Semisupervised Learning with Report-guided Pseudo Labels for Deep Learning-based Prostate Cancer Detection Using Biparametric MRI

Purpose

To evaluate a novel method of semisupervised learning (SSL) guided by automated sparse information from diagnostic reports to leverage additional data for deep learning-based malignancy detection in patients with clinically significant prostate cancer.

Materials and Methods

This retrospective study included 7756 prostate MRI examinations (6380 patients) performed between January 2014 and December 2020 for model development. An SSL method, report-guided SSL (RG-SSL), was developed for detection of clinically significant prostate cancer using biparametric MRI. RG-SSL, supervised learning (SL), and state-of-the-art SSL methods were trained using 100, 300, 1000, or 3050 manually annotated examinations. Performance on detection of clinically significant prostate cancer by RG-SSL, SL, and SSL was compared on 300 unseen examinations from an external center with a histopathologically confirmed reference standard. Performance was evaluated using receiver operating characteristic (ROC) and free-response ROC analysis. P values for performance differences were generated with a permutation test.

Results

At 100 manually annotated examinations, mean examination-based diagnostic area under the ROC curve (AUC) values for RG-SSL, SL, and the best SSL were 0.86 ± 0.01 (SD), 0.78 ± 0.03, and 0.81 ± 0.02, respectively. Lesion-based detection partial AUCs were 0.62 ± 0.02, 0.44 ± 0.04, and 0.48 ± 0.09, respectively. Examination-based performance of SL with 3050 examinations was matched by RG-SSL with 169 manually annotated examinations, thus requiring 14 times fewer annotations. Lesion-based performance was matched with 431 manually annotated examinations, requiring six times fewer annotations.

Conclusion

RG-SSL outperformed SSL in clinically significant prostate cancer detection and achieved performance similar to SL even at very low annotation budgets.Keywords: Annotation Efficiency, Computer-aided Detection and Diagnosis, MRI, Prostate Cancer, Semisupervised Deep Learning Supplemental material is available for this article. Published under a CC BY 4.0 license.

Artificial intelligence for prostate MRI: open datasets, available applications, and grand challenges

Artificial intelligence (AI) for prostate magnetic resonance imaging (MRI) is starting to play a clinical role for prostate cancer (PCa) patients. AI-assisted reading is feasible, allowing workflow reduction. A total of 3,369 multi-vendor prostate MRI cases are available in open datasets, acquired from 2003 to 2021 in Europe or USA at 3 T (n = 3,018; 89.6%) or 1.5 T (n = 296; 8.8%), 346 cases scanned with endorectal coil (10.3%), 3,023 (89.7%) with phased-array surface coils; 412 collected for anatomical segmentation tasks, 3,096 for PCa detection/classification; for 2,240 cases lesions delineation is available and 56 cases have matching histopathologic images; for 2,620 cases the PSA level is provided; the total size of all open datasets amounts to approximately 253 GB. Of note, quality of annotations provided per dataset highly differ and attention must be paid when using these datasets (e.g., data overlap). Seven grand challenges and commercial applications from eleven vendors are here considered. Few small studies provided prospective validation. More work is needed, in particular validation on large-scale multi-institutional, well-curated public datasets to test general applicability. Moreover, AI needs to be explored for clinical stages other than detection/characterization (e.g., follow-up, prognosis, interventions, and focal treatment).

Measuring Staff Satisfaction in Transportation System using AHP Method under Uncertainty

One of the most important challenges in today world is the transportation and satisfaction of this system. Measuring satisfaction in transportation has been done quantitatively so far and, more importantly, the measurement process has not usually been scientific. This study aimed at an accurate scientific measurement. Because of ambiguity, this paper discusses how to use the intuitionistic fuzzy method, in which both membership and non-membership functions were expressed. Therefore, initially, considering the results of previous researches, as well as studying the references and standards, the characteristics and basic criteria, a researcher-made questionnaire with a reliability of 0.95 was first performed and then the weights were determined using the AHP intuitionistic fuzzy method. After performing the above steps, the intuitionistic fuzzy satisfaction value was calculated at different levels and using the method, the final number of satisfaction was defuzzified.

