Benchmarking weakly-supervised deep learning pipelines for whole slide classification in computational pathology

Artificial intelligence (AI) can extract visual information from histopathological slides and yield biological insight and clinical biomarkers. Whole slide images are cut into thousands of tiles and classification problems are often weakly-supervised: the ground truth is only known for the slide, not for every single tile. In classical weakly-supervised analysis pipelines, all tiles inherit the slide label while in multiple-instance learning (MIL), only bags of tiles inherit the label. However, it is still unclear how these widely used but markedly different approaches perform relative to each other.We implemented and systematically compared six methods in six clinically relevant end-to-end prediction tasks using data from N=2980 patients for training with rigorous external validation. We tested three classical weakly-supervised approaches with convolutional neural networks and vision transformers (ViT) and three MIL-based approaches with and without an additional attention module. Our results empirically demonstrate that histological tumor subtyping of renal cell carcinoma is an easy task in which all approaches achieve an area under the receiver operating curve (AUROC) of above 0.9. In contrast, we report significant performance differences for clinically relevant tasks of mutation prediction in colorectal, gastric, and bladder cancer. In these mutation prediction tasks, classical weakly-supervised workflows outperformed MIL-based weakly-supervised methods for mutation prediction, which is surprising given their simplicity. This shows that new end-to-end image analysis pipelines in computational pathology should be compared to classical weakly-supervised methods. Also, these findings motivate the development of new methods which combine the elegant assumptions of MIL with the empirically observed higher performance of classical weakly-supervised approaches. We make all source codes publicly available at https://github.com/KatherLab/HIA, allowing easy application of all methods to any similar task.     

Nebulized colistin as the adjunctive treatment for ventilator-associated pneumonia: A systematic review and meta-analysis

PurposeNebulized colistin (NC) is a potential therapy for ventilator-associated pneumonia (VAP); however, the clinical efficacy and safety of NC remain unclear. This study investigated whether NC is an effective therapy for patients with VAP.Materials and methodsWe performed a search in Web of Science, PubMed, Embase, and the Cochrane Library to retrieve randomized controlled trials (RCTs) and observational studies published at any time until February 6, 2023. The primary outcome was clinical response. Secondary outcomes included microbiological eradication, overall mortality, length of mechanical ventilation (MV), length of intensive care unit stay (ICU-LOS), nephrotoxicity, neurotoxicity, and bronchospasm.ResultsSeven observational studies and three RCTs were included. Despite exhibiting a higher microbiological eradication rate (OR,2.21; 95%CI, 1.25–3.92) and the same nephrotoxicity risk (OR,0.86; 95%CI, 0.60–1.23), NC was not significantly different in clinical response (OR,1.39; 95%CI, 0.87–2.20), overall mortality (OR,0.74; 95%CI, 0.50–1.12), MV length (mean difference (MD),-2.5; 95%CI, −5.20–0.19), and the ICU-LOS (MD,-1.91; 95%CI, −6.66–2.84) than by the intravenous antibiotic. Besides, the risk of bronchospasm raised significantly (OR, 5.19; 95%CI, 1.05–25.52) among NC.ConclusionNC was associated with better microbiological outcomes but did not result in any remarkable changes in the prognosis of patients with VAP.     

The presentation,etiologies, pathophysiology,and treatment of pulmonary renal syndrome: A review of the literature

Pulmonary renal syndrome (PRS) is a constellation of different disorders that cause both rapidly progressive glomerulonephritis and diffuse alveolar hemorrhage. While antineutrophil cytoplasmic antibody associated vasculitis and anti-glomerular basement membrane disease are the predominant causes of PRS, numerous other mechanisms have been shown to cause this syndrome, including thrombotic microangiopathies, drug exposures, and infections, among others. This syndrome has high morbidity and mortality, and early diagnosis and treatment is imperative to improve outcomes. Treatment generally involves glucocorticoids and immunosuppressive agents, but treatment targeted to the underlying disorder can improve outcomes and mitigate side effects. Familiarity with the wide range of possible causes of PRS can aid the clinician in workup, diagnosis and early initiation of treatment. This review provides a summary of the clinical presentation, etiologies, pathophysiology, and treatment of PRS.     

