under the curve, sensitivity, specificity, positive predictive value and negative predictive value for csPCa were 0.79, 75%, 83%, 81% and 77%, respectively, for combined mpMRI and systematic biopsies, and improved after addition of PSMA-PET to 0.84, 87%, 80%, 81% and 86%, respectively (P < 0.001). On final histopathology 46/138 patients (33%) were not suitable for hemi-ablative FT. Addition of PSMA-PET correctly identified 26/46 (57%) non-suitable patients and resulted in 4/138 (3%) false-positive exclusions.

CONCLUSIONS

Addition of PSMA-PET to the conventional work-up by mpMRI and systematic biopsies could improve selection for hemi-ablative FT and guide exclusion of patients for whom whole-gland treatments might be a more suitable treatment option.

Using multiparametric Magnetic Resonance Imaging and Prostate Specific Membrane Antigen Positron Emission Tomography to detect and delineate the gross tumour volume of intraprostatic lesions - A systematic review and meta-analysis

BACKGROUND AND PURPOSE

Radiation therapy is used frequently for patients with prostate cancer. Dose escalation to intraprostatic lesions (IPLs) has been shown to improve oncologic outcomes, without increasing toxicity. Both multiparametric MRI (mpMRI) and PSMA PET can be used to identify IPLs.

MATERIALS AND METHODS

A systematic review was conducted to determine the ability of mpMRI, PSMA PET and their combination to detect IPLs prior to radical prostatectomy (RP) as correlated with the histology. Trials included patients that had mpMRI, PSMA PET, or both, prior to RP. The quality of the histopathological-radiological co-registration was assessed as high or low for each study. Recorded outcomes include sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). A meta-analysis was conducted using a bivariate model to determine the pooled sensitivity and specificity for each imaging modality. This systematic review was registered through PROSPERO (CRD42023389092).

RESULTS

Altogether, 42 studies were included in the systematic review. Of these, 20 could be included in the meta-analysis. The pooled sensitivity (95 % CI), specificity (95 % CI) and AUROC for mpMRI (n = 13 studies) were 64.7 % (50.2 % - 76.9 %), 86.4 % (79.7 % - 91.1 %), and 0.852; the pooled outcomes for PSMA PET (n = 12) were 75.7 % (64.0 % - 84.5 %), 87.1 % (80.2 % - 91.9 %), and 0.889; for their combination (n = 5), the pooled outcomes were 70.3 % (64.1 % - 75.9 %), 81.9 % (71.9 % - 88.8 %), and 0.796. When reviewing studies with a high-quality histopathological-radiological co-registration, IPL delineation recommendations varied by study and the imaging modality used.

CONCLUSION

All of mpMRI, PSMA PET or their combination were found to have very good diagnostic outcomes for detecting IPLs. Recommendations for delineating IPLs varied based on the imaging modalities used and between research groups. Consensus guidelines for IPL delineation would help with creating consistency for focal boost radiation treatments in future studies.

Digital image analysis and machine learning-assisted prediction of neoadjuvant chemotherapy response in triple-negative breast cancer

BACKGROUND

Pathological complete response (pCR) is associated with favorable prognosis in patients with triple-negative breast cancer (TNBC). However, only 30-40% of TNBC patients treated with neoadjuvant chemotherapy (NAC) show pCR, while the remaining 60-70% show residual disease (RD). The role of the tumor microenvironment in NAC response in patients with TNBC remains unclear. In this study, we developed a machine learning-based two-step pipeline to distinguish between various histological components in hematoxylin and eosin (H&E)-stained whole slide images (WSIs) of TNBC tissue biopsies and to identify histological features that can predict NAC response.

METHODS

H&E-stained WSIs of treatment-naïve biopsies from 85 patients (51 with pCR and 34 with RD) of the model development cohort and 79 patients (41 with pCR and 38 with RD) of the validation cohort were separated through a stratified eightfold cross-validation strategy for the first step and leave-one-out cross-validation strategy for the second step. A tile-level histology label prediction pipeline and four machine-learning classifiers were used to analyze 468,043 tiles of WSIs. The best-trained classifier used 55 texture features from each tile to produce a probability profile during testing. The predicted histology classes were used to generate a histology classification map of the spatial distributions of different tissue regions. A patient-level NAC response prediction pipeline was trained with features derived from paired histology classification maps. The top graph-based features capturing the relevant spatial information across the different histological classes were provided to the radial basis function kernel support vector machine (rbfSVM) classifier for NAC treatment response prediction.

