Expert knowledge for the recognition of leukemic cells

Rocio Ochoa-Montiel; Rocio Ochoa-Montiel; Gustavo Olague; Humberto Sossa; Humberto Sossa

doi:10.1364/AO.385208

Applied Optics
Vol. 59,
Issue 14,
pp. 4448-4460
(2020)
•https://doi.org/10.1364/AO.385208

Expert knowledge for the recognition of leukemic cells

Rocio Ochoa-Montiel, Gustavo Olague, and Humberto Sossa

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Rocio Ochoa-Montiel, Gustavo Olague, and Humberto Sossa, "Expert knowledge for the recognition of leukemic cells," Appl. Opt. 59, 4448-4460 (2020)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article
Editors' Pick

Check for updates

Related Topics
Table of Contents Category
- Imaging Systems, Image Processing, and Displays
Optics & Photonics Topics
?

The topics in this list come from the Optics and Photonics Topics applied to this article.

About this Article
History
- Original Manuscript: December 5, 2019
- Revised Manuscript: April 14, 2020
- Manuscript Accepted: April 15, 2020
- Published: May 8, 2020
Virtual Issues
Applied Optics Optics theory and practice in Iberoamerica (2020)

Abstract

This work shows the advantage of expert knowledge for leukemic cell recognition. In the medical area, visual analysis of microscopic images has regularly used biological samples to recognize hematological disorders. Nowadays, techniques of image recognition are needed to achieve an adequate identification of blood tissues. This paper presents a procedure to acquire expert knowledge from blood cell images. We apply Gaussian mixtures, evolutionary computing, and standard techniques of image processing to extract knowledge. This information feeds a support vector machine or multilayer perceptron to classify healthy or leukemic cells. Additionally, convolutional neural networks are used as a benchmark to compare our proposed method with the state of the art. We use a public database of 260 healthy and leukemic cell images. Results show that our traditional pattern recognition methodology matches deep learning accuracy since the recognition of blood cells achieves 99.63%, whereas the convolutional neural networks reach 97.74% on average. Moreover, the computational effort of our approach is minimal, while meeting the requirement of being explainable.

Full Article | PDF Article

More Like This

Automatic classification of acute and chronic myeloid leukemic cells with wide-angle label-free static cytometry

Linyan Xie, Qiao Liu, Changshun Shao, and Xuantao Su
Opt. Express 25(23) 29365-29373 (2017)

Deep learning-based light scattering microfluidic cytometry for label-free acute lymphocytic leukemia classification

Jing Sun, Lan Wang, Qiao Liu, Attila Tárnok, and Xuantao Su
Biomed. Opt. Express 11(11) 6674-6686 (2020)

Differentiation of normal and leukemic cells by 2D light scattering label-free static cytometry

Linyan Xie, Qiao Liu, Changshun Shao, and Xuantao Su
Opt. Express 24(19) 21700-21707 (2016)

Previous Article Next Article

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (12)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (9)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (24)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

1:	…
2:	while $C o n v e r g e n c e C r i t e r i o n = u n k n o w n$ do
3:	▹ $q_{t}$ defines a vector of the current vector population
4:	▹ $n v_{t}$ defines a vector of the new vector population
5:	for $t = 0; t < N_{p}; t + +$ do
6:	$r_{1} \leftarrow r a n d (N_{p})$
7:	▹ select a random index from $1, 2, \dots, N_{p}$
8:	$r_{2} \leftarrow r a n d (N_{p})$
9:	▹ select a random index from $1, 2, \dots, N_{p}$
10:	$r_{3} \leftarrow r a n d (N_{p})$
11:	▹ select a random index from $1, 2, \dots, N_{p}$
12:	$u_{t} \leftarrow q_{r 3} + F * (q_{r 1} - q_{r 2})$
13:	if $f (u_{t}) < f (q_{t})$ then $n v_{t} \leftarrow u_{t}$ ;
14:	else $n v_{t} \leftarrow q_{t}$ ;
15:	…

