Refine
Year of publication
Document Type
- Article (16)
- Doctoral Thesis (1)
Has Fulltext
- yes (17)
Is part of the Bibliography
- no (17)
Keywords
- Machine learning (2)
- Prostate cancer (2)
- ARDS (1)
- Antibiotics (1)
- Artificial intelligence (1)
- Bacterial pathogens (1)
- COVID-19 (1)
- CTLA-4 (1)
- Critical care (1)
- Diagnostic markers (1)
Institute
- Medizin (17) (remove)
Background: Intensive Care Resources are heavily utilized during the COVID-19 pandemic. However, risk stratification and prediction of SARS-CoV-2 patient clinical outcomes upon ICU admission remain inadequate. This study aimed to develop a machine learning model, based on retrospective & prospective clinical data, to stratify patient risk and predict ICU survival and outcomes. Methods: A Germany-wide electronic registry was established to pseudonymously collect admission, therapeutic and discharge information of SARS-CoV-2 ICU patients retrospectively and prospectively. Machine learning approaches were evaluated for the accuracy and interpretability of predictions. The Explainable Boosting Machine approach was selected as the most suitable method. Individual, non-linear shape functions for predictive parameters and parameter interactions are reported. Results: 1039 patients were included in the Explainable Boosting Machine model, 596 patients retrospectively collected, and 443 patients prospectively collected. The model for prediction of general ICU outcome was shown to be more reliable to predict “survival”. Age, inflammatory and thrombotic activity, and severity of ARDS at ICU admission were shown to be predictive of ICU survival. Patients’ age, pulmonary dysfunction and transfer from an external institution were predictors for ECMO therapy. The interaction of patient age with D-dimer levels on admission and creatinine levels with SOFA score without GCS were predictors for renal replacement therapy. Conclusions: Using Explainable Boosting Machine analysis, we confirmed and weighed previously reported and identified novel predictors for outcome in critically ill COVID-19 patients. Using this strategy, predictive modeling of COVID-19 ICU patient outcomes can be performed overcoming the limitations of linear regression models. Trial registration “ClinicalTrials” (clinicaltrials.gov) under NCT04455451.
Staphylococcus aureus proteins Sbi and Efb recruit human plasmin to degrade complement C3 and C3b
(2012)
Upon host infection, the human pathogenic microbe Staphylococcus aureus (S. aureus) immediately faces innate immune reactions such as the activated complement system. Here, a novel innate immune evasion strategy of S. aureus is described. The staphylococcal proteins surface immunoglobulin-binding protein (Sbi) and extracellular fibrinogen-binding protein (Efb) bind C3/C3b simultaneously with plasminogen. Bound plasminogen is converted by bacterial activator staphylokinase or by host-specific urokinase-type plasminogen activator to plasmin, which in turn leads to degradation of complement C3 and C3b. Efb and to a lesser extend Sbi enhance plasmin cleavage of C3/C3b, an effect which is explained by a conformational change in C3/C3b induced by Sbi and Efb. Furthermore, bound plasmin also degrades C3a, which exerts anaphylatoxic and antimicrobial activities. Thus, S. aureus Sbi and Efb comprise platforms to recruit plasmin(ogen) together with C3 and its activation product C3b for efficient degradation of these complement components in the local microbial environment and to protect S. aureus from host innate immune reactions.
Background: The inclusion of immune checkpoint inhibitors (ICIs) in therapeutic algorithms has led to significant survival benefits in patients with various metastatic cancers. Concurrently, an increasing number of neurological immune related adverse events (IRAE) has been observed. In this retrospective analysis, we examine the ICI-induced incidence of cerebral pseudoprogression and propose a classification system.
Methods: We screened our hospital information system to identify patients with any in-house ICI treatment for any tumor disease during the years 2007-2019. All patients with cerebral MR imaging (cMRI) of sufficient diagnostic quality were included. cMRIs were retrospectively analyzed according to immunotherapy response assessment for neuro-oncology (iRANO) criteria.
