IP Library › Granted Patent US 12,725,261
Granted Patent B2
US 12,725,261 · App. 18/282,756 · Granted Sep 1, 2026

Convolutional long short-term memory networks for rapid medical image segmentation

Inventors: Piotr Slomka (Los Angeles, CA); Aditya Killekar (Beverly Hills, CA); Sebastien Cadet (Los Angeles, CA); Damini Dey (Los Angeles, CA)
Assignee: CEDARS-SINAI MEDICAL CENTER
G06T7/11G06T7/0012G06T7/62G06T2207/10081G06T2207/20081G06T2207/20084G06T2207/30096G06T2207/30101
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,725,261
App. No.
18/282,756
Granted
Sep 1, 2026
Kind
B2
Abstract

Convolutional long short-term memory (LSTM) networks are leveraged to segment and/or quantify medical imaging data. A multi-branch architecture makes use of an attention branch and a main branch. The main branch includes a dense block followed by a segmentation head, and is configured to consider a single input slice of the imaging data. This main branch is able to segment larger and easier-to-classify targets. The attention branch, however, makes use of a sequential processor that includes a convolutional LSTM (ConvLSTM) followed by a segmentation head and an attention head. This attention branch is configured to process an input slice and additional adjacent slices. This attention branch is able to segment smaller and more difficult to classify targets.

Claims (60)

1 . A system, comprising:

one or more data processors; and

a non-transitory computer-readable storage medium containing instructions which, when executed on the one or more data processors, cause the one or more data processors to perform operations including:

receiving medical imaging data containing a plurality of ordered image slices;

accessing a multi-branch model associated with a target, wherein the multi-branch model includes a main branch and an attention branch, wherein the main branch includes a densely connected convolutional network (DenseNet) followed by a first segmentation head, wherein the attention branch includes a convolutional long short-term memory network (ConvLSTM) feeding into a second segmentation head and an attention head, wherein the multi-branch model is trained to receive sequential image slices and, for each sequential image slice, output a target mask indicative of one or more target regions within the sequential image slice identified as being the target, wherein a first output of the first segmentation head, a second output of the second segmentation head, and a third output of the attention head are used to generate the target mask;

providing the plurality of ordered image slices to the multi-branch model; and

generating, by the multi-branch model, a plurality of output target masks in response to providing the plurality of ordered image slices to the multi-branch model.

2 . The system of claim 1 , wherein the operations further comprise generating a quantitative score using the plurality of output target masks, wherein the quantitative score is indicative of a severity of a condition associated with the target.

3 . The system of claim 2 , wherein generating the quantitative score includes:

calculating a total target volume using the plurality of output target masks; and

generating a quantitative score using the total target volume.

4 . The system of claim 3 , wherein generating the quantitative score includes at least one of:

i) generating a coronary artery calcium score, wherein the target is calcium, and wherein the plurality of ordered image slices is associated with a computed tomography attenuation correction scan;

ii) generating a pneumonia burden score, wherein the target includes at least one of a ground-glass opacity lesion and a high opacity lesion, and wherein the plurality of ordered image slices is associated with a computed tomography study of a pleural cavity;

iii) generating a plaque volume measurement, wherein the target is plaque buildup, and wherein the plurality of ordered image slices is associated with a coronary computed tomography angiography study; or

iv) generating a stenosis severity score, wherein the target is plaque buildup, and wherein the plurality of ordered image slices is associated with a coronary computed tomography angiography study.

5 . The system of claim 1 , wherein, for each image slice of the plurality of ordered image slices, generating the plurality of output target masks includes:

generating a main branch output by performing elementwise multiplication of the first output and the third output;

generating an attention branch output by performing elementwise multiplication of the second output and the third output; and

generating a respective output target mask by performing elementwise addition of the main branch output and the attention branch output.

6 . The system of claim 1 , wherein each of the first segmentation head, the second segmentation head, and the attention head includes a 3×3 convolutional layer followed by a batch layer, followed by a leaky rectified linear unit layer, followed by an additional 3×3 convolutional layer, followed by an additional batch layer, followed by an additional leaky rectified linear unit layer, followed by a 1×1 convolutional layer, and wherein the attention head further includes a sigmoid layer following the 1×1 convolutional layer.

