The STOIC2021 COVID-19 AI challenge: Applying reusable training methodologies to private data.

Challenges drive the state-of-the-art of automated medical image analysis. The quantity of public training data that they provide can limit the performance of their solutions. Public access to the training methodology for these solutions remains absent. This study implements the Type Three (T3) challenge format, which allows for training solutions on private data and guarantees reusable training methodologies. With T3, challenge organizers train a codebase provided by the participants on sequestered training data. T3 was implemented in the STOIC2021 challenge, with the goal of predicting from a computed tomography (CT) scan whether subjects had a severe COVID-19 infection, defined as intubation or death within one month. STOIC2021 consisted of a Qualification phase, where participants developed challenge solutions using 2000 publicly available CT scans, and a Final phase, where participants submitted their training methodologies with which solutions were trained on CT scans of 9724 subjects. The organizers successfully trained six of the eight Final phase submissions. The submitted codebases for training and running inference were released publicly. The winning solution obtained an area under the receiver operating characteristic curve for discerning between severe and non-severe COVID-19 of 0.815. The Final phase solutions of all finalists improved upon their Qualification phase solutions.

A pathomic approach for tumor-infiltrating lymphocytes classification on breast cancer digital pathology images.

Background and objectives: The detection of tumor-infiltrating lymphocytes (TILs) could aid in the development of objective measures of the infiltration grade and can support decision-making in breast cancer (BC). However, manual quantification of TILs in BC histopathological whole slide images (WSI) is currently based on a visual assessment, thus resulting not standardized, not reproducible, and time-consuming for pathologists. In this work, a novel pathomic approach, aimed to apply high-throughput image feature extraction techniques to analyze the microscopic patterns in WSI, is proposed. In fact, pathomic features provide additional information concerning the underlying biological processes compared to the WSI visual interpretation, thus providing more easily interpretable and explainable results than the most frequently investigated Deep Learning based methods in the literature. Methods: A dataset containing 1037 regions of interest with tissue compartments and TILs annotated on 195 TNBC and HER2+ BC hematoxylin and eosin (H&E)-stained WSI was used. After segmenting nuclei within tumor-associated stroma using a watershed-based approach, 71 pathomic features were extracted from each nucleus and reduced using a Spearman's correlation filter followed by a nonparametric Wilcoxon rank-sum test and least absolute shrinkage and selection operator. The relevant features were used to classify each candidate nucleus as either TILs or non-TILs using 5 multivariable machine learning classification models trained using 5-fold cross-validation (1) without resampling, (2) with the synthetic minority over-sampling technique and (3) with downsampling. The prediction performance of the models was assessed using ROC curves. Results: 21 features were selected, with most of them related to the well-known TILs properties of having regular shape, clearer margins, high peak intensity, more homogeneous enhancement and different textural pattern than other cells. The best performance was obtained by Random-Forest with ROC AUC of 0.86, regardless of resampling technique. Conclusions: The presented approach holds promise for the classification of TILs in BC H&E-stained WSI and could provide support to pathologists for a reliable, rapid and interpretable clinical assessment of TILs in BC.

When Doctors and AI Interact: on Human Responsibility for Artificial Risks.

A discussion concerning whether to conceive Artificial Intelligence (AI) systems as responsible moral entities, also known as "artificial moral agents" (AMAs), has been going on for some time. In this regard, we argue that the notion of "moral agency" is to be attributed only to humans based on their autonomy and sentience, which AI systems lack. We analyze human responsibility in the presence of AI systems in terms of meaningful control and due diligence and argue against fully automated systems in medicine. With this perspective in mind, we focus on the use of AI-based diagnostic systems and shed light on the complex networks of persons, organizations and artifacts that come to be when AI systems are designed, developed, and used in medicine. We then discuss relational criteria of judgment in support of the attribution of responsibility to humans when adverse events are caused or induced by errors in AI systems.

Impact of pulse oximetry screening on the detection of duct dependent congenital heart disease: a Swedish prospective screening study in 39,821 newborns.