Deep learning-assisted prostate cancer detection on bi-parametric MRI: minimum training data size requirements and effect of prior knowledge

Objectives

To assess Prostate Imaging Reporting and Data System (PI-RADS)–trained deep learning (DL) algorithm performance and to investigate the effect of data size and prior knowledge on the detection of clinically significant prostate cancer (csPCa) in biopsy-naïve men with a suspicion of PCa.

Methods

Multi-institution data included 2734 consecutive biopsy-naïve men with elevated PSA levels (≥ 3 ng/mL) that underwent multi-parametric MRI (mpMRI). mpMRI exams were prospectively reported using PI-RADS v2 by expert radiologists. A DL framework was designed and trained on center 1 data (n = 1952) to predict PI-RADS ≥ 4 (n = 1092) lesions from bi-parametric MRI (bpMRI). Experiments included varying the number of cases and the use of automatic zonal segmentation as a DL prior. Independent center 2 cases (n = 296) that included pathology outcome (systematic and MRI targeted biopsy) were used to compute performance for radiologists and DL. The performance of detecting PI-RADS 4–5 and Gleason > 6 lesions was assessed on 782 unseen cases (486 center 1, 296 center 2) using free-response ROC (FROC) and ROC analysis.

Results

The DL sensitivity for detecting PI-RADS ≥ 4 lesions was 87% (193/223, 95% CI: 82–91) at an average of 1 false positive (FP) per patient, and an AUC of 0.88 (95% CI: 0.84–0.91). The DL sensitivity for the detection of Gleason > 6 lesions was 85% (79/93, 95% CI: 77–83) @ 1 FP compared to 91% (85/93, 95% CI: 84–96) @ 0.3 FP for a consensus panel of expert radiologists. Data size and prior zonal knowledge significantly affected performance (4%, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p<0.05$$\end{document}p<0.05).

Conclusion

PI-RADS-trained DL can accurately detect and localize Gleason > 6 lesions. DL could reach expert performance using substantially more than 2000 training cases, and DL zonal segmentation.

Key Points

• AI for prostate MRI analysis depends strongly on data size and prior zonal knowledge.

• AI needs substantially more than 2000 training cases to achieve expert performance.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00330-021-08320-y.

End-to-end prostate cancer detection in bpMRI via 3D CNNs: Effects of attention mechanisms, clinical priori and decoupled false positive reduction

We present a multi-stage 3D computer-aided detection and diagnosis (CAD) model² for automated localization of clinically significant prostate cancer (csPCa) in bi-parametric MR imaging (bpMRI). Deep attention mechanisms drive its detection network, targeting salient structures and highly discriminative feature dimensions across multiple resolutions. Its goal is to accurately identify csPCa lesions from indolent cancer and the wide range of benign pathology that can afflict the prostate gland. Simultaneously, a decoupled residual classifier is used to achieve consistent false positive reduction, without sacrificing high sensitivity or computational efficiency. In order to guide model generalization with domain-specific clinical knowledge, a probabilistic anatomical prior is used to encode the spatial prevalence and zonal distinction of csPCa. Using a large dataset of 1950 prostate bpMRI paired with radiologically-estimated annotations, we hypothesize that such CNN-based models can be trained to detect biopsy-confirmed malignancies in an independent cohort. For 486 institutional testing scans, the 3D CAD system achieves 83.69±5.22% and 93.19±2.96% detection sensitivity at 0.50 and 1.46 false positive(s) per patient, respectively, with 0.882±0.030 AUROC in patient-based diagnosis -significantly outperforming four state-of-the-art baseline architectures (U-SEResNet, UNet++, nnU-Net, Attention U-Net) from recent literature. For 296 external biopsy-confirmed testing scans, the ensembled CAD system shares moderate agreement with a consensus of expert radiologists (76.69%; kappa = 0.51±0.04) and independent pathologists (81.08%; kappa = 0.56±0.06); demonstrating strong generalization to histologically-confirmed csPCa diagnosis.

