PREDICTION OF BLOOD LACTATE LEVELS IN CHILDREN AFTER

CARDIAC SURGERY USING MACHINE LEARNING

Ms. M. Latha, C. Vennela, M. Charan sai, K. Pavani, A. Giresh kumar, B. Venkatasivakumar

1 Professor, Dept. of CSE, SEAT , [Link], AP, India2,3,4,5,6 UG Student, Dept. of CSE, SEAT, [Link] , AP, India

[4]. If necessary, a detailed study is carried out to

evaluate each isoform of the enzyme test, methods
Abstract: The abstract of the research paper of inactivation with urea and heat inhibition)
discusses the importance of measuring lactate
dehydrogenase (LDH) levels in blood tests to [5]. In the blood of a healthy person, LDH is
determine various pathologies and monitor a stable and has specific reference values. An
patient's health. LDH is an enzyme found in altered concentration of lactated hydrogenase does
different tissues of the body, and its concentration not diagnose a specific disease
can indicate tissue damage or cell death. The
study focuses on predicting blood lactate levels in [6]. The obtained values are compared with other
children after cardiac surgery using machine indicators of biochemical analysis
learning algorithms. By analysis the LDH levels
and other biochemical parameters, the researchers [7]. Unsatisfactory results are grounds for further
aim to improve post-operative care and enhance examination of a particular system or organ
the understanding of patient recovery dynamics. [8]. Lactate dehydrogenate is an enzyme that
The use of machine learning in predicting blood
oxidation of glucose and the formation of lactic
lactate levels showcases the potential for advanced
acid. LDH accumulates in the kidney, liver, heart
technology to revolutionize healthcare practices.
tissue, and muscle
Keywords: Lactate, Enzyme, Acid, Blood tests, [9-10]. A specific isoenzyme is responsible for
Lactate Dehydrogenate, Biochemical, Cardiac, each organelle—guideline values for blood levels
Surgery, Machine learning
in adults. Children's indicators are classified by
age category
I. Introduction - An increase in the level of the
enzyme in the blood occurs during the destruction [11]. With the catastrophic injury of the listed
of LDH-containing cells organs, the enzyme is released into the systemic
circulation through the affected areas
[1]. The type of isoenzyme increased in the
[12]. Preparation for the blood test for LDH It is
biochemical course determines the location of the
unnecessary to eat anything for twelve hours
damage; that is, the organ cells are subject to
before the test. Also, experts advise against
destruction
smoking and drinking alcoholic beverages,
[2]. Enzyme is forms are numbered 1 to 5, engaging in vigorous physical activity, and eating
depending on their location. Increased activity of a lot
one or type of lactate dehydrogenase is a sign of
[13]. Also, do not spoil the mood’s Before the
destructive change in the tissues of the organs in
examination, the patient should be quiet and sit on
which it is contained
the sidewalk in front of the office to bring his
[3]. With standard biochemical analysis, the form heart rate back to normal and calm the heart
usually indicates the total LDH index

1|Page
 [14-15]. A blood test for LDH is usually 2 Chaudhary, An The 20
recommended, along with other necessary paper 22
A., Kolhe, S., improve
procedures, to help determine the presence of a presents
specific disease in the patient's tissues and organs & Kamal, R. d an
random improve
[16]. Due to tissue necrosis or myocardial
d-RFC
infarction, patients pass the study several times forest
(Rando
after a specific time
classifie m
II. Literature Review: r for Forest
Classifi
Sl. Name of the Title Descrip Ye multi- er)
No. Aurthor(s) tion ar approac
class
1 Behrooz Blood explore 20 h for
classific
Mamandipoor, lactate the 06 multi-
ation class
Mahshid concent problem
Majd, Monica ration of disease
Moz, Venet predicti classifyi classific
Osmani on in ng ation
critical images problem
care by the .
patients object 3 Gnanamanick An This is a 20
: categori overvie status 23
handlin es they am, S. S.,
w of paper
g contain Priyadarisini, bacteria on a
missing in the l blight
V. B., destruc
values case of disease
a large Narayanan, N. tive rice
of rice disease
number N., and
of -bac-
Vasudevan, P., strategi
object terial
es for
categori & Kavitha, S. blight
its
es. manage (BB),
ment caused
by
Xantho
monas
oryzae
pv.
oryzae
(Xoo).