7 . The system of claim 1 , wherein the operations further comprise:

accessing an additional multi-branch model, wherein the multi-branch model is trained for segmentation according to the target, and wherein the additional multi-branch model is trained for segmentation according to an additional target;

providing the plurality of ordered image slices to the additional multi-branch model;

generating, by the additional multi-branch model, a plurality of output additional target masks in response to providing the plurality of ordered image slices to the additional multi-branch model, wherein the plurality of output additional target masks are indicative of one or more additional target regions within the plurality of ordered image slices identified as being the additional target; and

applying the plurality of output additional target masks to the plurality of output target masks to generate a quantitative score, wherein the quantitative score is indicative of a severity of a condition associated with the one or more targets.

8 . The system of claim 1 , wherein the operations further comprise: training the multi-branch model using a set of training data, wherein the set of training data includes a plurality of manually annotated image sets, wherein each of the plurality of manually annotated image sets includes a plurality of ordered, annotated image slices, and wherein each respective image slice of the plurality of ordered, annotated image slices includes at least one manually-applied annotation indicative of an identified target region within the respective image slice.

9 . The system of claim 1 , wherein the operations further comprise: presenting the medical imaging data using a display device, wherein presenting the medical imaging data using the display device includes applying a visually distinguishable feature to the medical imaging data based on the plurality of output target masks such that the one or more target regions visible in the medical imaging data are visually distinguishable from other regions within the medical imaging data.

10 . A computer-implemented method, comprising:

receiving medical imaging data containing a plurality of ordered image slices;

accessing a multi-branch model associated with a target, wherein the multi-branch model includes a main branch and an attention branch, wherein the main branch includes a densely connected convolutional network (DenseNet) followed by a first segmentation head, wherein the attention branch includes a convolutional long short-term memory network (ConvLSTM) feeding into a second segmentation head and an attention head, wherein the multi-branch model is trained to receive sequential image slices and, for each sequential image slice, output a target mask indicative of one or more target regions within the sequential image slice identified as being the target, wherein a first output of the first segmentation head, a second output of the second segmentation head, and a third output of the attention head are used to generate the target mask;

providing the plurality of ordered image slices to the multi-branch model; and

generating, by the multi-branch model, a plurality of output target masks in response to providing the plurality of ordered image slices to the multi-branch model.

11 . The computer-implemented method of claim 10 , further comprising generating a quantitative score using the plurality of output target masks, wherein the quantitative score is indicative of a severity of a condition associated with the target.

12 . The computer-implemented method of claim 11 , wherein generating the quantitative score includes:

calculating a total target volume using the plurality of output target masks; and

generating a quantitative score using the total target volume.

13 . The computer-implemented method of claim 12 , wherein generating the quantitative score includes at least one of:

i) generating a coronary artery calcium score, wherein the target is calcium, and wherein the plurality of ordered image slices is associated with a computed tomography attenuation correction scan;

ii) generating a pneumonia burden score, wherein the target includes at least one of a ground-glass opacity lesion and a high opacity lesion, and wherein the plurality of ordered image slices is associated with a computed tomography study of a pleural cavity;

iii) generating a plaque volume measurement, wherein the target is plaque buildup, and wherein the plurality of ordered image slices is associated with a coronary computed tomography angiography study; or

iv) generating a stenosis severity score, wherein the target is plaque buildup, and wherein the plurality of ordered image slices is associated with a coronary computed tomography angiography study.

14 . The computer-implemented method of claim 10 , wherein, for each image slice of the plurality of ordered image slices, generating the plurality of output target masks includes:

generating a main branch output by performing elementwise multiplication of the first output and the third output;

generating an attention branch output by performing elementwise multiplication of the second output and the third output; and

generating a respective output target mask by performing elementwise addition of the main branch output and the attention branch output.

15 . The computer-implemented method of claim 10 , wherein each of the first segmentation head, the second segmentation head, and the attention head includes a 3×3 convolutional layer followed by a batch layer, followed by a leaky rectified linear unit layer, followed by an additional 3×3 convolutional layer, followed by an additional batch layer, followed by an additional leaky rectified linear unit layer, followed by a 1×1 convolutional layer, and wherein the attention head further includes a sigmoid layer following the 1×1 convolutional layer.