Multi-Modal Siamese Network for Diagnostically Similar Lesion Retrieval in Prostate MRI

Multi-parametric prostate MRI (mpMRI) is a powerful tool to diagnose prostate cancer, though difficult to interpret even for experienced radiologists. A common radiological procedure is to compare a magnetic resonance image with similarly diagnosed cases. To assist the radiological image interpretation process, computerized Content-Based Image Retrieval systems (CBIRs) can therefore be employed to improve the reporting workflow and increase its accuracy. In this article, we propose a new, supervised siamese deep learning architecture able to handle multi-modal and multi-view MR images with similar PIRADS score. An experimental comparison with well-established deep learning-based CBIRs (namely standard siamese networks and autoencoders) showed significantly improved performance with respect to both diagnostic (ROC-AUC), and information retrieval metrics (Precision-Recall, Discounted Cumulative Gain and Mean Average Precision). Finally, the new proposed multi-view siamese network is general in design, facilitating a broad use in diagnostic medical imaging retrieval.

Interferometric optical testing to discriminate benign and malignant brain tumors

When normal cells suffer from the malignant disease, changes occur in cellular size and structure, and their constituents. These changes can affect the biological response of cells and tissues. Conventional hematoxylin and eosin staining, which is the gold standard of tumor diagnosis provide a very detailed view of the tissue by staining cell structures to diagnose a disease based on the deformation of the cells. However, this study shows changes in cells and tissues structure due to malignant diseases can also affect the optical properties. The purpose of this research was to assess and compare these effects to identify benign and malignant brain tissues from each other. Various samples of adult human brain tissues were studied using Mach-Zehnder interferometer as an optical method and using the Fourier transform method as an analytical process. The formed interference patterns of benign and malignant brain tumors were investigated to obtain the phase distribution of tumors. The obtained information demonstrated that the phase distribution of malignant brain tissues was different from benign brain tissues. The phase distribution approximately ranged from 10 to 120 rad for benign samples and from 10 to 160 rad for malignant samples. Moreover, the average of unwrapped phase distribution was 63.79 and 85.69 rad in benign and malignant samples, respectively. This results indicated that the proposed laser-based technique, Mach-Zehnder interferometer method, can be used as an auxiliary procedure along with histological techniques to identify benign and malignant brain tumors from each other. It is suggested that the unwrapped phase distribution of tissues be considered as an optical property for differentiating various brain tumors.

Universal in vivo Textural Model for Human Skin based on Optical Coherence Tomograms

Currently, diagnosis of skin diseases is based primarily on the visual pattern recognition skills and expertise of the physician observing the lesion. Even though dermatologists are trained to recognize patterns of morphology, it is still a subjective visual assessment. Tools for automated pattern recognition can provide objective information to support clinical decision-making. Noninvasive skin imaging techniques provide complementary information to the clinician. In recent years, optical coherence tomography (OCT) has become a powerful skin imaging technique. According to specific functional needs, skin architecture varies across different parts of the body, as do the textural characteristics in OCT images. There is, therefore, a critical need to systematically analyze OCT images from different body sites, to identify their significant qualitative and quantitative differences. Sixty-three optical and textural features extracted from OCT images of healthy and diseased skin are analyzed and, in conjunction with decision-theoretic approaches, used to create computational models of the diseases. We demonstrate that these models provide objective information to the clinician to assist in the diagnosis of abnormalities of cutaneous microstructure, and hence, aid in the determination of treatment. Specifically, we demonstrate the performance of this methodology on differentiating basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) from healthy tissue.

Evaluating Radioprotective Effect of Hesperidin on Acute Radiation Damage in the Lung Tissue of Rats

BACKGROUND

Oxidative stress plays an important role in the pathogenesis and progression of γ-irradiation-induced cellular damage, Lung is a radiosensitive organ and its damage is a dose-limiting factor in radiotherapy. The administration of dietary antioxidants has been suggested to protect against the succeeding tissue damage. The present study aimed to evaluate the radioprotective efficacy of Hesperidin (HES) against γ-irradiation-induced tissue damage in the lung of male rats.