OBJECTIVE: To evaluate the use of pulse oximetry to screen for early detection of life threatening congenital heart disease. DESIGN: Prospective screening study with a new generation pulse oximeter before discharge from well baby nurseries in West Götaland. Cohort study comparing the detection rate of duct dependent circulation in West Götaland with that in other regions not using pulse oximetry screening. Deaths at home with undetected duct dependent circulation were included. SETTING: All 5 maternity units in West Götaland and the supraregional referral centre for neonatal cardiac surgery. PARTICIPANTS: 39,821 screened babies born between 1 July 2004 and 31 March 2007. Total duct dependent circulation cohorts: West Götaland n=60, other referring regions n=100. MAIN OUTCOME MEASURES: Sensitivity, specificity, positive and negative predictive values, and likelihood ratio for pulse oximetry screening and for neonatal physical examination alone. RESULTS: In West Götaland 29 babies in well baby nurseries had duct dependent circulation undetected before neonatal discharge examination. In 13 cases, pulse oximetry showed oxygen saturations

Age- and gender-specific mortality rates in childhood hypertrophic cardiomyopathy.

AIMS: Hypertrophic cardiomyopathy (HCM) is the commonest inherited cause of sudden cardiac death in children; current guidelines suggest HCM screening after 12-15 years of age. The study aims to establish the age range at highest risk. METHODS AND RESULTS: Cohort study from six regional centres of paediatric cardiology, including children presenting with sudden death; n = 150 (59% = male; 39% familial HCM). Age- and gender-specific mortality was calculated, and compared with rates calculated from the Swedish National Cause of Death Registry. There were 56 deaths within the cohort, 39 were sudden arrhythmia deaths, with 31 at <19 years of age. Between 9-13.9 years of age annual sudden death mortality averages 7.2%, vs. 1.7% after 16 years of age; P = 0.025, odds ratio for proportions 3.75 [95% confidence intervals (CI) 1.18-11.91], similar in both familial and idiopathic HCM. The risk for sudden death peaks earlier in girls (10-11 years), with male preponderance after the age of 15. National cause of death statistics confirm that the mortality rate from HCM is significantly higher in the 8-16 year olds (0.112 per 100,000 age-specific population) than in the 17-30 year olds (0.055 per 100,000; 95% CI 0.011-0.099). CONCLUSION: In families with HCM, children should be screened at an early age.

High mortality rates among GHB abusers in Western Sweden.

BACKGROUND: GHB is a drug of abuse and acute poisonings have been an increasing medical problem over the last decade in Sweden. OBJECTIVES: To document all cases of GHB poisonings in Gothenburg during 1995-2004 and to record drug-related deaths to compare the toxicity of GHB with other illicit drugs, such as heroin and amphetamine. METHODS: The number of GHB-poisoned patients treated at the Sahlgrenska University Hospital has been recorded with the help of an in-house database. The number of deaths by illicit drugs was recorded during 2004. Seizures of the drugs GHB, 1,4-butanediol, and GBL were registered between 1996 and 2004. RESULTS: The number of poisoned patients was 259. The number of seizures with GHB was 743, GBL 343, and 1,4-butanediol 236. In 2004 the number of deaths was 6 with heroin, 7 with GHB, 32 with amphetamine, 6 with cocaine, and one with methadone. One patient with GHB poisoning died during hospital care. CONCLUSIONS: Intoxication by GHB has substantial morbidity and abuse of GHB has substantial mortality. The acute prognosis is good but long-term prognosis is insecure with an increased risk for drug dependency and an early death.

[GHB, GBL and butanediol poisonings--a serious problem in Western Sweden]. / GHB-, GBL- och butandiolförgiftningar--ett allvarligt problem i Västsverige.

Acute poisoning with GHB (Gamma-hydroxybutyrate, Gamma-hydroxybutyric acid) has been an increasing medical and social problem during the last decade in Sweden, especially on the west coast. The number of poisonings decreased in the beginning of this millennium but has again increased during the last years. At the same time the number of seizures by the police has increased similarly as well as the number of drug-related deaths. During 2004 the number of GHB-abuse related deaths in western Sweden was seven, approximately the same figures as for heroin. Two other substances which are transformed to GHB in the human body, GBL (gamma-butyrolactone) and 1,4-butanediol, have during the last years presented themselves among the same users as for GHB. Since GBL and butanediol are not classified as illicit drugs the possibilities for the police force to intervene and capture the drugs are severely restricted. Intoxication by GBL and butanediol has shown to be as dangerous as intoxication by GHB. Acute intoxications and abuse of these drugs is still a serious medical and social problem. A legal classification of GBL and butanediol as narcotics appears to be medically motivated.