Results: We identified 12 cases of cerebral pseudoprogression in 123 patients treated with ICIs and sufficient MRI. These patients were receiving ICI therapy for lung cancer (n=5), malignant melanoma (n=4), glioblastoma (n=1), hepatocellular carcinoma (n=1) or lymphoma (n=1) when cerebral pseudoprogression was detected. Median time from the start of ICI treatment to pseudoprogression was 5 months. All but one patient developed neurological symptoms. Three different patterns of cerebral pseudoprogression could be distinguished: new or increasing contrast-enhancing lesions, new or increasing T2 predominant lesions and cerebral vasculitis type pattern.
Conclusion: Cerebral pseudoprogression followed three distinct patterns and was detectable in 3.2% of all patients during ICI treatment and in 9.75% of the patients with sufficient brain imaging follow up. The fact that all but one of the affected patients developed neurological symptoms, which would be classified as progressive disease according to iRANO criteria, mandates vigilance in the diagnosis and treatment of ICI-induced cerebral lesions.
In der vorliegenden Arbeit wurde ein durch Zellschrumpfung aktivierter nichtselektiver Kationenkanal, der in der M1 Sammelrohrzellinie der Maus bereits früher beobachtet worden war [Volk et al., 1995], hinsichtlich seiner Selektivität, seines Inhibitorspektrums und seines Aktivierungsmechanismus mittels der Patch ClampTechnik näher charakterisiert, und es wurde geprüft, ob dieser oder ein ähnlicher Kanal auch von anderen Zellinien exprimiert wird, wobei außer epithelialen Zellinien (HT 29 , BSC1) auch solche aus Muskel (A10) und Ner vengewebe (Neuro2a) untersucht wurden. Zellschrumpfung in symmetrischer NaClLösung mit Zusatz von 100 mM Saccharose im Bad stimulierte einen von Kationen getragenen Einwärtsstrom. Während der initialen Phase der Aktivierung durch extrazelluläre Hyperosmolarität konnten Schaltvorgänge von Einzelka nälen beobachtet werden. Die Einzelkanalleitfähigkeit in den verschiedenen Zellinien betrug 15 bis 27 pS. Der Strom stieg nach wenigen Minuten um das 30 bis 60fache an. Austausch des extrazellulären Natriums durch das schlecht permeierende NMDG führte zum kompletten Sistieren des Einwärtsstroms. Die Aktivierung des nichtselektiven Kationenstroms war unab hängig von der absoluten Osmolarität der Badlösung. Sie wurde ausschließlich durch osmo tisch bedingte Volumenabnahme (Schrumpfung) der Zellen hervorgerufen, wobei die Zell schrumpfung der Aktivierung des Stroms geringgradig vorauseilte. Die Aktivierung durch Zellschrumpfung war reversibel und konnte mit dem gleichen Zeitverlauf und mit der glei chen Leitfähigkeitszunahme auch zweimal hintereinander ausgelöst werden. Die stimulierte Kationenleitfähigkeit war hochselektiv für Kationen über Anionen, aber nichtselektiv für die monovalenten Kationen NH 4 , Natrium, Kalium und Lithium (Leitfähig keitssequenz NH 4 > Na > K > Li ). Für divalente Kationen wie Kalzium und Barium war keine Leitfähigkeit meßbar. Flufenaminsäure (100 µM) hemmte den stimulierten Einwärts strom um mehr als 80 %, Diphenylamin2Carboxylsäure (DPC) und sein Derivat Dichlor DPC (DCDPC) hatten einen etwas geringeren Hemmeffekt. Noch schwächer hemmte LOE 908. Amilorid, Quecksilber und Bumetanid hatten praktisch keine Hemmwirkung auf den