Non-pharmacological interventions for spasticity in adults: An overview of systematic reviews

ObjectivesSpasticity causes significant long-term disability-burden, requiring comprehensive management. This review evaluates evidence from published systematic reviews of clinical trials for effectiveness of non-pharmacological interventions for improved spasticity outcomes.MethodsData sources: a literature search was conducted using medical and health science electronic (MEDLINE, EMBASE, CINAHL, PubMed, and the Cochrane Library) databases for published systematic reviews up to 15th June 2017. Data extraction and synthesis: two reviewers applied inclusion criteria to select potential systematic reviews, independently extracted data for methodological quality using Assessment of Multiple Systematic Reviews (AMSTAR). Quality of evidence was critically appraised with Grades of Recommendation, Assessment, Development and Evaluation (GRADE).ResultsOverall 18 systematic reviews were evaluated for evidence for a range of non-pharmacological interventions currently used in managing spasticity in various neurological conditions. There is “moderate” evidence for electro-neuromuscular stimulation and acupuncture as an adjunct therapy to conventional routine care (pharmacological and rehabilitation) in persons following stroke. “Low” quality evidence for rehabilitation programs targeting spasticity (such as induced movement therapy, stretching, dynamic elbow-splinting, occupational therapy) in stroke and other neurological conditions; extracorporeal shock-wave therapy in brain injury; transcranial direct current stimulation in stroke; transcranial magnetic stimulation and transcutaneous electrical nerve stimulation for other neurological conditions; physical activity programs and repetitive magnetic stimulation in persons with MS, vibration therapy for SCI and stretching for other neurological condition. For other interventions, evidence was inconclusive.ConclusionsDespite the available range of non-pharmacological interventions for spasticity, there is lack of high-quality evidence for many modalities. Further research is needed to judge the effect with appropriate study designs, timing and intensity of modalities, and associate costs of these interventions.     

Weight-based versus non-weight-based diltiazem dosing in the setting of atrial fibrillation with rapid ventricular response

PurposeThere is conflicting evidence to support the superiority of weight-based (WB) dosing of intravenous (IV) diltiazem over non-weight-based (NWB) dosing strategies in the management of atrial fibrillation (AFib) with rapid ventricular response (RVR).MethodsA retrospective review evaluated patients presenting to the emergency department (ED) in AFib with RVR and receiving IV diltiazem from 2015 to 2018. Those receiving a NWB dose were compared with those receiving a WB dose based on actual body weight (ABW). Secondary analyses evaluated safety profiles of the regimens and compared response in groups defined by ABW or ideal body weight (IBW).ResultsA total of 371 patients were included in the analysis. No significant difference was observed in achieving a therapeutic response (66.5% vs. 73.1%, p = 0.18) or adverse events between the groups. Patients receiving a WB dose were significantly more likely to have a HR < 100 bpm than those receiving a NWB dose (40.9% vs. 53.5%, p = 0.01). When groups were defined by IBW, WB dosing was associated with a significantly higher incidence of achieving a therapeutic response (62.7% vs. 74.3%, p = 0.02).ConclusionIn patients presenting with AF with RVR, there was no significant difference in achieving a therapeutic response between the two strategies. A WB dosing approach did result in a greater proportion of patients with a HR < 100 bpm. The utilization of IBW for WB dosing may result in an increased achievement of a therapeutic response.     