III. Methodology: The application of machine

learning algorithms to predict blood lactate levels
in children after cardiac surgery. The study
compares the (PBLL) Prediction of Blood
Lactate Levels , (BLCP) Blood Lactate

2|Page
Concentration Prediction , (PBLV) Prediction of • LDH in child from 2 to 12 years old -
Blood Lactate Values , (MLPB) Machine Learning 295-300 U / L
Predicts Blood Lactate Levels & (MLPA)
• LDH adults - 235-250 U / L
Machine-Learning Prediction Algorithm .
• Data Collection: Gathering the data from III. DATASET: Davidson et al.2 has
pediatric patients who underwent cardio compiled a corpus con around 24783
surgery. Collecting the information tweets annotated the text by crowd
sourcing. This dataset represents three
such as pre-operative health status, surgery
classes of labels as “hate speech”,
details, and post-operative recovery data, “offensive language ” and “neither”. They
including blood lactate levels. begin with a hate speech lexicon
• Data Preprocessing : Clean the dataset by
handling missing values and removing
irrelevant Data.
• Feature Selection: - Identify relevant
features that could impact blood lactate
levels post-surgery.
• Data Splitting: Split the dataset into
training and testing sets to evaluate the containing words and phrases identified by
model's performance. internet users as hate speech, compiled by
[Link]. Using the TwitterAPI they
• Model Selection: Choose machine learning
searched for tweets containing terms from
algorithms suitable for prediction tasks, the lexicon, result- ing in a sample of
considering algorithms like Random tweets from 33,458 Twitter users. They
Forest, Support Vector Machines, decision extracted the time-line for each user,
tree. resulting in a set of 85.4 million tweets.
From this corpus they then took a random
• Model Training: Train the selected models
sample of 25k tweets containing terms
using the training dataset, adjusting from the lexicon and had them manually
hyperparameters as needed. coded by CrowdFlower (CF) workers. CF
• Hyperparameter Tuning Fine-tune Workers were asked to label each tweet as
hyperparameters to improve the model's one of the three categories. Each data file
contains 5 columns: Count, Hate- Speech,
performance. Consider techniques like
Offensive language, Neither, Class. In this
cross-validation for robust evaluation. study, we consider” Hate Speech” and
• Results Interpretation: Were interpreted to ”Offensive language” as offensive
understand the factors that impact blood language. To be precise, we have 20620
lactate levels in children after cardiac tweets out of 24783 total tweets as
offensive language (i.e., about 83% of
surgery.
tweets) which confirms that the dataset is
• The amount of LDH levels in the blood of highly imbalanced. Also, Fig. 2 shows that
different age groups is as follows: offensive messages tends to be shorter than
normal Ensuring that the input data
• LDH in a newborn child - 2000 U / L meets these requirements is crucial for the
successful training and deployment of a
• LDH For a child under two years - 425 U /
Random Forest model. Proper
L preprocessing, handling of missing values,
and careful consideration of the

3|Page
 characteristics of the dataset contribute to
the algorithm's effectiveness and
generalization
 Output Format :The output of the
Random Forest algorithm is multifaceted
and includes various components that
provide insights into the model's
performance and predictions. The primary
outputs are associated with both the
training phase, where the model learns
from the data, and the prediction phase,
where it makes predictions on new or Fig 1: Blood Lactate Levels prediction model
unseen instances.
 Trained Random Forest Model: The
most fundamental output is the trained
Random Forest model itself. This consists
of an ensemble of decision trees, each
capturing different aspects of the
relationships within the training data. The
model retains the knowledge gained during
the training phase, encapsulating the
patterns and decision rules learned from
the input features and target variable.
 Prediction Results: When the trained
Random Forest model is applied to new
data, it produces predictions. For
classification tasks, the output typically
includes predicted class labels for each
instance. For regression tasks, the output
Fig 2: Lactate level prediction
consists of predicted numerical values.
 IV. RESULTS AND
These predictions are generated by
DISCUSSION :The proposed Prediction
aggregating the individual predictions
of Blood Lactate Levels (PBLL) model has
from each tree in the ensemble . fields, and
compared with the existing Blood lactate
social sciences.
concentration prediction (BLCP),
Prediction of blood lactate values (PBLV),
Machine learning predicts blood lactate
levels (MLPB) and machine-learning
prediction algorithm (MLPA). In blood
serum, lactate dehydrogenase exists in five
isozyme types—LDH-1, LDH-2, LDG-3,
LDG-4, and LDH-5. These are specialized
molecular forms of lactate dehydrogenase
that have functional properties.
Biochemists describe LDH as a tetramer
consisting of two types of a specific

