As I have a weakness for studies that challenge the dogma, I was ecstatic when I stumbled upon Carson et al.’s bold study on blood transfusions. In their study they selected 2000 patients with cardiovascular disease that were undergoing surgery due to hip fracture. Interestingly, there was no difference in 3-year mortality when randomizing between transfusion Hemoglobin thresholds of 100 g/L or 80 g/L! Continue reading
One of the successful insights to training neural networks has been the rectified linear unit, or short the ReLU, as a fast alternative to the traditional activation functions such as the sigmoid or the tanh. One of the major advantages of the simle ReLu is that it does not saturate at the upper end, thus the network is able to distinguish a poor answer from a really poor answer and correct accordingly.
A modification to the ReLU, the Leaky ReLU, that would not saturate in the opposite direction has been tested but did not help. Interestingly in a recent paper by the Microsoft© deep learning team, He et al. revisited the subject and introduced a Parametric ReLU, the PReLU, achieving superhuman performance on the imagenet. The PReLU learns the parameter α (alpha) and adjusts it through basic gradient descent.
In this tutorial I will benchmark a few different implementations of the ReLU and PReLU together with Theano. The benchmark test will be on the MNIST database, mostly for convenience. Continue reading
I recently e-mailed Freakonomics asking them to look into the economic incentives that govern health-care. Guess if I was thrilled when I found these two recent gems (although I can’t take any credit):
- How Do We Know What Really Works in Healthcare?
- How Many Doctors Does It Take to Start a Healthcare Revolution?
Freakonomics delivers as always with a fresh, interesting scientific perspective. As a supporter since a few years I want to recommend these to anyone interested in the topic. As I have written earlier, I find it perplexing that so many arthroscopies are being performed while so much evidence is pointing against this. There is probably a vast combination of incentives that result in this, these podcasts provide some nice insights.
htmlTable-function has perhaps been one of my most successful projects. I developed it in order to get tables matching those available in top medical journals. As the function has grown I’ve decided to separate it from my Gmisc-package into a separate package, and at the time of writing this I’ve just released the 1.3 version. While
htmlTable allows for creating plain tables without any fancy formatting (see usage vignette) it is primarily aimed at complex tables. In this post I’ll try to show you what you can do and how to tame some of the more advanced features. Continue reading
Costa et. al. performed recently a multi-center study including 18 trauma centers throughout the United Kingdom. They randomized patients with distal radius fractures to volar locking plates or Kirschner wires. They managed to randomize 461 patients and follow-up 90% after 1 year using self-reported scores. They could not find any difference after 1 year regarding their primary outcome, PRWE-score (−1.3 points, 95% CI -4.5 to 1.8). This was well below their stipulated minimally clinically important difference of 6 points. Could this be the end of the volar locking plate era? Continue reading
Today is a good day to start parallelizing your code. I’ve been using the parallel package since its integration with R (v. 2.14.0) and its much easier than it at first seems. In this post I’ll go through the basics for implementing parallel computations in R, cover a few common pitfalls, and give tips on how to avoid them. Continue reading
One, never-ending thread on this blog seems to be the degenerative meniscus. Gauffin et al. recently published a randomized controlled trial comparing surgery vs physiotherapy with favorable results for surgery regarding pain. Could this be the turnaround for one of the most controversial, but common orthopaedic procedures? Continue reading
Surgical treatment for trigger finger consists of cleaving the A1 pulley, a simple procedure, considered to be low risk. Even so, there can still be scar tenderness, nerve injury, tendon bowing, and infection. It is therefore nice to see Wojahn et al.’s paper on long-term follow-up after cortisone injections. Injection is a simple alternative that consists of an injection angling 45° distally at the A1 pulley with 1 mL 40 mg/mL Depo-Medrol. Continue reading
Nordenvall et al.  recently report in PLOS ONE an interesting study on cruciate ligament (CL) injuries where they could not find any protective effect on knee osteoarthritis (OA) from reconstructive surgery. It is a nation-wide study based on > 60,000 patients where they used the National Swedish Patient Register in order to find ICD-codes and surgical procedures code to identify patients with CL injury and those with a subsequent procedure. While the early results showed slightly less OA, there was an increased risk of 1.4 (95% confidence interval, 1.2 to 1.7) after > 10 years for those with surgery. Continue reading
I love the study from McDowell et al. where they tested different methods for keeping casts dry. It is simple but takes up a both common and important issue that we doctors frequently forget about. They compare six different methods and it seems that double plastic bags secured with duct tape provide the best protection for the money. Continue reading