16 . The computer-implemented method of claim 10 , further comprising:

accessing an additional multi-branch model, wherein the multi-branch model is trained for segmentation according to the target, and wherein the additional multi-branch model is trained for segmentation according to an additional target;

providing the plurality of ordered image slices to the additional multi-branch model;

generating, by the additional multi-branch model, a plurality of output additional target masks in response to providing the plurality of ordered image slices to the additional multi-branch model, wherein the plurality of output additional target masks are indicative of one or more additional target regions within the plurality of ordered image slices identified as being the additional target; and

applying the plurality of output additional target masks to the plurality of output target masks to generate a quantitative score, wherein the quantitative score is indicative of a severity of a condition associated with the one or more targets.

17 . The computer-implemented method of claim 10 , further comprising: training the multi-branch model using a set of training data, wherein the set of training data includes a plurality of manually annotated image sets, wherein each of the plurality of manually annotated image sets includes a plurality of ordered, annotated image slices, and wherein each respective image slice of the plurality of ordered, annotated image slices includes at least one manually-applied annotation indicative of an identified target region within the respective image slice.

18 . The computer-implemented method of claim 10 , further comprising: presenting the medical imaging data using a display device, wherein presenting the medical imaging data using the display device includes applying a visually distinguishable feature to the medical imaging data based on the plurality of output target masks such that the one or more target regions visible in the medical imaging data are visually distinguishable from other regions within the medical imaging data.

19 . A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause a data processing apparatus to perform operations including:

receiving medical imaging data containing a plurality of ordered image slices;

accessing a multi-branch model associated with a target, wherein the multi-branch model includes a main branch and an attention branch, wherein the main branch includes a densely connected convolutional network (DenseNet) followed by a first segmentation head, wherein the attention branch includes a convolutional long short-term memory network (ConvLSTM) feeding into a second segmentation head and an attention head, wherein the multi-branch model is trained to receive sequential image slices and, for each sequential image slice, output a target mask indicative of one or more target regions within the sequential image slice identified as being the target, wherein a first output of the first segmentation head, a second output of the second segmentation head, and a third output of the attention head are used to generate the target mask;

providing the plurality of ordered image slices to the multi-branch model; and

generating, by the multi-branch model, a plurality of output target masks in response to providing the plurality of ordered image slices to the multi-branch model.