Subtype of Blood Cell	Cell Size (µm)	Nucleus	Chromatin
Neutrophil	10–15	2–5 lobes connected by thin filaments without visible chromatin	Coarsely clumped
Band neutrophil	10–15	Constricted and chromatin visible within the thinnest part	Coarsely clumped
Band neutrophil		Constricted and chromatin visible within the thinnest part	Coarsely clumped
Lymphocyte	7–18	Round to oval; may be slightly indented; occasional nucleoli	Condensed to deeply condensed
Monocyte	12–20	Variable; may be round horseshoe or kidney shaped; often has folds producing	Moderately clumped; lacy brain-like convolutions
Eosinophil	12–17	2–3 lobes connected by thin filaments without visible chromatin	Coarsely clumped
Basophil	10–14	Usually two lobes connected by thin filaments without visible chromatin	Coarsely clumped

Num. of Exec.	Without Image Data Augmentation				With Image Data Augmentation
	LeNet		AlexNet		LeNet		AlexNet
	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)
1	82.05	71.22	97.44	250.83	84.62	62.28	96.15	251.48
2	82.05	54.45	98.72	240.92	69.23	33.78	98.72	245.48
3	84.62	67.03	97.44	229.41	74.36	63.72	98.72	236.55
4	69.23	46.64	98.72	226.92	87.18	84.19	98.72	224.61
5	74.36	41.92	92.31	233.86	70.51	44.78	97.44	242.61
6	85.90	68.08	97.44	232.13	94.87	88.27	98.72	224.03
7	74.36	52.86	96.15	234.78	76.92	52.41	97.44	238.78
8	82.05	44.86	96.15	229.86	79.49	59.16	96.15	231.22
9	80.77	39.27	98.72	229.11	83.33	59.45	94.87	232.98
10	82.05	45.97	94.87	232.73	89.74	80.88	92.31	233.75
11	82.05	49.98	97.44	234.14	88.46	79.36	97.44	242.69
12	78.21	28.50	93.59	232.56	85.90	63.58	97.44	229.33
13	79.49	67.08	93.59	239.14	91.03	78.45	100	231.25
14	91.03	71.16	97.44	228.31	67.95	39.14	98.72	226.59
15	92.31	90.16	97.44	235.38	83.33	62.17	98.72	230.56
16	91.03	111.06	96.15	232.59	79.49	66.64	100.00	234.42
17	80.77	70.28	97.44	231.53	89.74	86.02	97.44	236.06
18	80.77	31.91	98.72	237.36	78.21	57.83	96.15	236.06
19	80.77	37.09	97.44	229.30	88.46	55.02	96.15	225.30
20	84.62	66.20	97.44	232.72	91.03	52.14	100.00	234.31
21	66.67	31.69	100.00	238.27	79.49	47.88	93.59	231.97
22	70.51	46.22	94.87	232.19	85.90	65.56	96.15	236.69
23	85.90	79.67	98.72	231.13	80.77	57.75	97.44	235.95
24	83.33	46.66	98.72	231.25	88.46	79.17	98.72	235.80
25	69.23	35.22	97.44	227.36	80.77	69.38	100.00	237.83
26	92.31	86.50	100.00	227.42	76.92	63.33	100.00	237.83
27	79.49	64.53	97.44	230.88	83.33	82.09	97.44	233.16
28	84.62	52.86	97.44	240.66	83.33	44.73	100.00	233.88
29	80.77	45.38	97.44	247.23	97.44	62.48	98.72	236.70
30	88.46	79.81	94.87	241.03	84.62	41.92	98.72	231.58
Average	81.32	57.48	97.05	234.03	83.16	62.78	97.74	234.70
$σ^{2}$	$\pm 44.85$		$\pm 3.42$		$\pm 51.32$		$\pm 3.70$
$σ$	$\pm 6.69$		$\pm 1.84$		$\pm 7.16$		$\pm 1.92$
Outlier detected?	No		No		No		No
Critical value of Z:	2.90		2.90		2.90		2.90