RESULTS

The tile-level prediction pipeline achieved 86.72% accuracy for histology class classification, while the patient-level pipeline achieved 83.53% NAC response (pCR vs. RD) prediction accuracy of the model development cohort. The model was validated with an independent cohort with tile histology validation accuracy of 83.59% and NAC prediction accuracy of 81.01%. The histological class pairs with the strongest NAC response predictive ability were tumor and tumor tumor-infiltrating lymphocytes for pCR and microvessel density and polyploid giant cancer cells for RD.

CONCLUSION

Our machine learning pipeline can robustly identify clinically relevant histological classes that predict NAC response in TNBC patients and may help guide patient selection for NAC treatment.

Digital image analysis and machine learning-assisted prediction of neoadjuvant chemotherapy response in triple-negative breast cancer

Background

Pathological complete response (pCR) is associated with favorable prognosis in patients with triple-negative breast cancer (TNBC). However, only 30-40% of TNBC patients treated with neoadjuvant chemotherapy (NAC) show pCR, while the remaining 60-70% show residual disease (RD). The role of the tumor microenvironment (TME) in NAC response in patients with TNBC remains unclear. In this study, we developed a machine learning-based two-step pipeline to distinguish between various histological components in hematoxylin and eosin (H&E)-stained whole slide images (WSIs) of TNBC tissue biopsies and to identify histological features that can predict NAC response.

Methods

H&E-stained WSIs of treatment-naïve biopsies from 85 patients (51 with pCR and 34 with RD) were separated through a stratified 8-fold cross validation strategy for the first step and leave one out cross validation strategy for the second step. A tile-level histology label prediction pipeline and four machine learning classifiers were used to analyze 468,043 tiles of WSIs. The best-trained classifier used 55 texture features from each tile to produce a probability profile during testing. The predicted histology classes were used to generate a histology classification map of the spatial distributions of different tissue regions. A patient-level NAC response prediction pipeline was trained with features derived from paired histology classification maps. The top graph-based features capturing the relevant spatial information across the different histological classes were provided to the radial basis function kernel support vector machine (rbfSVM) classifier for NAC treatment response prediction.

Results

The tile-level prediction pipeline achieved 86.72% accuracy for histology class classification, while the patient-level pipeline achieved 83.53% NAC response (pCR vs. RD) prediction accuracy. The histological class pairs with the strongest NAC response predictive ability were tumor and tumor tumor-infiltrating lymphocytes for pCR and microvessel density and polyploid giant cancer cells for RD.

Conclusion

Our machine learning pipeline can robustly identify clinically relevant histological classes that predict NAC response in TNBC patients and may help guide patient selection for NAC treatment.

[68Ga]Ga-PSMA and [68Ga]Ga-RM2 PET/MRI vs. Histopathological Images in Prostate Cancer: A New Workflow for Spatial Co-Registration

This study proposed a new workflow for co-registering prostate PET images from a dual-tracer PET/MRI study with histopathological images of resected prostate specimens. The method aims to establish an accurate correspondence between PET/MRI findings and histology, facilitating a deeper understanding of PET tracer distribution and enabling advanced analyses like radiomics. To achieve this, images derived by three patients who underwent both [68Ga]Ga-PSMA and [68Ga]Ga-RM2 PET/MRI before radical prostatectomy were selected. After surgery, in the resected fresh specimens, fiducial markers visible on both histology and MR images were inserted. An ex vivo MRI of the prostate served as an intermediate step for co-registration between histological specimens and in vivo MRI examinations. The co-registration workflow involved five steps, ensuring alignment between histopathological images and PET/MRI data. The target registration error (TRE) was calculated to assess the precision of the co-registration. Furthermore, the DICE score was computed between the dominant intraprostatic tumor lesions delineated by the pathologist and the nuclear medicine physician. The TRE for the co-registration of histopathology and in vivo images was 1.59 mm, while the DICE score related to the site of increased intraprostatic uptake on [68Ga]Ga-PSMA and [68Ga]Ga-RM2 PET images was 0.54 and 0.75, respectively. This work shows an accurate co-registration method for histopathological and in vivo PET/MRI prostate examinations that allows the quantitative assessment of dual-tracer PET/MRI diagnostic accuracy at a millimetric scale. This approach may unveil radiotracer uptake mechanisms and identify new PET/MRI biomarkers, thus establishing the basis for precision medicine and future analyses, such as radiomics.