stimulierten Einwärtsstrom. Maitotoxin, ein Aktivator mancher nichtselektiver Kationenka näle, hatte unter isoomotischen Bedingungen keine stimulatorische Wirkung auf den Strom. Im Gegensatz zum Ca 2 aktivierten nichtselektiven Kationenkanal, der in exzidierten Membranflecken beobachtet wird, war die Aktivierung des schrumpfungsinduzierten nicht selektiven Kationenkanals unabhängig von der zytoplasmatischen Ca 2 Konzentration. Das Ausmaß der Stimulation war allerdings vermindert, wenn die Pipettenlösung hohe Konzen trationen von Kalziumchelatoren (10 mM EGTA oder 10 mM BAPTA) enthielt. Dagegen hemmte Magnesiumentzug die Antwort auf Zellschrumpfung fast vollständig. ATP in Kon zentrationen bis 10 mM in der Pipettenlösung konnte die Aktivierung des Einwärtsstroms nicht verhindern, nur reduzieren. Dagegen unterdrückte ATPVerarmung der Zellen (durch Inkubation mit Rotenon und 2Deoxyglukose) die Antwort auf Zellschrumpfung völlig. Auch durch Zusatz von 1 mM ATP zur Pipettenlösung konnte die Antwort in ATPverarmten Zel len nicht wiedergewonnen werden, obwohl die Zellen nach Auswaschen der Stoffwechsel gifte wieder normal reagierten. Die Proteinkinasehemmstoffe Staurosporin und Calphostin C (beide in der Konzentration von 1 µM) konnten die Antwort unterdrücken. Einige auf das Zytoskelett wirksame Substanzen (Cytochalasin D und Taxol) hatten ebenfalls hemmende Wirkungen. Die Befunde der vorliegenden Arbeit zeigen, daß der schrumpfungsaktivierte nichtselekti ve Kationenkanal offenbar ubiquitär exprimiert wird. Der Aktivierungsmechanismus ist kom plex und abhängig von intrazellulärem ATP und Magnesium. Vermutlich sind Proteinkinasen und Zytoskelettelemente an der Kanalaktivierung beteiligt. Der Kanal dürfte im Rahmen der Zellvolumenregulation von Bedeutung sein und spielt möglicherweise eine Rolle für Zellpro liferation und Apoptose.
Hepatitis C virus (HCV) substantially affects lipid metabolism, and remodeling of sphingolipids appears to be essential for HCV persistence in vitro. The aim of the current study is the evaluation of serum sphingolipid variations during acute HCV infection. We enrolled prospectively 60 consecutive patients with acute HCV infection, most of them already infected with human immunodeficiency virus (HIV), and serum was collected at the time of diagnosis and longitudinally over a six-month period until initiation of antiviral therapy or confirmed spontaneous clearance. Quantification of serum sphingolipids was performed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Spontaneous clearance was observed in 11 out of 60 patients (18.3%), a sustained viral response (SVR) in 43 out of 45 patients (95.5%) receiving an antiviral treatment after follow-up, whereas persistence of HCV occurred in six out of 60 patients (10%). C24-ceramide (C24-Cer)-levels increased at follow-up in patients with spontaneous HCV eradication (p < 0.01), as compared to baseline. Sphingosine and sphinganine values were significantly upregulated in patients unable to clear HCV over time compared to patients with spontaneous clearance of HCV infection on follow-up (p = 0.013 and 0.006, respectively). In summary, the persistence of HCV after acute infection induces a downregulation of C24Cer and a simultaneous elevation of serum sphingosine and sphinganine concentrations.