Effect of Omega-3 Dosage on Cardiovascular Outcomes: An Updated Meta-Analysis and Meta-Regression of Interventional Trials

ObjectivesTo quantify the effect of eicosapentaenoic (EPA) and docosahexaenoic (DHA) acids on cardiovascular disease (CVD) prevention and the effect of dosage.MethodsThis study is designed as a random effects meta-analysis and meta-regression of randomized control trials with EPA/DHA supplementation. This is an update and expanded analysis of a previously published meta-analysis which covers all randomized control trials with EPA/DHA interventions and cardiovascular outcomes published before August 2019. The outcomes included are myocardial infarction (MI), coronary heart disease (CHD) events, CVD events (a composite of MI, angina, stroke, heart failure, peripheral arterial disease, sudden death, and non-scheduled cardiovascular surgical interventions), CHD mortality and fatal MI. The strength of evidence was assessed using the Grading of Recommendations Assessment, Development, and Evaluation framework.ResultsA total of 40 studies with a combined 135,267 participants were included. Supplementation was associated with reduced risk of MI (relative risk [RR], 0.87; 95% CI, 0.80 to 0.96), high certainty number needed to treat (NNT) of 272; CHD events (RR, 0.90; 95% CI, 0.84 to 0.97), high certainty NNT of 192; fatal MI (RR, 0.65; 95% CI, 0.46 to 0.91]), moderate certainty NNT = 128; and CHD mortality (RR, 0.91; 95% CI, 0.85 to 0.98), low certainty NNT = 431, but not CVD events (RR, 0.95; 95% CI, 0.90 to 1.00). The effect is dose dependent for CVD events and MI.ConclusionCardiovascular disease remains the leading cause of death worldwide. Supplementation with EPA and DHA is an effective lifestyle strategy for CVD prevention, and the protective effect probably increases with dosage.     

Treatment outcome of continuation of intravenous amikacin for Mycobacterium abscessus pulmonary disease with a persistent culture positivity after the treatment initiation

IntroductionWhether prolonged intravenous amikacin treatment would lead to better treatment results in patients with Mycobacterium abscessus subspecies abscessus (M. abscessus) pulmonary disease (PD) is unknown. We investigated the efficacy of continued amikacin treatment for the microbiological outcome of M. abscessus PD patients with persistent culture positivity after treatment initiation.MethodsWe retrospectively evaluated 62 patients with M. abscessus PD who were treated with intravenous amikacin and beta-lactams along with a macrolide-based regimen at 3 tertiary referral centers in South Korea. The intravenous antibiotic treatment duration was determined by the attending physician.ResultsThe median treatment durations with amikacin and beta-lactam in the 62 patients were 25.1 and 8.2 weeks, respectively. The overall microbiological cure rate was 29.0%. Among the 62 patients, 44 showed persistent culture positivity at 8 weeks after treatment with an amikacin-containing multidrug regimen. The median parenteral amikacin treatment duration after 8 weeks in these patients was 18.0 weeks. The conditional probability of microbiological cure with continuation of the amikacin-containing regimen in these patients was 18.2% (95% confidence interval 8.2–32.7). Additionally, the conditional probability of microbiological cure in the 34 patients with persistent culture positivity at 12 weeks was 8.8% (95% confidence interval 1.9–23.7). After 16 weeks, the conditional probability of microbiological cure decreased further, reaching 0% at 28 weeks after treatment initiation.ConclusionThe continuation of intravenous amikacin therapy was usually not followed by culture conversion in M. abscessus PD patients with persistent sputum culture positivity after treatment initiation.     

Construction and Application of Chain Management Information System for Cancer Pain

BackgroundWith the rapid development of information technology, hospital information systems (HISs) have been deeply applied in the medical field and have shown broad application prospects. There are still some noninteroperable clinical information systems that pose an obstacle to the effective coordination of care, such as cancer pain management.AimTo construct a chain management information system for cancer pain and explore its clinical application effect.MethodsA quasiexperimental study was conducted in the inpatient department of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine. A total of 259 patients were nonrandomly divided into two groups: the experimental group (after the system was applied, n = 123) and the control group (before the system was applied, n = 136). The cancer pain management evaluation form score, patient satisfaction with pain control, pain score at admission and discharge, and the worst pain score during hospitalization were compared between the two groups.ResultsCompared with the control group, the score of the cancer pain management evaluation form was significantly higher (p < .05). There were no statistically significant differences in worst pain intensity, pain score at admission and discharge, and patients’ satisfaction with pain control between the two groups.ConclusionsThe cancer pain chain management information system can enable nurses to evaluate and record pain in a more standardized way, but it has no significant effect on the pain intensity of cancer patients.     