Comparison of Multiparametric Magnetic Resonance Imaging with Prostate-Specific Membrane Antigen Positron-Emission Tomography Imaging in Primary Prostate Cancer Diagnosis: A Systematic Review and Meta-Analysis

Multiparametric magnetic-resonance imaging (mpMRI) has proven utility in diagnosing primary prostate cancer. However, the diagnostic potential of prostate-specific membrane antigen positron-emission tomography (PSMA PET) has yet to be established. This study aims to systematically review the current literature comparing the diagnostic performance of mpMRI and PSMA PET imaging to diagnose primary prostate cancer. A systematic literature search was performed up to December 2021. Quality analyses were conducted using the QUADAS-2 tool. The reference standard was whole-mount prostatectomy or prostate biopsy. Statistical analysis involved the pooling of the reported diagnostic performances of each modality, and differences in per-patient and per-lesion analysis were compared using a Fisher’s exact test. Ten articles were included in the meta-analysis. At a per-patient level, the pooled values of sensitivity, specificity, and area under the curve (AUC) for mpMRI and PSMA PET/CT were 0.87 (95% CI: 0.83−0.91) vs. 0.93 (95% CI: 0.90−0.96, p < 0.01); 0.47 (95% CI: 0.23−0.71) vs. 0.54 (95% CI: 0.23−0.84, p > 0.05); and 0.84 vs. 0.91, respectively. At a per-lesion level, the pooled sensitivity, specificity, and AUC value for mpMRI and PSMA PET/CT were lower, at 0.63 (95% CI: 0.52−0.74) vs. 0.79 (95% CI: 0.62−0.92, p < 0.001); 0.88 (95% CI: 0.81−0.95) vs. 0.71 (95% CI: 0.47−0.90, p < 0.05); and 0.83 vs. 0.84, respectively. High heterogeneity was observed between studies. PSMA PET/CT may better confirm the presence of prostate cancer than mpMRI. However, both modalities appear comparable in determining the localisation of the lesions.

Correlation Between Quantitative PSMA PET Parameters and Clinical Risk Factors in Non-Metastatic Primary Prostate Cancer Patients

Background

PSMA PET is frequently used for staging of prostate cancer patients. Furthermore, there is increasing interest to use PET information for personalized local treatment approaches in surgery and radiotherapy, especially for focal treatment strategies. However, it is not well established which quantitative imaging parameters show highest correlation with clinical and histological tumor aggressiveness.

Methods

This is a retrospective analysis of 135 consecutive patients with non-metastatic prostate cancer and PSMA PET before any treatment. Clinical risk parameters (PSA values, Gleason score and D'Amico risk group) were correlated with quantitative PET parameters maximum standardized uptake value (SUVmax), mean SUV (SUVmean), tumor asphericity (ASP) and PSMA tumor volume (PSMA-TV).

Results

Most of the investigated imaging parameters were highly correlated with each other (correlation coefficients between 0.20 and 0.95). A low to moderate, however significant, correlation of imaging parameters with PSA values (0.19 to 0.45) and with Gleason scores (0.17 to 0.31) was observed for all parameters except ASP which did not show a significant correlation with Gleason score. Receiver operating characteristics for the detection of D'Amico high-risk patients showed poor to fair sensitivity and specificity for all investigated quantitative PSMA PET parameters (Areas under the curve (AUC) between 0.63 and 0.73). Comparison of AUC between quantitative PET parameters by DeLong test showed significant superiority of SUVmax compared to SUVmean for the detection of high-risk patients. None of the investigated imaging parameters significantly outperformed SUVmax.

Conclusion

Our data confirm prior publications with lower number of patients that reported moderate correlations of PSMA PET parameters with clinical risk factors. With the important limitation that Gleason scores were only biopsy-derived in this study, there is no indication that the investigated additional parameters deliver superior information compared to SUVmax.