MATERIALS AND METHODS

Thirty two rats were divided into four groups. Rats in Group 1 received PBS and underwent sham irradiation. Rats in Group 2 received HES and underwent sham irradiation. Rats in Group 3 received PBS and underwent γ-irradiation. Rats in Group 4 received HES and underwent γ-irradiation. These rats were exposed to γ-radiation 18 Gy using a single fraction cobalt-60 unit, and were administered HES (100 mg/kg/d, b.w, orally) for 7 days prior to irradiation. Rats in each group were sacrificed 24 hours after radiotherapy (RT) for the determination of superoxide dismutase (SOD), glutathione (GSH), malondialdehyde (MDA) and histopathological evaluations.

RESULTS

Compared to group 1, the level of SOD and GSH significantly decreased and MDA level significantly increased in group 3 at 24 h following irradiation, (p=0.001, p<0.001, p=0.001), respectively. A statistically significant difference in all parameters was observed for rats in group 4 as compared to group 3 (p<0.05). Histopathological results 24 hours after RT showed that radiation has increased inflammation, lymphocyte, macrophage and neutrophil compared to group 1 ( p<0.0125). Oral administration of HES before RT significantly decreased macrophage and neutrophil when compared to group 3 (p<0.0125), but partly there was inflammation and lymphocyte that indicated there was no significant difference when compared to group 3 (p>0.0125).

CONCLUSION

Oral administration of HES was found to offer protection against γ-irradiation- induced pulmonary damage and oxidative stress in rats, probably by exerting a protective effect against inflammatory disorders via its free radical scavenging and membrane stabilizing ability.

[Significant association between nonsyndromic cleft lip with or without cleft palate and IRF6rs2235371 polymorphism in Iranian familiar population]

According to previous studies the IRF6rs2235371 polymorphism is a risk factor for NSCL/P in different populations. However our recent study revealed no correlation between IRF6rs642961 and NSCL/P in our population. In the present study we have investigated the relationship between IRF6rs2235371 and NSCL/P in same group to determine whether IRF6rs2235371 is a risk factor in our population as well. We analyzed the IRF6rs2235371 genotype in a subset of the Iranian population using the Polymerase Chain Reaction technique. The PCR products were digested with DpnII. Chi-square test was applied to analyze the obtained result. The patients were supplied by the Cleft Lip and Palate Clinic of the Isfahan University of Medical Science. A clinician ascertained the non-syndromic status of all patients and that no clefting drugs, ethanol or smoking were abused during pregnancy. The control group was selected from unaffected subjects with no history of NSCL/P in their families. 107 patients from 107 Iranian unrelated families and 100 controls were screened. There was a significant association between the IRF6rs2235371 genotype sand an increased NSCL/P risk. Our data indicates that the IRF6rs2235371 variation can increase the risk of NSCL/P in the Iranian population. This result is in contrast with the results of our recent study on the correlation between the IRF6rs642961 polymorphism and NSCL/P in the same group.

The effect of creatine supplementation on muscle fatigue and physiological indices following intermittent swimming bouts

We evaluated the effect of Creatine (Cr) supplementation on muscle fatigue and physiological indices after intermittent swimming bouts in trained swimmers. Sixteen healthy non-elite swimmers (19±4 years, 75±12 kg) were randomly assigned into two groups of either Cr supplementation or placebo and performed six repeated sprints swimming bouts of 50-m departing every 120 seconds. The Cr group was supplemented 4 times a day for 6 days. Blood lactate, Creatine Kinase (CK), creatinine, heart rate, best repeated sprint (RSb) and mean repeated sprint (RSm) times, and percentage of speed decrement (%Dec) were measured at the various phases of swimming bouts. Repeated measure ANOVA and independent t-student tests showed CK and blood lactate concentration increased gradually after the third and sixth swimming bouts. % Dec in Cr group was significantly lower after 3rd swimming bout, also heart rate in Cr group was associated with lower increase in HR mean (P<0.05) compared to placebo. These results suggest that Cr supplementation may improve swimming performance and reduce increased blood lactate levels following intermittent sprint swimming bouts. In conclusion Cr supplementation in trained swimmers may improve anaerobic performance and heart rate variations independent of the effect of intensive sprint swimming bouts.