Stability of benzylpenicillin for continuous intravenous infusions: An isotonic formulation for therapeutic use and a low-dose formulation for clinical trial

IntroductionThe objectives of this study were to develop a stability-indicating high performance liquid chromatography (HPLC) assay for benzylpenicillin (BPC) in pharmaceutical fluids, and to investigate the stability of (i) isotonic citrate-buffered BPC solutions at the clinically relevant concentration of 30 mg/mL, and (ii) low concentration citrate-buffered BPC intravenous infusions (5–30 μg/mL).MethodsThe stability of isotonic BPC solutions containing 3.4 or 7.2 mg/mL sodium citrate was compared against contemporary hypertonic solutions. The HPLC assay was shown to be stability-indicating following acidic, alkali, oxidative and elevated temperature stress testing.ResultsAfter 7 d storage at 4 °C and 24 h at 35 °C, the concentrations of isotonic BPC 30 mg/mL solutions containing 3.4 and 7.2 mg/mL sodium citrate were 96% and 95% respectively, compared to day 0. After 3 d at 4 °C and 24 h at room temperature (22 °C), the concentrations of isotonic BPC solutions with 3.4 and 7.2 mg/mL sodium citrate were 99% and 96% respectively, compared to day 0. These data were comparable to the hypertonic solutions and meet pharmacopeial stability requirements. Low concentration BPC infusions showed 0.5% and 2.5% degradation after 24 h storage at 22 °C and 35 °C, respectively.ConclusionsThe isotonic BPC 30 mg/mL formulation is simple to prepare and may offer clinical benefits in settings where hypertonic solutions are problematic. This study provides assurance that high- and low-dose isotonic BPC infusions are stable at room temperature and our findings may be applicable to in vitro studies of BPC.     

An Electronic Health Record–Compatible Model to Predict Personalized Treatment Effects From the Diabetes Prevention Program: A Cross-Evidence Synthesis Approach Using Clinical Trial and Real-World Data

ObjectiveTo develop an electronic health record (EHR)-based risk tool that provides point-of-care estimates of diabetes risk to support targeting interventions to patients most likely to benefit.Patients and MethodsA risk prediction model was developed and validated in a large observational database of patients with an index visit date between January 1, 2012, and December 31, 2016, with treatment effect estimates from risk-based reanalysis of clinical trial data. The risk model development cohort included 1.1 million patients with prediabetes from the OptumLabs Data Warehouse (OLDW); the validation cohort included a distinct sample of 1.1 million patients in OLDW. The randomly assigned clinical trial cohort included 3081 people from the Diabetes Prevention Program (DPP) study.ResultsEleven variables reliably obtainable from the EHR were used to predict diabetes risk. This model validated well in the OLDW (C statistic = 0.76; observed 3-year diabetes rate was 1.8% (95% confidence interval [CI], 1.7 to 1.9) in the lowest-risk quarter and 19.6% (19.4 to 19.8) in the highest-risk quarter). In the DPP, the hazard ratio (HR) for lifestyle modification was constant across all levels of risk (HR, 0.43; 95% CI, 0.35 to 0.53), whereas the HR for metformin was highly risk dependent (HR, 1.1; 95% CI, 0.61 to 2.0 in the lowest-risk quarter vs HR, 0.45; 95% CI, 0.35 to 0.59 in the highest-risk quarter). Fifty-three percent of the benefits of population-wide dissemination of the DPP lifestyle modification and 73% of the benefits of population-wide metformin therapy can be obtained by targeting the highest-risk quarter of patients.ConclusionThe Tufts–Predictive Analytics and Comparative Effectiveness DPP Risk model is an EHR-compatible tool that might support targeted diabetes prevention to more efficiently realize the benefits of the DPP interventions.     