Explainable AI for CNN-based prostate tumor segmentation in multi-parametric MRI correlated to whole mount histopathology

Automatic prostate tumor segmentation is often unable to identify the lesion even if multi-parametric MRI data is used as input, and the segmentation output is difficult to verify due to the lack of clinically established ground truth images. In this work we use an explainable deep learning model to interpret the predictions of a convolutional neural network (CNN) for prostate tumor segmentation. The CNN uses a U-Net architecture which was trained on multi-parametric MRI data from 122 patients to automatically segment the prostate gland and prostate tumor lesions. In addition, co-registered ground truth data from whole mount histopathology images were available in 15 patients that were used as a test set during CNN testing. To be able to interpret the segmentation results of the CNN, heat maps were generated using the Gradient Weighted Class Activation Map (Grad-CAM) method. The CNN achieved a mean Dice Sorensen Coefficient 0.62 and 0.31 for the prostate gland and the tumor lesions -with the radiologist drawn ground truth and 0.32 with whole-mount histology ground truth for tumor lesions. Dice Sorensen Coefficient between CNN predictions and manual segmentations from MRI and histology data were not significantly different. In the prostate the Grad-CAM heat maps could differentiate between tumor and healthy prostate tissue, which indicates that the image information in the tumor was essential for the CNN segmentation.

Feasibility, pitfalls and results of a structured concept-development phase for a randomized controlled phase III trial on radiotherapy in primary prostate cancer patients

Objective

Failure rate in randomized controlled trials (RCTs) is > 50%, includes safety-problems, underpowered statistics, lack of efficacy, lack of funding or insufficient patient recruitment and is even more pronounced in oncology trials. We present results of a structured concept-development phase (CDP) for a phase III RCT on personalized radiotherapy (RT) in primary prostate cancer (PCa) patients implementing prostate specific membrane antigen targeting positron emission tomography (PSMA-PET).

Materials and methods

The 1 yr process of the CDP contained five main working packages: (i) literature search and scoping review, (ii) involvement of individual patients, patients’ representatives and patients’ self-help groups addressing the patients’ willingness to participate in the preparation process and the conduct of RCTs as well as the patient informed consent (PIC), (iii) involvement of national and international experts and expert panels (iv) a phase II pilot study investigating the safety of implementation of PSMA-PET for focal dose escalation RT and (v) in-silico RT planning studies assessing feasibility of envisaged dose regimens and effects of urethral sparing in focal dose escalation.

Results

(i) Systematic literature searches confirmed the high clinical relevance for more evidence on advanced RT approaches, in particular stereotactic body RT, in high-risk PCa patients. (ii) Involvement of patients, patient representatives and randomly selected males relevantly changed the PIC and initiated a patient empowerment project for training of bladder preparation. (iii) Discussion with national and international experts led to adaptions of inclusion and exclusion criteria. (iv) Fifty patients were treated in the pilot trial and in- and exclusion criteria as well as enrollment calculations were adapted accordingly. Parallel conduction of the pilot trial revealed pitfalls on practicability and broadened the horizon for translational projects. (v) In-silico planning studies confirmed feasibility of envisaged dose prescription. Despite large prostate- and boost-volumes of up to 66% of the prostate, adherence to stringent anorectal dose constraints was feasible. Urethral sparing increased the therapeutic ratio.

Conclusion

The dynamic framework of interdisciplinary working programs in CDPs enhances robustness of RCT protocols and may be associated with decreased failure rates. Structured recommendations are warranted to further define the process of such CDPs in radiation oncology trials.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-022-09434-2.

Investigating the heterogeneity of viscoelastic properties in prostate cancer using MR elastography at 9.4T in fresh prostatectomy specimens

PURPOSE

To quantify the heterogeneity of viscoelastic tissue properties in prostatectomy specimens from men with prostate cancer (PC) using MR elastography (MRE) with histopathology as reference.

METHODS

Twelve fresh prostatectomy specimens were examined in a preclinical 9.4T MRI scanner. Maps of the complex shear modulus (|G*| in kPa) with its real and imaginary part (G' and G" in kPa) were calculated at 500 Hz. Prostates were divided into 12 segments for segment-wise measurement of viscoelastic properties and histopathology. Coefficients of variation (CVs in %) were calculated for quantification of heterogeneity.

