Peering Into the Black Box of Artificial Intelligence: Evaluation Metrics of Machine Learning Methods.

OBJECTIVE: Machine learning (ML) and artificial intelligence (AI) are rapidly becoming the most talked about and controversial topics in radiology and medicine. Over the past few years, the numbers of ML- or AI-focused studies in the literature have increased almost exponentially, and ML has become a hot topic at academic and industry conferences. However, despite the increased awareness of ML as a tool, many medical professionals have a poor understanding of how ML works and how to critically appraise studies and tools that are presented to us. Thus, we present a brief overview of ML, explain the metrics used in ML and how to interpret them, and explain some of the technical jargon associated with the field so that readers with a medical background and basic knowledge of statistics can feel more comfortable when examining ML applications. CONCLUSION: Attention to sample size, overfitting, underfitting, cross validation, as well as a broad knowledge of the metrics of machine learning, can help those with little or no technical knowledge begin to assess machine learning studies. However, transparency in methods and sharing of algorithms is vital to allow clinicians to assess these tools themselves.

Carotid artery stenting: Current state of evidence and future directions.

Both carotid endarterectomy (CEA) and carotid artery stenting (CAS) are common treatments for carotid artery stenosis. Several randomized controlled trials (RCTs) have compared CEA to CAS in the treatment of carotid artery stenosis. These studies have suggested that CAS is more strongly associated with periprocedural stroke; however, CEA is more strongly associated with myocardial infarction. Published long-term outcomes report that CAS and CEA are similar. A reduction in complications associated with CAS has also been demonstrated over time. The symptomatic status of the patient and history of previous CEA or cervical radiotherapy are significant factors when deciding between CEA or CAS. Numerous carotid artery stents are available, varying in material, shape and design but with minimal evidence comparing stent types. The role of cerebral protection devices is unclear. Dual antiplatelet therapy is typically prescribed to prevent in-stent thrombosis, and however, evidence comparing periprocedural and postprocedural antiplatelet therapy is scarce, resulting in inconsistent guidelines. Several RCTs are underway that will aim to clarify some of these uncertainties. In this review, we summarize the development of varying techniques of CAS and studies comparing CAS to CEA as treatment options for carotid artery stenosis.

Media messaging in diagnosis of acute CXR pathology: an interobserver study among residents.

The objectives of the study were to determine whether diagnostic accuracy and reliability by on-call teams is affected by communicating chest radiograph (CXR) images via instant messaging on smartphones in comparison to viewing on a workstation. 12 residents viewed 100 CXR images each with a 24% positive rate for significant or acute findings sent to their phones via a popular instant messaging application and reported their findings if any. After an interval of 42 days they viewed the original DICOM images on personal computers and again reported their findings. There were no statistically significant differences in accuracy, agreement, sensitivity, specificity, positive predictive value or negative predictive value between desktop workstation viewed images and images sent via the mobile application. Media messaging is a useful adjunct for quick second opinions on radiological images, without significant decay in diagnostic accuracy. If technical, ethical and legal issues are addressed, it could be incorporated into practice as a useful adjunct.

Meta-analysis of the clinicopathological characteristics and peri-operative outcomes of colorectal cancer in obese patients.

BACKGROUND: The effect of obesity on the clinicopathological characteristics of colorectal cancer (CRC) has not been clearly characterized. This meta-analysis assesses the pathological and perioperative outcomes of obese patients undergoing surgical resection for CRC. METHODS: Meta-analysis was performed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Databases were searched for studies reporting outcomes for obese and non-obese patients undergoing primary CRC resection, based on body-mass index measurement. Results were reported as mean differences or pooled odds ratios (OR) with 95% confidence intervals (95% CI). RESULTS: A total of 2183 citations were reviewed; 29 studies comprising 56,293 patients were ultimately included in the analysis, with an obesity rate of 19.3%. Obese patients with colorectal cancer were more often female (OR 1.2, 95% CI 1.1-1.2, p<0.001) but there was no difference in the proportion of rectal cancers, T4 tumours, tumour differentiation or margin positivity. Obese patients were significantly more likely to have lymph node metastases (OR 1.2, 95% CI 1.1-1.2, p<0.001), have a lower nodal yield, were associated with a longer duration of surgery, more blood loss and conversions to open surgery (OR 2.6, 95% CI 1.6-4.0, p<0.001) but with no difference in length of stay or post-operative mortality. CONCLUSION: This meta-analysis demonstrates that obese patients undergoing resection for CRC are more likely to have node positive disease, longer surgery and higher failure rates of minimally invasive approaches. The challenges of colorectal cancer resection in obese patients are emphasized.