Background: Since there is no standardized and effective treatment for advanced uveal melanoma (UM), the prognosis is dismal once metastases develop. Due to the availability of immune checkpoint blockade (ICB) in the real-world setting, the prognosis of metastatic UM has improved. However, it is unclear how the presence of hepatic and extrahepatic metastasis impacts the response and survival after ICB. Methods: A total of 178 patients with metastatic UM treated with ICB were included in this analysis. Patients were recruited from German skin cancer centers and the German national skin cancer registry (ADOReg). To investigate the impact of hepatic metastasis, two cohorts were compared: patients with liver metastasis only (cohort A, n = 55) versus those with both liver and extra-hepatic metastasis (cohort B, n = 123). Data were analyzed in both cohorts for response to treatment, progression-free survival (PFS), and overall survival (OS). The survival and progression probabilities were calculated with the Kaplan–Meier method. Log-rank tests, χ2 tests, and t-tests were performed to detect significant differences between both cohorts. Results: The median OS of the overall population was 16 months (95% CI 13.4–23.7) and the median PFS, 2.8 months (95% CI 2.5–3.0). The median OS was longer in cohort B than in cohort A (18.2 vs. 6.1 months; p = 0.071). The best objective response rate to dual ICB was 13.8% and to anti-PD-1 monotherapy 8.9% in the entire population. Patients with liver metastases only had a lower response to dual ICB, yet without significance (cohort A 8.7% vs. cohort B 16.7%; p = 0.45). Adverse events (AE) occurred in 41.6%. Severe AE were observed in 26.3% and evenly distributed between both cohorts. Conclusion: The survival of this large cohort of patients with advanced UM was more favorable than reported in previous benchmark studies. Patients with both hepatic and extrahepatic metastasis showed more favorable survival and higher response to dual ICB than those with hepatic metastasis only.
Objectives: To analyze the performance of radiological assessment categories and quantitative computational analysis of apparent diffusion coefficient (ADC) maps using variant machine learning algorithms to differentiate clinically significant versus insignificant prostate cancer (PCa). Methods: Retrospectively, 73 patients were included in the study. The patients (mean age, 66.3 ± 7.6 years) were examined with multiparametric MRI (mpMRI) prior to radical prostatectomy (n = 33) or targeted biopsy (n = 40). The index lesion was annotated in MRI ADC and the equivalent histologic slides according to the highest Gleason Grade Group (GrG). Volumes of interest (VOIs) were determined for each lesion and normal-appearing peripheral zone. VOIs were processed by radiomic analysis. For the classification of lesions according to their clinical significance (GrG ≥ 3), principal component (PC) analysis, univariate analysis (UA) with consecutive support vector machines, neural networks, and random forest analysis were performed. Results: PC analysis discriminated between benign and malignant prostate tissue. PC evaluation yielded no stratification of PCa lesions according to their clinical significance, but UA revealed differences in clinical assessment categories and radiomic features. We trained three classification models with fifteen feature subsets. We identified a subset of shape features which improved the diagnostic accuracy of the clinical assessment categories (maximum increase in diagnostic accuracy ΔAUC = + 0.05, p < 0.001) while also identifying combinations of features and models which reduced overall accuracy. Conclusions: The impact of radiomic features to differentiate PCa lesions according to their clinical significance remains controversial. It depends on feature selection and the employed machine learning algorithms. It can result in improvement or reduction of diagnostic performance.
Introduction: It has been proposed that individual genetic variation contributes to the course of severe infections and sepsis. Recent studies of single nucleotide polymorphisms (SNPs) within the endotoxin receptor and its signaling system showed an association with the risk of disease development. This study aims to examine the response associated with genetic variations of TLR4, the receptor for bacterial LPS, and a central intracellular signal transducer (TIRAP/Mal) on cytokine release and for susceptibility and course of severe hospital acquired infections in distinct patient populations. Methods: Three intensive care units in tertiary care university hospitals in Greece and Germany participated. 375 and 415 postoperative patients and 159 patients with ventilator associated pneumonia (VAP) were included. TLR4 and TIRAP/Mal polymorphisms in 375 general surgical patients were associated with risk of infection, clinical course and outcome. In two prospective studies, 415 patients following cardiac surgery and 159 patients with newly diagnosed VAP predominantly caused by Gram-negative bacteria were studied for cytokine levels in-vivo and after ex-vivo monocyte stimulation and clinical course. Results: Patients simultaneously carrying polymorphisms in TIRAP/Mal and TLR4 and patients homozygous for the TIRAP/Mal SNP had a significantly higher risk of severe infections after surgery (odds ratio (OR) 5.5; confidence interval (CI): 1.34 - 22.64; P = 0.02 and OR: 7.3; CI: 1.89 - 28.50; P < 0.01 respectively). Additionally we found significantly lower circulating cytokine levels in double-mutant individuals with ventilator associated pneumonia and reduced cytokine production in an ex-vivo monocyte stimulation assay, but this difference was not apparent in TIRAP/Mal-homozygous patients. In cardiac surgery patients without infection, the cytokine release profiles were not changed when comparing different genotypes. Conclusions: Carriers of mutations in sequential components of the TLR signaling system may have an increased risk for severe infections. Patients with this genotype showed a decrease in cytokine release when infected which was not apparent in patients with sterile inflammation following cardiac surgery.