20 . The computer-program product of claim 19 , wherein the operations further comprise: presenting the medical imaging data using a display device, wherein presenting the medical imaging data using the display device includes applying a visually distinguishable feature to the medical imaging data based on the plurality of output target masks such that the one or more target regions visible in the medical imaging data are visually distinguishable from other regions within the medical imaging data.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 18, 2023
From: SLOMKA, PIOTR; KILLEKAR, ADITYA; CADET, SEBASTIEN; DEY, DAMINI
To: CEDARS-SINAI MEDICAL CENTER
Reel/Frame 064941/0379 →
Continuity (5)
Provisional Application 63301222 · Jan 20, 2022
Provisional Application 63272081 · Oct 26, 2021
Provisional Application 63249354 · Sep 28, 2021
Provisional Application 63163466 · Mar 19, 2021
Related Publication 20250111516A1 · Apr 3, 2025
References Cited (74)
US 10579865B2 · Yoo · 2020 [cited by examiner]
US 20180373924A1 · Yoo · 2018 [cited by examiner]
US 20190130562A1 · Liu et al. · 2019 [cited by applicant]
US 20200167929A1 · Wang · 2020 [cited by applicant]
US 20210006755A1 · Kim · 2021 [cited by examiner]
US 20210398676A1 · Evans · 2021 [cited by examiner]
US 20220005168A1 · Kim · 2022 [cited by examiner]
US 20220384045A1 · Zimmerman · 2022 [cited by examiner]
US 20230196567A1 · Aykut · 2023 [cited by examiner]
WO 2021050765A1 · 2021 [cited by applicant]
International Search Report in International Patent Application No. PCT/US2022/020954 mailed Jun. 1, 2022, 2 pp. [cited by applicant]
Written Opinion in International Patent Application No. PCT/US2022/020954 mailed Jun. 1, 2022, 9 pp. [cited by applicant]
Ramanan, R.V. et al., “Incidental chest computed tomography findings in asymptomatic covid-19 patients. A multicentre indian perspective,” The Indian Journal of Radiology & Imaging, vol. 31, No. Suppl 1, p. S45, 2021. [cited by applicant]
Rubin, G.D. et al., “The Role of Chest Imaging in Patient Management during the COVID-19 Pandemic: A Multinational Consensus Statement from the Fleischner Society.” Radiology. 2020;296(1):172-180. [cited by applicant]
Saood A. and I. Hatem, COVID-19 lung CT image segmentation using deep learning methods: U-Net versus SegNet. BMC Medical Imaging, 2021. 21(1): p. 19. [cited by applicant]
Sharifpour, M. et al., “C-Reactive protein as a prognostic indicator in hospitalized patients with COVID-19.” PLOS ONE. 2020;15(11):e0242400. [cited by applicant]
Shi, X., et al., Convolutional LSTM network: A machine learning approach for precipitation now casting. arXiv preprint arXiv: 1506.04214, 2015. [cited by applicant]
Tao, A., K. Sapra, and B. Catanzaro, Hierarchical multi-scae attention for semantic segmentation. arXiv preprint arXiv: 2005.10821, 2020. [cited by applicant]
Wang, G., et al.,, A Noise-Robust Framework for Automatic Segmentation of COVID-19 Pneumonia Lesions From CT Images. IEEE Trans Med Imaging, 2020. 39(8): p. 2653-2663. [cited by applicant]
Wang, Y., et al., Temporal Changes of CT Findings in 90 Patients with COVID-19 Pneumonia: A longitudinal study. Radiology, 2020. 296(2): p. E55-e64. [cited by applicant]
Xu, B., et al., Empirical evaluation of rectified activation in convolutional network. arXiv preprint arXiv: 1505.00853, 2015. [cited by applicant]
Yang, R., et al., Chest CT Severity Score: An Imaging Tool for Assessing Severe COVID-19. Radiology: Cardiothoracic Imaging, 2020. 2(2): p. e200047. [cited by applicant]
Zhang, K., et al., Clinically Aplicable AI System for Accurate Diagnosis, Quantitative Measurements, and Prognosis of Covid-19 Pneumonia Using Computed Tomography. Cell, 2020. 181(6): p. 1423-1433.e11. [cited by applicant]
Zhao, S., et al., Region mutual information loss for semantic segmentation. arXiv preprint arXiv: 1910.12037, 2019. [cited by applicant]
Ai, T., et al., Correlation of Chest CT and RT-PCR Testing for Coronavirus Disease 2019(COVID-19) in China: A Report of 1014 Cases. Radiology, 2020. 296(2): p. E32-E40. [cited by applicant]
Albano, D., et al., “Incidental findings suggestive of covid-19 in asymptomatic patients undergoing nuclear medicine procedures in a high-prevalence region,” Journal of Nuclear Medicine, vol. 61, No. 5, pp. 632-636, 202… [cited by applicant]