4|Page
 number of subunits - M (from the English
word "muscle") and H (from the English
word "heart"). Thus, the location of
localization of one or another type of LDH
is indicated. These are clinical signs of
myocardial infarction, cirrhosis, cancerous
pancreatic necrosis, pulmonary
thromboembolism, diseases of the renal
apparatus, and other pathologies associated
with tissue destruction and death of
cellular structures. Based on the study's
results, a specific disease was not
diagnosed. The dependence on the
indicators is the basis for a detailed patient
examination. The comparison of High Lactate dehydrogenate is a necessary isoenzyme
efficiency Management has shown in the found in the blood. It is through lactic acid
following table 1, produced in the body and glucose oxidation. If
LDH is normal, there is no such component
Table 1: Comparison of High efficiency because it is independently destroyed and excreted
Management (in %) from the body. However, when the level of LDH
INPUT BLC PBL MLP MLP PBL increases or decreases in clinical practice, it
S P V B A L reveals some diseases in the human body. The
evaluation based on the High efficiency
100 66.4 68.40 37.49 50.08 82.5
Management has shown in the fig 3. In a cut off
1 6
point, the proposed PBLL reached 85.41% of
200 66.3 68.42 37.32 49.81 82.0
High efficiency Management. In this same range,
0 6
the existing BLCP achieved 69.38%, PBLV
300 66.2 69.30 38.05 50.11 82.1 reached 72.13%, MLPB achieved 41.39% and
8 8
MLPA obtained 53.62%. The proposed model
400 69.3 72.13 41.39 53.62 85.4 achieved better High efficiency Management
8 1 results while compared with the existing models.
500 70.5 73.45 42.12 54.94 85.7 Considering that lactate dehydrogenate is a
8 9 component of many blood cells in blood diseases,
600 71.1 74.28 43.01 55.48 86.3 it is not surprising that it is used as a marker to
9 6 monitor various blood diseases, including liver
71.6 74.68 43.09 55.78 86.0 diseases. Hepatitis of various diseases causes the
0 6 death of hepatocytes - liver cells, which include
the biochemical compound LDH. An increase in
lactate dehydrogenate is detected in coronary
insufficiency, congestion in the liver or. A patient
has cardiac arrhythmia - all indicators are
standard, but if it is treated with electrical
impulses, the enzyme level will increase. With the
help of the enzyme lactate dehydrogenate in the
liver, kidney, and skeleton, it is diagnosed if the
patient has various pathological conditions. The

5|Page
comparison of Enzyme Management has shown in predict ion in critical care. In Digital Personalized
the following table 2, Health and Medicine (pp. 73-77). IOS Press.