Alternative polyadenylation (APA) is a widespread mechanism that contributes to the sophisticated dynamics of gene regulation. Approximately 50% of all protein-coding human genes harbor multiple polyadenylation (PA) sites; their selective and combinatorial use gives rise to transcript variants with differing length of their 3' untranslated region (3'UTR). Shortened variants escape UTR-mediated regulation by microRNAs (miRNAs), especially in cancer, where global 3'UTR shortening accelerates disease progression, dedifferentiation and proliferation. Here we present APADB, a database of vertebrate PA sites determined by 3' end sequencing, using massive analysis of complementary DNA ends. APADB provides (A)PA sites for coding and non-coding transcripts of human, mouse and chicken genes. For human and mouse, several tissue types, including different cancer specimens, are available. APADB records the loss of predicted miRNA binding sites and visualizes next-generation sequencing reads that support each PA site in a genome browser. The database tables can either be browsed according to organism and tissue or alternatively searched for a gene of interest. APADB is the largest database of APA in human, chicken and mouse. The stored information provides experimental evidence for thousands of PA sites and APA events. APADB combines 3' end sequencing data with prediction algorithms of miRNA binding sites, allowing to further improve prediction algorithms. Current databases lack correct information about 3'UTR lengths, especially for chicken, and APADB provides necessary information to close this gap. Database URL: http://tools.genxpro.net/apadb/
Empiric antibiotics are often used in combination with mechanical debridement to treat patients suffering from periodontitis and to eliminate disease-associated pathogens. Until now, only a few next generation sequencing 16S rDNA amplicon based publications with rather small sample sizes studied the effect of those interventions on the subgingival microbiome. Therefore, we studied subgingival samples of 89 patients with chronic periodontitis (solely non-smokers) before and two months after therapy. Forty-seven patients received mechanical periodontal therapy only, whereas 42 patients additionally received oral administered amoxicillin plus metronidazole (500 and 400 mg, respectively; 3x/day for 7 days). Samples were sequenced with Illumina MiSeq 300 base pairs paired end technology (V3 and V4 hypervariable regions of the 16S rDNA). Inter-group differences before and after therapy of clinical variables (percentage of sites with pocket depth ≥ 5mm, percentage of sites with bleeding on probing) and microbiome variables (diversity, richness, evenness, and dissimilarity) were calculated, a principal coordinate analysis (PCoA) was conducted, and differential abundance of agglomerated ribosomal sequence variants (aRSVs) classified on genus level was calculated using a negative binomial regression model. We found statistically noticeable decreased richness, and increased dissimilarity in the antibiotic, but not in the placebo group after therapy. The PCoA revealed a clear compositional separation of microbiomes after therapy in the antibiotic group, which could not be seen in the group receiving mechanical therapy only. This difference was even more pronounced on aRSV level. Here, adjunctive antibiotics were able to induce a microbiome shift by statistically noticeably reducing aRSVs belonging to genera containing disease-associated species, e.g., Porphyromonas, Tannerella, Treponema, and Aggregatibacter, and by noticeably increasing genera containing health-associated species. Mechanical therapy alone did not statistically noticeably affect any disease-associated taxa. Despite the difference in microbiome modulation both therapies improved the tested clinical parameters after two months. These results cast doubt on the relevance of the elimination and/or reduction of disease-associated taxa as a main goal of periodontal therapy.
