A Fluorescence Lifetime Retrieval Algorithm Based on LSTM Neural Network

Biyu YANG; Yue XU

doi:10.3788/gzxb20225106.0618002

Journals >Acta Photonica Sinica >Volume 51 >Issue 6 >Page 0618002 > Article

Acta Photonica Sinica
Vol. 51, Issue 6, 0618002 (2022)

A Fluorescence Lifetime Retrieval Algorithm Based on LSTM Neural Network

Biyu YANG¹ and Yue XU^1、2、*

Author Affiliations

¹College of Integrated Circuit Science and Engineering，Nanjing University of Posts and Telecommunications，Nanjing 210023，China

²National and Local Joint Engineering Laboratory of RF Integration and Micro-assembly Technology，Nanjing 210023，China

show less

DOI: 10.3788/gzxb20225106.0618002 Cite this Article

Biyu YANG, Yue XU. A Fluorescence Lifetime Retrieval Algorithm Based on LSTM Neural Network[J]. Acta Photonica Sinica, 2022, 51(6): 0618002 Copy Citation Text

show less

Fig. 1. TCSPC method for fluorescence lifetime detection

Download full size | View in the Article

Fig. 2. Flow chart of the construction of the LSTM neural network model

Download full size | View in the Article

Fig. 3. Structure of LSTM

Download full size | View in the Article

Fig. 4. Test results of the same neurons with the different layers

Download full size | View in the Article

Fig. 5. Test results of the different neurons with the same layers

Download full size | View in the Article

Fig. 6. The structure of the network

Download full size | View in the Article

Fig. 7. Histogram of fluorescence lifetime decay

Download full size | View in the Article

Fig. 8. The restoration results of 32×32 array

Download full size | View in the Article

Fig. 9. Comparison of CMM，LSM and LSTM retrieval range

Download full size | View in the Article

Fig. 10. The comparison of array imaging when

τ_{1}

=5 ns，

τ_{2}

=10 ns，

τ_{3}

=15 ns

Download full size | View in the Article

Fig. 11. The comparison of array imaging when

τ_{1}

=40 ns，

τ_{2}

=60 ns，

τ_{3}

=80 ns

Download full size | View in the Article

Theoretical fluorescence lifetime τ/ns	Retrieval algorithm	Retrieval value/ns	R_acc/%	Standard deviation/ps
5	CMM	5.46	90.80%	88.40
	LSM	5.47	90.60%	186.20
	LSTM	4.90	98.00%	98.26
10	CMM	10.87	91.30%	158.08
	LSM	10.98	90.20%	355.92
	LSTM	10.12	98.80%	181.25
40	CMM	31.36	78.40%	302.40
	LSM	45.24	86.90%	1 892.79
	LSTM	41.46	96.32%	1 383.94
80	CMM	38.89	48.61%	320.77
	LSM	92.13	84.84%	4 892.03
	LSTM	81.65	97.94%	3 392.69
90	CMM	39.73	44.14%	332.59
	LSM	103.63	84.86%	5 540.91
	LSTM	86.87	96.52%	2 794.33

Table 1. Comparison of CMM、LSM and LSTM algorithms for retrieving single point when the photon count is 5 000

View in the Article

Theoretical fluorescence lifetime τ/ns	Retrieval algorithm	Retrieval value/ns	R_acc/%	Standard deviation/ps
5	CMM	5.45	91.00%	54.20
	LSM	5.39	92.20%	93.78
	LSTM	5.09	98.20%	61.48
10	CMM	10.73	92.70%	122.72
	LSM	10.81	91.90%	231.10
	LSTM	9.89	98.90%	106.12
40	CMM	30.36	75.90%	229.64
	LSM	40.97	97.58%	834.33
	LSTM	40.71	98.23%	831.88
80	CMM	39.57	49.46%	218.26
	LSM	90.36	87.05%	2 495.16
	LSTM	80.18	99.78%	2 263.88
90	CMM	40.56	45.07%	236.90
	LSM	101.62	87.09%	3 088.81
	LSTM	87.72	97.47%	2 301.81

Table 2. Comparison of CMM、LSM and LSTM algorithms for retrieving single point when the photon count is 10 000

Biyu YANG, Yue XU. A Fluorescence Lifetime Retrieval Algorithm Based on LSTM Neural Network[J]. Acta Photonica Sinica, 2022, 51(6): 0618002

Download Citation

Tools

Save the article for my favorites

Paper Information