INPUT BLC PBL MLP MLP PBL [2] Mamandipoor, B., Yeung, W., Agha-Mir-
S P V B A L Salim, L., Stone, D. J., Osmani, V., & Celi, L. A.
100 57.9 72.03 40.00 54.28 82.4 (2022). Predict ion of blood lactate values in crit
0 0 ically ill pat ients: a retrospect ive mult i-center
cohort study. Journal of clinical monitoring and
200 59.3 74.00 42.42 56.48 82.3 computing, 36(4), 1087-1097.
9 9
300 60.1 75.13 42.83 57.28 83.5 [3] Sughimoto, K., Levman, J., Baig, F., Berger,
9 9 D., Oshima, Y., Kurosawa, H., ... & Miyaji, K.
400 62.5 76.32 44.43 57.95 84.0 (2022). Machine learning predicts blood lactate
2 7 levels in children after cardiac surgery in paediat
ric ICU. Cardiology in the Young, 1-8.
500 63.5 76.71 46.75 59.38 85.5
3 0 [4] Mahmoud, E., Al Dhoayan, M., Bosaeed, M.,
600 64.1 78.23 48.00 60.47 86.6 Al Johani, S., & Arabi, Y. M. (2021). Developing
7 6 machine-learning predict ion algorithm for
bacteremia in admitted pat ients. Infect ion and
Drug Resistance, 14,757.
V. CONCLUSION: Since the development of
LDH indices occurs under the influence of dying [5] Heldt, F. S., Vizcaychipi, M. P., Peacock, S.,
tissue, the prediction depends on many factors. Cinelli, M., McLachlan, L., Andreott i, F., ... &
That is the degree of necrotic changes, the Khan, R. T. (2021). Early risk assessment for
localization of dying tissue, the extent of COVID-19 pat ients from emergency department
damage, the presence of concomitant diseases, data using machine learning. Scient ific reports,
and the effectiveness of treatment. When 11(1), 1-13.
detecting deaths in the early stages and not in
vital organs - the prognosis is helpful , and the [6] Aktar, S., Ahamad, M. M., Rashed-Al-
appointment of effective treatment and Mahfuz, M., Azad, A. K. M., Uddin, S., Kamal,
compliance with all recommended preventive A. H. M., ... & Moni, M. A. (2021). Machine
measures. If the tissue of the heart, liver, brain, learning approach to predict ing COVID-19
or other vital organs dies, the prognosis is disease severity based on clinical blood test data:
significantly worse. Here, the degree of stat ist ical analysis and model development.
JMIR medical informatics, 9(4), e25884.
mortality plays an important role, as well as the
speed and effectiveness of the prescribed
[7] Olaetxea, I., Valero, A., Lopez, E., Lafuente,
treatment. LDH in biochemical blood tests is a H., Izeta, A., Jaunarena,I., & Seifert , A. (2020).
parameter that clarifies the nature of many Machine learning-assisted Raman spectroscopy
specific and nonspecific diseases. LDH (lactate for pH and lactate sensing in body fluids. Analyt
dehydrogenase) is needed to determine the ical Chemist ry, 92(20), 13888-13895.
pathology in laboratory research and sometimes
to monitor the dynamics of the patient's [8] Seo, D. W., Yi, H., Park, B., Kim, Y. J., Jung,
recovery. D. H., Woo, I., ... &
Kim, W. Y. (2020). Predict ion of Adverse Events
in Stable Non-Variceal Gastrointest inal Bleeding
REFERENCES: Using Machine Learning. Journal of clinical
[1] Mamandipoor, B., Majd, M., Moz, M., & medicine, 9(8), 2603.
Osmani, V. (2020). Blood lactate concentrat ion

6|Page
[9] Giannini, H. M., Ginest ra, J. C., Chivers, C.,
Draugelis, M., Hanish,
A., Schweickert , W. D., ... & Umscheid, C. A.
(2019). A machine learning algorithm to predict
severe sepsis and sept ic shock: Development,
implementat ion and impact on clinical pract ice. .
Crit ical care medicine, 47(11), 1485.

[10] Kim, K. A., Choi, J. Y., Yoo, T. K., Kim, S.

K., Chung, K., & Kim, D. W. (2013). Mortality
predict ion of rats in acute hemorrhagic shock
using machine learning techniques. Medical &
biological engineering & comput ing, 51(9), 1059-
1067.

[11] Convert ino, V. A., Johnson, M. C.,

Alarhayem, A., Nicholson, S. E., Chung, K. K.,
DeRosa, M., & Eastridge, B. J. (2021).
Compensatory reserve detects subclinical shock
with more expedit ious predict ion for need of life
‐ saving intervent ions compared to systolic blood
pressure and blood lactate. Transfusion, 61, S167-
S173.
[12] Rawson, T. M., Hernandez, B., Moore, L. S.
P., Blandy, O., Herrero, P., Gilchrist , M., ... &
Holmes, A. H. (2019). Supervised machine
learning for the predict ion of infect ion on
admission to hospital: a prospect ive observat
ional cohort study. Journal of Ant imicrobial
Chemotherapy, 74(4), 1108-1115.

[13] Iwase, S., Nakada, T. A., Shimada, T., Oami,

T., Shimazui, T., Takahashi, N., ... & Kawakami,
E. (2022). Predict ion algorithm for ICU mortality
and length of stay using machine learning. Scient
ific Reports, 12(1), 1-9.

[14] Wu, W., & Zhou, Z. (2021). A

comprehensive way to access hospital death
predict ion model for Acute mesenteric ischemia:
a combinat ion of t radit ional stat istics and
machine learning. Internat ional Journal of
General Medicine, 14, 591.

[15] Huyut, M. T., & Üst ündağ, H. (2022). P

redict ion of diagnosis and prognosis of COVID-
19 disease by blood gas parameters using decision
t rees machine learning model: A ret rospect ive
observat ional study. Medical Gas Research,
12(2), 60.

7|Page