Early Prediction of Sepsis From Clinical Data

1. Résumé
2. Objectives
3. The Data

• Try this Binder for EDA :

• Try this Binder for Presentation :

• Try this Binder for Demonstartion :

• Six ML Algorithm

• MLP

Résumé :

Intensive care units (ICU) are the part of hospitals where decision making is very important, complex and critical at the same time. Deciding correctly and at the right time who is the patient who needs critical care and who does not can reduce the pressure on these units, ensuring good management and allocation of resources which are both limited and very expensive. This good management will guarantee a considerable saving in costs by reducing them considerably. In fact, people whose condition is critically ill as a result of accidents, surgical complications, serious respiratory problems, infections and serious injuries can achieve death if they are not treated with technology and sophisticated equipment allowing close and constant monitoring and maintenance of bodily functions. In particular, sepsis is a major health problem and life-threatening illness that occurs when the body's response to infection causes tissue damage, organ failure, or death. Worldwide, around 30 million people develop sepsis and 6 million people die from sepsis each year. The costs of managing sepsis in hospitals exceed those of any other health problem. The majority of these costs relate to patients who develop sepsis while in hospital. Early detection and antibiotic treatment of sepsis improves outcomes. However, although professional critical care societies have proposed new clinical criteria that facilitate recognition of sepsis, the basic need for early detection and treatment remains unmet. It is in this context that the objective of this professional master's degree is to develop a machine learning-based model for the early and real-time prediction of sepsis from clinical data in intensive care units. This involves designing and implementing a functional predictive model to automatically identify a patient's risk of sepsis and make a positive or negative prediction of sepsis for each time window in the patient's clinical record.

Objectives:

Sepsis is a major public health concern with significant morbidity, mortality, and healthcare expenses. Early detection and antibiotic treatment of sepsis improve outcomes. However, although professional critical care societies have proposed new clinical criteria that aid sepsis recognition, the fundamental need for early detection and treatment remains unmet. In response, researchers have proposed algorithms for early sepsis detection, but directly comparing such methods has not been possible because of different patient cohorts, clinical variables and sepsis criteria, prediction tasks, evaluation metrics, and other differences. To address these issues, the PhysioNet/Computing in Cardiology Challenge 2019 facilitated the development of automated, open-source algorithms for the early detection of sepsis from clinical data.

To do this the Knowledge Discovery in Databases (KDD) process will be adopted. This process consists of 4 steps:

1. Selecting and extracting the multivariate time series data set from the database.
2. Pre-processed and cleaned up the time series into a tidy dataset by exploring the data, handling missing data, and removing noise / outliers.
3. Building a predictive model allowing to associate the time series of biomedical measurements with a real-time severity indicator (probability of mortality) by implementing several machine learning algorithms.
4. Evaluation and validation. The use of two databases provides an opportunity to validate the different predictive models that will be produced.

The Data

Click here to download the complete training database (42 MB), consisting of two parts: training set A (20,336 subjects) and B (20,000 subjects). Each training data file provides a table with measurements over time. Each column of the table provides a sequence of measurements over time (e.g., heart rate over several hours), where the header of the column describes the measurement. Each row of the table provides a collection of measurements at the same time (e.g., heart rate and oxygen level at the same time). The table is formatted in the following way:

HR	O2Sat	Temp	...	HospAdmTime	ICULOS	SepsisLabel
NaN	NaN	NaN	...	-50	1	0
86	98	NaN	...	-50	2	0
75	NaN	NaN	...	-50	3	1
99	100	35.5	...	-50	4	1

There are 40 time-dependent variables HR, O2Sat, Temp ..., HospAdmTime, which are described here. The final column, SepsisLabel, indicates the onset of sepsis according to the Sepsis-3 definition, where 1 indicates sepsis and 0 indicates no sepsis. Entries of NaN (not a number) indicate that there was no recorded measurement of a variable at the time interval.

Note: spaces were added to this example to improve readability. They will not be present in the data files.

TABLE 1. Feature Summary

Row	Measurement	Description
1	HR	Heart rate (beats/min)
2	O2Sat	Pulse oximetry (%)
3	Temp	Temperature (°C)
4	SBP	Systolic BP (mm Hg)
5	MAP	Mean arterial pressure (mm Hg)
6	DBP	Diastolic BP (mm Hg)
7	Resp	Respiration rate (breaths/min)
8	Etco2	End tidal carbon dioxide (mm Hg)
9	BaseExcess	Excess bicarbonate (mmol/L)
10	Hco3	Bicarbonate (mmol/L)
11	Fio2	Fraction of inspired oxygen (%)
12	pH	pH
13	Paco2	Partial pressure of carbon dioxide fromarterial blood (mm Hg)
14	Sao2	Oxygen saturation from arterial blood (%)
15	AST	Aspartate transaminase (IU/L)
16	BUN	Blood urea nitrogen (mg/dL)
17	Alkalinephos	Alkaline phosphatase (IU/L)
18	Calcium	Calcium (mg/dL)
19	Chloride	Chloride (mmol/L)
20	Creatinine	Creatinine (mg/dL)
21	Bilirubin	direct Direct bilirubin (mg/dL)
22	Glucose	Serum glucose (mg/dL)
23	Lactate	Lactic acid (mg/dL)
24	Magnesium	Magnesium (mmol/dL)
25	Phosphate	Phosphate (mg/dL)
26	Potassium	Potassium (mmol/L)
27	Bilirubin total	Total bilirubin (mg/dL)
28	TroponinI	Troponin I (ng/mL)
29	Hct	Hematocrit (%)
30	Hgb	Hemoglobin (g/dL)
31	PTT	Partial thromboplastin time (s)
32	WBC	Leukocyte count (count/L)
33	Fibrinogen	Fibrinogen concentration (mg/dL)
34	Platelets	Platelet count (count/mL)
35	Age	Age (yr)
36	Gender	Female (0) or male (1)
37	Unit 1	Administrative identifier for ICU unit (medical ICU); false (0) or true (1)
38	Unit 2	Administrative identifier for ICU unit (surgical ICU); false (0) or true (1)
39	HospAdmTime	Time between hospital and ICU admission (hours since ICU admission)
40	ICULOS	ICU length of stay (hours since ICU admission)
41	SepsisLabel	For septic patients, SepsisLabel is 1 if $t \geq t_{sepsis}-6$ and $0$ if $t < t_{sepsis}-6$.For nonseptic patients, SepsisLabel is 0.

Clinical time series data provided :

• Vital signs (rows 1–8),
• Laboratory values (rows 9–34)
• Demographics (rows 35–40)
• Outcome row 41.

Try this Binder for EDA :