Num. of. Exec.	Images without Post-Processing				Images with Post-Processing
	LeNet		AlexNet		LeNet		AlexNet
	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)
1	66.67	56.45	92.31	271.63	58.97	63.44	85.90	251.23
2	75.64	57.02	92.31	271.83	53.85	49.47	92.31	249.13
3	73.08	56.08	89.74	259.75	57.69	34.86	89.74	238.31
4	74.36	71.03	87.18	256.72	65.38	69.50	93.59	253.14
5	76.92	62.38	89.74	266.78	44.87	52.23	88.46	246.19
6	66.67	42.20	87.18	263.55	52.56	43.50	85.90	242.05
7	70.51	54.50	85.90	258.45	52.56	34.67	92.31	245.14
8	71.79	51.98	91.03	261.92	47.44	41.27	88.46	237.70
9	62.82	39.47	88.46	261.86	48.72	52.50	88.46	246.30
10	80.77	63.23	84.62	258.38	51.28	34.09	85.90	243.38
11	56.41	54.03	88.46	262.64	52.56	52.14	85.90	241.31
12	67.95	47.59	91.03	249.52	60.26	51.64	85.90	240.88
13	71.79	60.56	92.31	258.97	56.41	33.14	89.74	240.78
14	67.95	38.72	88.46	264.14	62.82	48.81	92.31	244.55
15	64.10	52.88	87.18	263.48	51.28	58.64	84.62	232.00
16	78.21	61.39	88.46	259.63	57.69	36.58	91.03	237.41
17	76.92	66.16	91.03	263.55	61.54	70.06	87.18	230.63
18	73.08	46.66	92.31	264.45	53.85	42.50	87.18	229.81
19	66.67	50.94	85.90	260.53	53.85	53.28	83.33	237.08
20	60.26	45.02	88.46	259.67	60.26	46.28	94.87	233.48
21	73.08	44.39	92.31	264.17	51.28	63.89	85.90	236.11
22	61.54	33.14	85.90	259.95	67.95	51.52	89.74	240.80
23	76.92	47.44	94.87	258.59	57.69	30.47	89.74	234.20
24	67.95	64.97	91.03	263.64	56.41	47.91	92.31	229.22
25	65.38	66.66	89.74	262.73	60.26	66.95	92.31	234.64
26	67.95	43.03	89.74	261.48	50.00	37.63	87.18	232.81
27	71.79	35.44	91.03	260.66	55.13	32.92	84.62	237.98
28	70.51	53.78	91.03	258.55	69.23	55.98	91.03	235.16
29	70.51	45.36	89.74	254.39	55.13	55.38	89.74	236.28
30	50.00	44.34	83.33	268.34	57.69	55.73	91.03	237.22
Average	69.27	51.89	89.36	261.67	56.15	48.90	88.89	239.26
$σ^{2}$	$\pm 45.17$		$\pm 7.05$		$\pm 33.15$		$\pm 9.11$
$σ$	$\pm 6.72$		$\pm 2.65$		$\pm 5.75$		$\pm 3.01$
Outlier detected?	No		No		No		No
Critical value of Z:	2.90		2.90		2.90		2.90

Num. of. Exec.	Images without Post-Processing				Images with Post-Processing
	LeNet		AlexNet		LeNet		AlexNet
	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)	Acc.	Time(s)
1	70.51	99.67	89.74	260.53	57.69	54.52	74.36	262.61
2	65.38	36.50	92.31	253.77	50.00	60.56	69.23	264.02
3	83.33	50.58	91.03	246.89	50.00	46.48	84.62	253.14
4	75.64	39.00	91.03	250.27	51.28	35.09	75.64	249.59
5	78.21	40.94	88.46	260.77	50.00	48.08	79.49	257.95
6	76.92	84.27	87.18	247.25	56.41	42.38	74.36	246.11
7	69.23	37.03	92.31	251.38	50.00	68.38	79.49	249.98
8	76.92	36.13	87.18	247.53	51.28	57.28	73.08	247.19
9	52.56	35.16	89.74	250.42	47.44	84.69	82.05	251.78
10	66.67	30.63	92.31	246.02	51.28	50.27	75.64	251.22
11	82.05	72.44	93.59	244.22	53.85	43.91	71.79	248.02
12	76.92	75.81	85.90	245.48	48.72	41.33	84.62	246.05
13	73.08	46.16	94.87	246.22	51.28	37.55	83.33	252.30
14	66.67	47.91	91.03	251.28	50.00	63.38	73.08	252.39
15	71.79	32.06	82.05	247.91	53.85	54.73	82.05	247.78
16	71.79	43.02	91.03	247.30	50.00	35.05	79.49	245.61
17	85.90	49.06	93.59	246.58	60.26	35.23	78.21	248.75
18	73.08	47.78	94.87	244.78	50.00	44.64	76.92	254.14
19	83.33	39.28	92.31	248.88	57.69	39.16	79.49	254.92
20	74.36	44.34	91.03	248.75	57.69	65.95	80.77	253.56
21	60.26	46.66	94.87	247.80	50.00	45.23	76.92	252.02
22	51.28	35.66	87.18	249.44	50.00	68.52	75.64	249.56
23	53.85	37.27	91.03	251.20	51.28	56.91	76.92	249.72
24	76.92	41.98	83.33	250.63	53.85	57.36	79.49	259.58
25	65.38	62.38	85.90	252.25	50.00	39.84	74.36	255.80
26	73.08	42.28	93.59	247.48	50.00	49.33	80.77	252.41
27	75.64	44.61	80.77	247.44	50.00	38.50	79.49	253.28
28	71.79	57.80	92.31	246.94	69.23	54.91	88.46	255.14
29	71.79	36.08	91.03	244.22	48.72	29.06	79.49	258.34
30	79.49	49.17	84.62	245.19	50.00	33.97	74.36	251.02
Average	71.29	48.05	89.87	248.96	52.39	49.43	78.12	252.47
$σ^{2}$	$\pm 76.06$		$\pm 14.66$		$\pm 20.15$		$\pm 18.13$
$σ$	$\pm 8.72$		$\pm 3.82$		$\pm 4.48$		$\pm 4.25$
Outlier detected?	No		No		No		No
Critical value of Z:	2.90		2.90		2.90		2.90

