World J Oncol

World Journal of Oncology, ISSN 1920-4531 print, 1920-454X online, Open Access

Article copyright, the authors; Journal compilation copyright, World J Oncol and Elmer Press Inc

Journal website https://www.wjon.org

Original Article

Volume 15, Number 1, February 2024, pages 81-89

Application of Machine Learning and Deep EfficientNets in Distinguishing Neonatal Adrenal Hematomas From Neuroblastoma in Enhanced Computed Tomography Images

Figure 2. Radiomics feature selection with the least absolute shrinkage and selection operator (LASSO) regression model. (a) The vertical lines indicate that the optimal value of the LASSO tuning parameter λ is 0.0168. (b) LASSO coefficient profile plot with different log (λ) was shown. Nine radiomic features with non-zero coefficients were selected.

Figure 3. The receiver operating characteristic (ROC) curves of the support vector machine (SVM), random forest, Extra Trees, gradient boosting, AdaBoost, and multi-layer perceptron (MLP) in the test cohorts.

Figure 4. Performance of the lesion classification model trained with the abdominal computed tomography (CT) images. (a) Confusion matrix was drawn using the test dataset (0: neonatal adrenal hematomas; 1: neonatal adrenal neuroblastoma). (b) Image evaluation by the receiver operating characteristic (ROC) curve was drawn for neuroblastoma. Area under the curve (AUC) was calculated.

Figure 5. Computed tomography (CT) of the neonatal adrenal neuroblastoma and hematoma: original image (a, c) and heat map (b, d).

**Table 1.** Clinical Characteristics in the Cohorts
Variable	Adrenal hematomas (N = 36)	Adrenal neuroblastoma (N = 40)	P-value
*P < 0.05. Continuous data are presented as median (interquartile range). Categorical data are presented as numbers (%). NSE: neuron-specific enolase; LDH: lactate dehydrogenase.
Time onset
Prenatal	4 (11.1%)	30 (75.0%)	< 0.001*
Postnatal	32 (88.9%)	10 (25.0%)
Sex
Male	19 (52.8%)	25 (62.5%)	0.391
Female	17 (47.2%)	15 (37.5%)
Localization
Left	5 (13.9%)	15 (37.5%)	0.003*
Right	31 (86.1%)	21 (52.5%)
Bilateral	0 (0%)	4 (10.0%)
Weight (kg)	3.56 (3.06 - 3.98)	4.50 (3.65 - 6.50)	0.001*
NSE (ng/mL)	31.95 (22.37 - 42.17)	45.65 (26.77 - 72.30)	0.037*
Ferritin (µg/L)	573.95 (359.85 - 769.23)	318.90 (216.10 - 426.00)	< 0.001*
LDH (U/L)	667.00 (441.50 - 881.50)	299.00 (264.00 - 448.50)	< 0.001*

Table 1. Clinical Characteristics in the Cohorts

Variable

Adrenal hematomas (N = 36)

Adrenal neuroblastoma (N = 40)

P-value

*P < 0.05. Continuous data are presented as median (interquartile range). Categorical data are presented as numbers (%). NSE: neuron-specific enolase; LDH: lactate dehydrogenase.

Time onset

Prenatal

4 (11.1%)

30 (75.0%)

< 0.001*

Postnatal

32 (88.9%)

10 (25.0%)

Sex

Male

19 (52.8%)

25 (62.5%)

0.391

Female

17 (47.2%)

15 (37.5%)

Localization

Left

5 (13.9%)

15 (37.5%)

0.003*

Right

31 (86.1%)

21 (52.5%)

Bilateral

0 (0%)

4 (10.0%)

Weight (kg)

3.56 (3.06 - 3.98)

4.50 (3.65 - 6.50)

0.001*

NSE (ng/mL)

31.95 (22.37 - 42.17)

45.65 (26.77 - 72.30)

0.037*

Ferritin (µg/L)

573.95 (359.85 - 769.23)

318.90 (216.10 - 426.00)

< 0.001*

LDH (U/L)

667.00 (441.50 - 881.50)

299.00 (264.00 - 448.50)

< 0.001*

**Table 2.** Diagnostic Performance of Different Models in Training and Test Cohorts
Model name	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
CI: confidence interval; MLP: multi-layer perceptron; SVM: support vector machine; RF: random forest; AUC: area under the curve; NPV: negative prediction value; PPV: positive prediction value.
SVM
Training	0.967 (0.900 - 1.00)	0.974	0.944	1.000	1.000	0.952
Test	0.985 (0.958 - 1.00)	0.946	1.000	0.895	0.900	1.000
RF
Training	1.000 (1.000 - 1.000)	1.000	1.000	1.000	1.000	1.000
Test	0.937 (0.865 - 1.000)	0.892	0.778	1.000	1.000	0.826
Extra Trees
Training	1.000 (1.000 - 1.000)	1.000	1.000	1.000	1.000	1.000
Test	0.958 (0.905 - 1.000)	0.892	0.778	1.000	1.000	0.826
Gradient boosting
Training	1.000 (1.000 - 1.000)	1.000	1.000	1.000	1.000	1.000
Test	0.827 (0.699 - 0.956)	0.838	0.889	1.000	0.800	0.882
AdaBoost
Training	0.967 (0.900 - 1.00)	1.000	1.000	1.000	1.000	1.000
Test	0.882 (0.771 - 0.993)	0.838	0.778	0.895	0.875	0.810
MLP
Training	0.969 (0.921 - 1.000)	0.921	0.944	0.900	0.895	0.947
Test	0.971 (0.929 - 1.000)	0.919	0.833	1.000	1.000	0.864

Table 2. Diagnostic Performance of Different Models in Training and Test Cohorts

Model name

AUC (95% CI)

Accuracy

Sensitivity

Specificity

PPV

NPV

CI: confidence interval; MLP: multi-layer perceptron; SVM: support vector machine; RF: random forest; AUC: area under the curve; NPV: negative prediction value; PPV: positive prediction value.

SVM

Training

0.967 (0.900 - 1.00)

0.974

0.944

1.000

0.952

Test

0.985 (0.958 - 1.00)

0.946

1.000

0.895

0.900

1.000

Training

1.000 (1.000 - 1.000)

1.000

Test

0.937 (0.865 - 1.000)

0.892

0.778

1.000

0.826

Extra Trees

Training

1.000 (1.000 - 1.000)

1.000

Test

0.958 (0.905 - 1.000)

0.892

0.778

1.000

0.826

Gradient boosting

Training

1.000 (1.000 - 1.000)

1.000

Test

0.827 (0.699 - 0.956)

0.838

0.889

1.000

0.800

0.882

AdaBoost

Training

0.967 (0.900 - 1.00)

1.000

Test

0.882 (0.771 - 0.993)

0.838

0.778

0.895

0.875

0.810

MLP

Training

0.969 (0.921 - 1.000)

0.921

0.944

0.900

0.895

0.947

Test

0.971 (0.929 - 1.000)

0.919

0.833

1.000

0.864