Aim: We investigated the long-term impact of adjunctive systemic antibiotics on periodontal disease progression. Periodontal therapy is frequently supplemented by systemic antibiotics, although its impact on the course of disease is still unclear.
Material & Methods: This prospective, randomized, double-blind, placebo-controlled multi-centre trial comprising patients suffering from moderate to severe periodontitis evaluated the impact of rational adjunctive use of systemic amoxicillin 500 mg plus metronidazole 400 mg (3x/day, 7 days) on attachment loss. The primary outcome was the percentage of sites showing further attachment loss (PSAL) ≥1.3 mm after the 27.5 months observation period. Standardized therapy comprised mechanical debridement in conjunction with antibiotics or placebo administration, and maintenance therapy at 3 months intervals.
Results: From 506 participating patients, 406 were included in the intention to treat analysis. Median PSAL observed in placebo group was 7.8% compared to 5.3% in antibiotics group (Q25 4.7%/Q75 14.1%; Q25 3.1%/Q75 9.9%; p < 0.001 respectively).
Conclusions: Both treatments were effective in preventing disease progression. Compared to placebo, the prescription of empiric adjunctive systemic antibiotics showed a small absolute, although statistically significant, additional reduction in further attachment loss. Therapists should consider the patient's overall risk for periodontal disease when deciding for or against adjunctive antibiotics prescription.
Brain metastases are a common finding upon initial diagnosis of otherwise locally limited non-small cell lung cancer. We present a retrospective case series describing three cases of patients with symptomatic, synchronous brain metastases and resectable lung tumors. The patients received local ablative treatment of the brain metastases followed by neoadjuvant immunochemotherapy with pemetrexed, cisplatin, and pembrolizumab. Afterwards, resection of the pulmonary lesion with curative intent was performed. One patient showed progressive disease 12 months after initial diagnosis, and passed away 31 months after initial diagnosis. Two of the patients are still alive and maintain a good quality of life with a progression-free survival and overall survival of 28 and 35 months, respectively, illustrating the potential of novel combinatorial treatment approaches.
The demand to develop convergent technology platforms, such as bio-functionalized medical devices, is rapidly increasing. However, the loss of biological function of the effector molecules during sterilization represents a significant and general problem. Therefore, we have developed and characterized a nano-coating (NC) formulation capable of maintaining the functionality of proteins on biological-device combination products. As a proof of concept, the NC preserved the structural and functional integrity of an otherwise highly fragile antibody immobilized on polyurethane during deleterious sterilizing irradiation (≥ 25 kGy). The NC procedure enables straight-forward terminal sterilization of bio-functionalized materials while preserving optimal conditioning of the bioactive surface.
Our purpose was to analyze the robustness and reproducibility of magnetic resonance imaging (MRI) radiomic features. We constructed a multi-object fruit phantom to perform MRI acquisition as scan-rescan using a 3 Tesla MRI scanner. We applied T2-weighted (T2w) half-Fourier acquisition single-shot turbo spin-echo (HASTE), T2w turbo spin-echo (TSE), T2w fluid-attenuated inversion recovery (FLAIR), T2 map and T1-weighted (T1w) TSE. Images were resampled to isotropic voxels. Fruits were segmented. The workflow was repeated by a second reader and the first reader after a pause of one month. We applied PyRadiomics to extract 107 radiomic features per fruit and sequence from seven feature classes. We calculated concordance correlation coefficients (CCC) and dynamic range (DR) to obtain measurements of feature robustness. Intraclass correlation coefficient (ICC) was calculated to assess intra- and inter-observer reproducibility. We calculated Gini scores to test the pairwise discriminative power specific for the features and MRI sequences. We depict Bland Altmann plots of features with top discriminative power (Mann–Whitney U test). Shape features were the most robust feature class. T2 map was the most robust imaging technique (robust features (rf), n = 84). HASTE sequence led to the least amount of rf (n = 20). Intra-observer ICC was excellent (≥ 0.75) for nearly all features (max–min; 99.1–97.2%). Deterioration of ICC values was seen in the inter-observer analyses (max–min; 88.7–81.1%). Complete robustness across all sequences was found for 8 features. Shape features and T2 map yielded the highest pairwise discriminative performance. Radiomics validity depends on the MRI sequence and feature class. T2 map seems to be the most promising imaging technique with the highest feature robustness, high intra-/inter-observer reproducibility and most promising discriminative power.