Ashari, M.A. et al., “Strategies for radiology departments in handling the COVID-19 pandemic.” Diagn Interv Radiol. 2020;26(4):296-300. [cited by applicant]
Bai, H.X., et al., Artifical Intelligence Augmentation of Ragiologist Performance in Distinguishing COVID-19 from Pneumonia of Other Origin at Chest CT. Radiology, 2020. 296(3): p. E156-E165. [cited by applicant]
Bernheim, A., et al., Chest CT Findings in Coronavirus Disease-19 (COVID-19): Relationship to Duration of Infection. Radiology, 2020. 295(3): p. 200463. [cited by applicant]
Boger, B., et al., Systematic review with meta-analysis of the accuracy of diagnostic tests for COVID-19. Am J Infect Control, 2021. 49(1): p. 21-29. [cited by applicant]
Chaganti, S., et al., Automated Quantification of CT Patterns Associated with COVID-19 from Chest CT. Radiology: Aritificial Intelligence, 2020. 2(4): p. e20048. [cited by applicant]
Chang, Y-.C. et al., “Pulmonary Sequelae in Convalescent Patients after Severe Acute Respiratory Syndrome: Evaluation with Thin-Section CT.” Radiology. 2005;236(3):1067-1075. [cited by applicant]
Chartrand, G. et al., “Deep Learning: A Primer for Radiologists.” RadioGraphics. 2017;37(7):2113-2131. [cited by applicant]
Cicek, O., et al., 3D U-Net: learning dense volumetric segmentation from sparse annotation. in International conference on medical image computing and computer-assissted intervention. 2016. Springer. [cited by applicant]
Colombi, D. et al. “Well-aerated Lung on Admitting Chest CT to Predict Adverse Outcome in COVID-19 Pneumonia.” Radiology. 2020;296(2):E86-e96. [cited by applicant]
Commandeur, F., et al., Deep Learning for Quantification of Epicardial and Thoracic Adipose Tissue From Non-Contrast CT. IEEE Trans Med Imaging, 2018. 37(8): p. 1835-1846. [cited by applicant]
De Gonzalez, A.B., et al., “Projected cancer risks from computed tomographic scans performed in the united states in 2007,” Archives of internal medicine, vol. 169, No. 22, pp. 2071-2077, 2009. [cited by applicant]
DeLong, E.R. et al., “Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.” Biometrics. 1988;44(3):837-845. [cited by applicant]
Deng, J. et al., “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition, pp. 248-255, IEEE, 2009. [cited by applicant]
Extended European Search Report in Europan Patent Application No. 22772290.7, mailed Dec. 19, 2024 (8 pages). [cited by applicant]
Fan, D.-P. et al., “Inf-net: Automatic covid-19 lung infection segmentation from ct images,” IEEE Transactions on Medical Imaging, vol. 39, No. 8m pp. 2626-2637, 2020. [cited by applicant]
Francone, M., et al., Chest CT score in COVID-19 patients: correlatoin with disease severity and short-term prognisis. Eur Radiol, 2020. 30(12): p. 6808-6817. [cited by applicant]
Gieraerts, C., et al., Prognostic Value and Reproductibility of AI -assisted Analysis of Lung Involvement in COVID-19 on Low-Dose Submilisievert Chest CT: Sample Size Implications for Clinical Trials. Radiology: Cardiot… [cited by applicant]
Gietema, H.A. et al., “Quantifying the Extent of Emphysema:: Factors Associated with Radiologists' Estimations and Quantitative Indices of Emphysema Severity Using the ECLIPSE Cohort.” Academic Radiology. 2011;18(6):661… [cited by applicant]
Glorot, X. et al., “Understanding the difficulty of training deep feedforward neural networks,” in Proceedings of the thirteenth international conference on artificial intelligence and statistics, pp. 249-256, JMLR Work… [cited by applicant]
Grodecki, K. et al., “Rapid quantification of COVID-19 pneumonia burden from computed tomography with convolutional LSTM networks.” ArXiv. 2021:arXiv:2104.00138v00131. [cited by applicant]
Grodecki, K., et al., Epicardial adipose tissue is assicoated with extent of pneumonia and adverse outcomes in patients with COVID-19. Metabolism, 2021. 115L p. 154436. [cited by applicant]
Grodecki, K., et al., Quantitative Burden of COVID-19 Pneumonia on Chest CT Predicts Adverse Outcomes: A Post-Hoc Analysis of a Prospective International Registry. Radiology: Cardiothoracic Imaging, 2020. 2(5): p. e2003… [cited by applicant]
Guan, X. et al., “Quantitative and semi-quantitative CT assessments of lung lesion burden in COVID-19 pneumonia.” Scientific Reports. 2021;11(1):5148. [cited by applicant]
Guendel, S. et al. “Multi-task learning for chest x-ray abnormality classification on noisy labels.” arXiv preprint arXiv:1905.06362 (2019). [cited by applicant]