The interactive effects of exercise type and environment temperature on HSP72 in active females

The purpose of this study was to determine the effects of endurance and/ or weight training in normal and/ or moderate heat environments on heat shock protein (HSP72) response in active young females. Forty-five healthy active female (21±3 year, 61.4±10.2 kg, VO2max 37.2±5.3 mL/kg/min) were randomLy assigned into five groups; endurance training+normal heat (ET+NH), endurance training+moderate heat (ET+MH), weight training+normal heat (WT+NH), weight training+moderate heat (WT+MH) and heat group (HG). The training protocols include treadmill running to exhaustion on a treadmill at %65-75 V02max in the ET+NH and ET+MH groups and 4 sets of eccentric actions of elbow flexors (two sets with %50 and two sets with %60 1RM) in the WT+NH and WT+MH groups. Fasting blood samples were collected from non-dominant arm at three phases; consist of rest (baseline) and mid-test and 30 min after the training protocols. Endurance training and/ or weight training, in particularly, in the moderate heat ambient resulted in an increase in the HSP72, creatin kinase (CK) and cortisol values, as compared to ET+NH and WT+NH groups and the mid-test and baseline phases. However, the HSP72 and CK concentrations significantly increased in WT+ NH group, as compared to ET+NH group. Furthermore, WT+MH group showed significant increase in HSP72 levels, as compared to ET+ NH group. These findings suggest the muscle damage related to factors such as; exercise type, environment temperature and increase in exercise-induced body temperature.

Endurance exercise training and diferuloyl methane supplement: changes in neurotrophic factor and oxidative stress induced by lead in rat brain

Lead is a highly neurotoxic agent that particularly affects the developing central nervous system. In the current study we investigated the neuroprotective effects of exercise training and/or diferuloyl methane (DM) supplement, which is known as curcumin, on lead acetate-induced neurotoxicity in the rat hippocampus. Sixty rats were randomly divided into six groups: 1) lead acetate, 2) DM supplement, 3) endurance training, 4) training+ DM supplement, 5) sham and 6) base. The rats in the training groups performed treadmill running consisting of 15 to 22 m · min(-1) for 25 to 64 min, 5 times a week for 8 weeks. All groups except sham received lead acetate (20 mg · kg(-1)), whereas the sham group received DM solvent. In addition, the DM and training + DM groups received DM solution (30 mg · kg(-1)) intraperitoneally. Chronic administration of lead acetate resulted in a significant increase in the malondialdehyde (MDA) in plasma, but not in the hippocampus. In addition, it led to significantly decreased brain-derived neurotrophic factor (BDNF) in the hippocampus and total antioxidant capacity (TAC) levels, as compared to the sham group. Treadmill running, DM supplementation, or both resulted in a significant decrease in MDA levels and significantly increased BDNF and TAC levels, as compared to the lead acetate group. These results provide a rationale for an inhibitory role of DM supplement and regular exercise in the attenuation of lead-induced neurotoxicity.

An epidemiological and clinical study on scorpionism in hospitalized children in khuzestan, iran

BACKGROUND

Scorpion sting is a public health problem in Khuzestan, South-West Iran. The aims of the current study were to monitor the hospitalized children, due to scorpion sting, and releasing more clinical and epidemiologic data related to scorpionism in this Province.

METHODS

In this retrospective study, the data of scorpion sting victims, among the hospitalized children in Abuzar Children Hospital of Ahvaz Jundi Shapur University of Medical Sciences, was analyzed from the points of epidemiological and clinical aspects in 2006.

RESULTS

The scorpion species of 18 files out of 57 were recognized accurately, using Farzanpay's keyof Iranian scorpions, as Androctonus crassicauda (Scorpionida: Buthidae) and Hemiscorpious lepturus (Scorpionida: Hemiscorpiidae).

