Report Open Access

Portable Early Prediction of Sepsis from Clinical Data on Intel Myriad X

Priyanka Kumar Mathur

Sepsis is a life-threatening condition where microbes present in the blood stream cause an 
unregulated immune response from the body which can result in tissue damage, multi-organ failure 
and eventually death. It affects 30 million people worldwide and causes 6 million deaths. Studies 
have indicated that every hour sepsis goes undetected, patient mortality increases 4-8%, thus early 
detection of the disease is necessary to decrease mortality rates and provide better patient 
outcomes.  
This project is derived from this year’s 2019 Computing in Cardiology Challenges, which focuses on 
the early detection of sepsis using machine learning algorithms. The two primary focuses of the 
project include (1) developing a model which can successfully detect sepsis early and (2) 
demonstrating the clinical viability of this deep learning model by implementing the model on Intel’s 
Neural Compute Stick 2 (Myriad X VPU processor) which is a portable USB device that can 
implement and deploy deep learning models. 
While developing the model, the data was preprocessed in the following ways: (1) linear interpolation 
was used to fill all missing values and values not collected were 0, (2) synthetic minority oversample 
technique (SMOT) was used to balance the data set, (3) patient file lengths were standardized to 
have only 3 and 17 hours of patient data, (4) all 40 parameters were used (although a variety of 
subsets were tested). Data preprocessing was done using MATLAB. 
The model was developed in python using Tensorflow and Keras. The model itself was a seven layer 
dense neural network (DNN) with Leaky Rectified Linear Unit (ReLU) activation functions for all 
hidden layers and a sigmoid activation function for the output layer. This was paired with binary cross 
entropy loss.  
The resulting model detected sepsis well, with a 76% sensitivity (correctly identified sepsis) and 80% 
specificity (correctly identified healthy patients). The model was successfully implemented on Intel’s 
Neural compute stick 2 (Myriad X VPU processor) with an average of 10ms processing time to 
determine whether the patient had sepsis, demonstrating clinical viability. Further testing including 
tuning hyperparameters to further increase sensitivity and specificity is currently being done.

Files (922.6 kB)
Name Size
Report_Priyanka_Mathur.pdf
md5:67951523e09f886f0509d824d1416378
922.6 kB Download
206
195
views
downloads
All versions This version
Views 206206
Downloads 195195
Data volume 179.9 MB179.9 MB
Unique views 186186
Unique downloads 178178

Share

Cite as