Acne‐like skin reactions frequently occur in patients undergoing treatment with drugs inhibiting the epidermal growth factor receptor. Recently, the effects of vitamin K1 containing cream (Reconval K1) as prophylactic skin treatment in addition to doxycycline were explored in a double‐blind randomized phase II trial (EVITA) in patients with metastatic colorectal cancer receiving cetuximab. EVITA demonstrated a trend towards less severe skin rash in Reconval K1‐treated patients using the tripartite WoMo skin reaction grading score as a thorough tool for quantification of drug related skin reactions. This gender‐specific analysis of the EVITA trial evaluated the application of the WoMo score for assessment of epidermal growth factor receptor (EGFR)‐related skin toxicities according to treatment arm and gender. To show the robustness of results parametric and non‐parametric statistical analyses were conducted. All three parts of the WoMo score independently demonstrated the superiority of the treatment arm (Reconval K1) regarding a significant reduction in acneiform skin reactions in women. Men did not benefit from Reconval K1 cream at any time point in none of the WoMo score analyses. The treatment effect in women was confirmed by the use of skin rash categories based on the final WoMo overall score and mixed effect longitudinal multiple linear regression analysis. The WoMo score represents a sensitive tool for studies exploiting treatments against EGFR mediated acne‐like skin rash. Part C of the WoMo score seems to be sufficient for quantification of drug related skin toxicities in further studies. Standard WoMo skin reaction score values for future studies are provided.
Background: Prostate cancer is a major health concern in aging men. Paralleling an aging society, prostate cancer prevalence increases emphasizing the need for efcient diagnostic algorithms.
Methods: Retrospectively, 106 prostate tissue samples from 48 patients (mean age,
66 ± 6.6 years) were included in the study. Patients sufered from prostate cancer (n = 38) or benign prostatic hyperplasia (n = 10) and were treated with radical prostatectomy or Holmium laser enucleation of the prostate, respectively. We constructed tissue microarrays (TMAs) comprising representative malignant (n = 38) and benign (n = 68) tissue cores. TMAs were processed to histological slides, stained, digitized and assessed for the applicability of machine learning strategies and open–source tools in diagnosis of prostate cancer. We applied the software QuPath to extract features for shape, stain intensity, and texture of TMA cores for three stainings, H&E, ERG, and PIN-4. Three machine learning algorithms, neural network (NN), support vector machines (SVM), and random forest (RF), were trained and cross-validated with 100 Monte Carlo random splits into 70% training set and 30% test set. We determined AUC values for single color channels, with and without optimization of hyperparameters by exhaustive grid search. We applied recursive feature elimination to feature sets of multiple color transforms.
Results: Mean AUC was above 0.80. PIN-4 stainings yielded higher AUC than H&E and
ERG. For PIN-4 with the color transform saturation, NN, RF, and SVM revealed AUC of 0.93 ± 0.04, 0.91 ± 0.06, and 0.92 ± 0.05, respectively. Optimization of hyperparameters improved the AUC only slightly by 0.01. For H&E, feature selection resulted in no increase of AUC but to an increase of 0.02–0.06 for ERG and PIN-4.
Conclusions: Automated pipelines may be able to discriminate with high accuracy between malignant and benign tissue. We found PIN-4 staining best suited for classifcation. Further bioinformatic analysis of larger data sets would be crucial to evaluate the reliability of automated classifcation methods for clinical practice and to evaluate potential discrimination of aggressiveness of cancer to pave the way to automatic precision medicine.