CONCLUSION

The most scorpionism emergencies among the children in the Khuzestan should be paid to those species. However, H. lepturus sting emergencies are in the top of attention among the children.

Seroprevalence of hepatitis B in Nahavand, Islamic Republic of Iran

We determined the seroprevalence of hepatitis B in Nahavand in a sample of 1824 subjects > 5 years in 2002. Face-to-face interviews were conducted and blood samples taken. The association between risk factor and hepatitis B was assessed using logistic regression. The prevalence of HBsAg positive cases was 2.3%, and HBcAb and HBsAb were isolated in 7.8% and 11.6% of the participants respectively; 11.9% were positive for both HBcAb and HBsAb. History of surgery and imprisonment were the major risk factors for infection with odds ratios of 2.14 (95% CI: 1.22-3.05) and 3.57 (95% CI: 1.68-5.4) respectively.

Cytologic findings of medulloblastoma in crush smears

OBJECTIVE

To describe the cytologic findings of medulloblastomas on intraoperative crush preparation smears and compare them with the findings of other central nervous system tumors.

STUDY DESIGN

The intraoperative crush preparation smears of 19 cases of medulloblastoma were studied (6 undifferentiated type and 13 well-differentiated type). The findings were compared with those of a control group consisting of 31 astrocytomas, 22 ependymomas, 18 oligodendrogliomas, 27 meningiomas, 17 schwannomas, 17 pituitary adenomas, 3 lymphomas, 5 hemangioblastomas, 5 chordomas and 11 metastatic tumors.

RESULTS

Medulloblastomas revealed clusters and isolated small, round malignant cells with hyperchromatic nuclei and indistinct cytoplasm. Typical Homer-Wright rosettes were seen in the well-differentiated type, but they were poorly formed in the undifferentiated type. Tumor cannibalism, target inclusions, cytoplasmic vacuoles and prominent multiple nucleoli were noticed frequently in the undifferentiated type. The control group (metastatic tumors and high grade astrocytomas) rarely showed tumor cannibalism or multiple nucleoli.

CONCLUSION

Smears of the undifferentiated type of medulloblastomas frequently revealed tumor cannibalism, cytoplasmic vacuoles, target inclusions and prominent multiple nucleoli. These findings have been rarely reported. The prognosis of the undifferentiated type of medulloblastoma was poor.

Ménière disease caused by an anomalous vein of the vestibular aqueduct

The precise cause of Ménière disease remains unclear. Multiple causes have been proposed with most experimental evidence pointing to impaired fluid resorption by the endolymphatic duct and sac as the final common pathway in development of hydrops. We report a unique case of Ménière disease secondary to compression of the endolymphatic duct and sac by an anomalous vein of the vestibular aqueduct. The resulting mechanical obstruction led to excessive accumulation of endolymph with clinical signs of Ménière disease. We review the literature and discuss proposed pathogenesis of disease. The finding of this anomalous vein provides further evidence that anatomical obstruction of the endolymphatic duct and sac may lead to Ménière-like symptoms. This unique example of an anatomical variant offers additional insight into the pathophysiology of endolymphatic hydrops.

Effect of diabetes on the healing process of synthetic grafts implanted in dogs: a preliminary study

The deleterious effects of diabetes in peripheral vascular surgery and blood access are generally considered to be associated with the greater failure rate of vascular prostheses. To understand better these phenomena, we investigated the healing characteristics of an ePTFE graft in dogs. The animals were made diabetic after total pancreatectomy and the grafts implanted for scheduled periods of 24 h, 48 h, 1 wk and 1 month, either as thoracic (series I) or abdominal (series II) aortic substitutes. The same implantations were performed in non-diabetic dogs used as controls. The luminal surfaces of grafts implanted in diabetic dogs, either as thoracic or abdominal bypasses, proved to be more invaded by platelets as a result of the blood's increased platelet aggregation properties. The presence of endothelial-like cells spreading over the pannus after 1 month implantation was strikingly different in diabetic versus non-diabetic dogs. In the first group, the cells were fragile and less abundant, whereas in the second they were more adhesive, elongated and orientated in the direction of the blood flow.