Habouzit, V. et al., “Incidental finding of covid-19 lung infection in 18f-fdg pet/ct: What should we do?,” Clinical nuclear medicine, 2020. [cited by applicant]
Hansell, D.M. et al., “Society: glossary of terms for thoracic imaging.” Radiology. 2008;246(3):697-722. [cited by applicant]
He, K., et al., Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. in Proceedings of the IEEE international conference on computer vision. 2015. [cited by applicant]
Henry, B.M. et al., “Lactate dehydrogenase levels predict coronavirus disease 2019 (COVID-19) severity and mortality: A pooled analysis.” Am J Emerg Med. 2020;38(9):1722-1726. [cited by applicant]
Huang, G., et al. Densely connected convolutional networks. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2017. [cited by applicant]
Kazerooni, E.A. et al. “Thin-section CT obtained at 10-mm increments versus limited three-level thin-section CT for idiopathic pulmonary fibrosis: correlation with pathologic scoring.” AJR Am J Roentgenol. 1997;169(4):9… [cited by applicant]
Khatami, F., et al., A meta-analysis of accuracy and sensitivity of chest CT and RT-PCR in COVID-19 diagnosis. Sci Rep, 2020. 10(1): p. 22402. [cited by applicant]
Kingma, D.P. et al., “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014. [cited by applicant]
Krizhevsky, A., I. Sutskever, and G.E. Hinton, Imagenet classification with deep convolutional nerual networks. Advances in neural information processing systems, 2012. 25: p. 1097-1105. [cited by applicant]
Landis, J.R. et al., “The Measurement of Observer Agreement for Categorical Data.” Biometrics. 1977;33(1):159-174. [cited by applicant]
Ledig, C. et al., “Photo-realistic single image super-resolution using a generative adversarial network,” in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4681-4690, 2017. [cited by applicant]
Li, K., et al., CT image visual quantitative evaluation and clinical classification of coronavirus disease (COVID-19)). Eur Radiol, 2020. 30(8): p. 4407-4416. [cited by applicant]
Li, L., et al., Using Artificial Intellifence to Detect COVID-19 and Community-acquired Pneumonia Based on Pulmonary CT: Evaluation of the Diagnostic Accuracy. Radiology, 2020. 296(2): p. E65-E-71. [cited by applicant]
Li, S. et al., “Predictive value of chest CT scoring in COVID-19 patients in Wuhan, China: A retrospective cohort study.” Respiratory Medicine. 2021; 176:106271. [cited by applicant]
Lin, T.-Y. et al., “Focal loss for dense object detection,” in Proceedings of the IEEE international confrence on computer vision, pp. 2980-2988, 2017. [cited by applicant]
Meyes, R., “Ablation studies in artificial neural networks,” arXiv preprint arXiv:1901.08644, 2019. [cited by applicant]
Milletari F., N. Navab, and S.-A. Ahmadi. V-net: Fully convolutional neural networks for volumetric medical image segmentation. In 2016 fourth international conference on 3D vision (3DV). 2016. IEEE. [cited by applicant]
Morozov, S. et al., “Mosmeddata: Chest ct scans with covid-19 related findings dataset,” arXiv preprint arXiv:2005.06465, 2020. [cited by applicant]
Neveu, S. et al., “Incidental diagnosis of covid-19 pneumonia on chest computed tomography,” Diagnostic and interventional imaging, vol. 101, No. 7-8, pp. 457-461, 2020. [cited by applicant]
Nunes, N. et al., “A multi-modal deep learning method for classifying chest radiology exams.” Advances in Databases and Information Systems; (Lecture Notes in Computer Science; Lect. Notes Computer), Springer Internatio… [cited by applicant]
Pallardy, A. et al., “Incidental findings suggestive of covid-19 in asymptomatic cancer patients undergoing 18f-fdg pet/ct in a low prevalence region,” European Journal of Nuclear Medicine and Molecular Imaging, vol. 48… [cited by applicant]
Pan, F., et al., Time Course of Lung Changes at Chest CR during Recovery from Coronavirus Disease 2019 (COVID-19). Radiology, 2020. 295(3): p. 715-721. [cited by applicant]
Pencina, M.J. et al., Extensions of net reclassification improvement calculations to measure the usefulness of new biomarkers. Stat Med. 2011;30(1):11-21. [cited by applicant]
Pfeuffer, A. et al., “Semantic segmentation of video sequences with convolutional Istms,” in 2019 IEEE Intelligent Vehicles Symposium (IV), pp. 1441-1447, IEEE, 2019. [cited by applicant]