A multimodal deep learning model for cardiac resynchronisation therapy response prediction

We present a novel multimodal deep learning framework for cardiac resynchronisation therapy (CRT) response prediction from 2D echocardiography and cardiac magnetic resonance (CMR) data. The proposed method first uses the ‘nnU-Net’ segmentation model to extract segmentations of the heart over the full cardiac cycle from the two modalities. Next, a multimodal deep learning classifier is used for CRT response prediction, which combines the latent spaces of the segmentation models of the two modalities. At test time, this framework can be used with 2D echocardiography data only, whilst taking advantage of the implicit relationship between CMR and echocardiography features learnt from the model. We evaluate our pipeline on a cohort of 50 CRT patients for whom paired echocardiography/CMR data were available, and results show that the proposed multimodal classifier results in a statistically significant improvement in accuracy compared to the baseline approach that uses only 2D echocardiography data. The combination of multimodal data enables CRT response to be predicted with 77.38% accuracy (83.33% sensitivity and 71.43% specificity), which is comparable with the current state-of-the-art in machine learning-based CRT response prediction. Our work represents the first multimodal deep learning approach for CRT response prediction.     

Loeffler’s Endocarditis: An Integrated Multimodality Approach

1. Download : Download high-res image (245KB)
2. Download : Download full-size image

     

What do lymphedema patients expect from a treatment and what do they achieve? A descriptive study

PurposeLymphedema is characterized by swelling and fibroadipose tissue deposition that is a physically, psychologically, and socially debilitating condition due to chronic and progressive nature of the disease. Treatment benefit evaluation from the patient's perspective is important for medical decision-making. The aim of this study is to investigate important treatment goals and benefits of treatment from the patients’ perspective.MethodEighty-one patients with lymphedema, lipoedema, or lipolymphoedema who are currenlty treated or who underwent previous treatment were included in the study. Socio-demographic data was recorded. Important goals and benefit from treatment were assessed with Patients Needs Questionnaire and Patient Benefit Questionnaire which are sub-questionnaires of Patient Benefit Index-Lymphedema.ResultsThe most important expectation and needed item was “To find a clear diagnosis and therapy” (n:59, 72%). The least important item for the lymphedema patients was “To feel more attractive” (n:9, 11%). Most beneficial effect of treatment was “To have no fear that the disease will become worse” (n:37, 45.7%). “To have fewer out of pocket treatment expenses” was rated as the least beneficial effect of treatment (n:24, 29.6%).ConclusionsIt is important to identify patients' needs and expectations. Patients should be referred for treatment according to their needs. The effectiveness of the treatment should be evaluated objectively. Patient education should be considered as a part of the effective treatment to teach patients how to control their lymphedema. A clear diagnosis and access to treatment should be ensured for lymphedema patients. Regulations for health insurance benefit coverage are needed to cover cost of compression garments.     

Etomidate as an induction agent for endotracheal intubation in critically ill patients: A meta-analysis of randomized trials

PurposeWe performed a meta-analysis of randomized controlled trials to evaluate if etomidate impacted mortality in critically ill adults when compared with other induction agents.Materials and methodsWe searched PubMed, Embase, and the Cochrane Central Register of Controlled Trials for randomized controlled trials which compared etomidate with any other induction agent in critically ill adult patients undergoing endotracheal intubation. The primary outcome was mortality at the main timepoint defined by the study. We conducted a fixed-effects meta-analysis for the risk ratio. Using that risk ratio and 95% confidence interval, we then estimated the probability of any harm (RR > 1) and the number needed to harm ≤100 (RR ≥ 1.05).ResultsWe included 11 randomized trials comprising 2704 patients. We found that etomidate increased mortality (319/1359 [23%] vs. 267/1345 [20%]; risk ratio (RR) = 1.16; 95% confidence interval (CI), 1.01–1.33; P = 0.03; I² = 0%; number needed to harm = 31). The probabilities of any increase and a 1% increase (NNH ≤100) in mortality were 98.1% and 92.1%, respectively.ConclusionsThis meta-analysis found a high probability that etomidate increases mortality when used as an induction agent in critically ill patients with a number needed to harm of 31.     

Functional Regurgitation of Atrioventricular Valves and Atrial Fibrillation: An Elusive Pathophysiological Link Deserving Further Attention

1. Download : Download high-res image (158KB)
2. Download : Download full-size image