RESULTS

Group-averaged values of cancerous vs. benign segments were significantly increased: |G*| of 12.13 kPa vs. 6.14 kPa, G' of 10.84 kPa vs. 5.44 kPa and G" of 5.45 kPa vs. 2.92 kPa, all p < 0.001. In contrast, CVs were significantly increased for benign segments: 23.59% vs. 26.32% (p = 0.014) for |G*|, 27.05% vs. 37.84% (p < 0.003) for G', and 36.51% vs. 50.37% (p = 0.008) for G".

DISCUSSION

PC is characterized by a stiff yet homogeneous biomechanical signature, which may be due to the unique nondestructive growth pattern of PC with intervening stroma, providing a rigid scaffold in the affected area. In turn, increased heterogeneity in benign prostate segments may be attributable to the presence of different prostate zones with involvement by specific nonmalignant pathology.

Radiomics in prostate cancer: an up-to-date review

Prostate cancer (PCa) is the most common worldwide diagnosed malignancy in male population. The diagnosis, the identification of aggressive disease, and the post-treatment follow-up needs a more comprehensive and holistic approach. Radiomics is the extraction and interpretation of images phenotypes in a quantitative manner. Radiomics may give an advantage through advancements in imaging modalities and through the potential power of artificial intelligence techniques by translating those features into clinical outcome prediction. This article gives an overview on the current evidence of methodology and reviews the available literature on radiomics in PCa patients, highlighting its potential for personalized treatment and future applications.

A review of artificial intelligence in prostate cancer detection on imaging

A multitude of studies have explored the role of artificial intelligence (AI) in providing diagnostic support to radiologists, pathologists, and urologists in prostate cancer detection, risk-stratification, and management. This review provides a comprehensive overview of relevant literature regarding the use of AI models in (1) detecting prostate cancer on radiology images (magnetic resonance and ultrasound imaging), (2) detecting prostate cancer on histopathology images of prostate biopsy tissue, and (3) assisting in supporting tasks for prostate cancer detection (prostate gland segmentation, MRI-histopathology registration, MRI-ultrasound registration). We discuss both the potential of these AI models to assist in the clinical workflow of prostate cancer diagnosis, as well as the current limitations including variability in training data sets, algorithms, and evaluation criteria. We also discuss ongoing challenges and what is needed to bridge the gap between academic research on AI for prostate cancer and commercial solutions that improve routine clinical care.

PSMA-PET/MRI-Based Focal Dose Escalation in Patients with Primary Prostate Cancer Treated with Stereotactic Body Radiation Therapy (HypoFocal-SBRT): Study Protocol of a Randomized, Multicentric Phase III Trial

Technical advances in radiotherapy (RT) treatment planning and delivery have substantially changed RT concepts for primary prostate cancer (PCa) by (i) enabling a reduction of treatment time, and by (ii) enabling safe delivery of high RT doses. Several studies proposed a dose-response relationship for patients with primary PCa and especially in patients with high-risk features, as dose escalation leads to improved tumor control. In parallel to the improvements in RT techniques, diagnostic imaging techniques like multiparametric magnetic resonance imaging (mpMRI) and positron-emission tomography targeting prostate-specific-membrane antigen (PSMA-PET) evolved and enable an accurate depiction of the intraprostatic tumor mass for the first time. The HypoFocal-SBRT study combines ultra-hypofractionated RT/stereotactic body RT, with focal RT dose escalation on intraprostatic tumor sides by applying state of the art diagnostic imaging and most modern RT concepts. This novel strategy will be compared with moderate hypofractionated RT (MHRT), one option for the curative primary treatment of PCa, which has been proven by several prospective trials and is recommended and carried out worldwide. We suspect an increase in relapse-free survival (RFS), and we will assess quality of life in order to detect potential changes.

Radiomics in prostate cancer imaging for a personalized treatment approach - current aspects of methodology and a systematic review on validated studies

Prostate cancer (PCa) is one of the most frequently diagnosed malignancies of men in the world. Due to a variety of treatment options in different risk groups, proper diagnostic and risk stratification is pivotal in treatment of PCa. The development of precise medical imaging procedures simultaneously to improvements in big data analysis has led to the establishment of radiomics - a computer-based method of extracting and analyzing image features quantitatively. This approach bears the potential to assess and improve PCa detection, tissue characterization and clinical outcome prediction. This article gives an overview on the current aspects of methodology and systematically reviews available literature on radiomics in PCa patients, showing its potential for personalized therapy approaches. The qualitative synthesis includes all imaging modalities and focuses on validated studies, putting forward future directions.