No. Execution	Accuracy	Time (s)	No. Execution	Accuracy	Time (s)
1	93.94	351.16	16	100.00	341.14
2	93.94	343.88	17	96.67	339.11
3	100	341.27	18	100.00	335.78
4	100	340.23	19	100.00	341.44
5	100.00	338.34	20	100.00	343.56
6	100.00	341.84	21	100.00	339.95
7	96.67	341.36	22	100.00	339.95
8	100.00	337.31	23	100.00	339.39
9	96.97	342.84	24	100.00	339.28
10	100.00	344.05	25	100.00	338.02
11	100.00	340.83	26	100.00	339.05
12	100.00	340.33	27	100.00	339.92
13	100.00	342.47	28	96.67	341.59
14	100.00	338.30	29	100.00	335.81
15	100.00	340.41	30	100.00	341.27
Average:				99.19	340.85
$σ^{2}$				$\pm 3.12$
$σ$				$\pm 1.77$
Outlier detected?				Yes
Critical value of Z:				2.90

	MLP		SVM
No. Exec.	Accuracy	Time(s)	Accuracy	Time(s)
1	99.46	2.89	97.00	0.63
2	99.46	0.55	100.00	0.07
3	99.85	0.50	97.00	0.07
4	99.62	0.51	100.00	0.02
5	99.54	0.52	100.00	0.11
6	99.46	0.50	100.00	0.01
7	99.62	0.53	99.00	0.01
8	99.69	0.59	99.00	0.01
9	99.62	0.49	99.00	0.01
10	99.46	0.51	99.00	0.01
11	99.62	0.50	99.00	0.01
12	99.62	0.56	100.00	0.01
13	99.69	0.53	99.00	0.01
14	99.77	0.51	100.00	0.01
15	99.69	0.50	100.00	0.01
16	99.69	0.50	100.00	0.01
17	99.69	0.50	97.00	0.01
18	99.69	0.48	100.00	0.01
19	99.54	0.47	100.00	0.01
20	99.69	0.48	99.00	0.01
21	99.54	0.49	99.00	0.01
22	99.54	0.49	99.00	0.01
23	99.85	0.49	100.00	0.01
24	99.69	0.46	97.00	0.01
25	99.69	0.52	99.00	0.01
26	99.77	0.61	99.00	0.01
27	99.62	0.47	100.00	0.01
28	99.54	0.48	100.00	0.01
29	99.62	0.48	100.00	0.01
30	99.62	0.48	97.00	0.01
Average	99.63	0.59	99.00	0.04
$σ^{2}$	$\pm 0.016$		$\pm 0.00$
$σ$	$\pm 0.11$		$\pm 0.01$
Outlier detected?	No		No
Critical value of Z:	2.90		2.90

	With Data Augmentation			Without Data Augmentation
Classifier Used	Original images	Imgs. without post-proc.	Imgs. with post-proc.	Original images	Imgs. without post-proc.	Imgs. with post-proc.
LeNet	83.16	69.27	56.15	81.32	71.79	52.39
AlexNet	97.74	89.36	88.89	97.05	89.87	78.12
	Features extracted from images with post-processing
MLP	99.63
SVM	99.00

Abstract

Cited By

Figures (12)

Tables (9)

Equations (24)

Applied Optics