Periodic Reporting for period 1 - DL-MechanoPredictor (Deep learning derived mechanical biomarkers for cancer therapy prediction)
Berichtszeitraum: 2022-09-01 bis 2024-02-29
Tumors are highly heterogeneous and differ considerably not only between tumors of different types but also among tumors of the same type. Therefore, therapeutic outcomes vary among patients with cancer separating patients between responders and non-responders to the treatment. The prediction of a patient’s response to treatment is of high importance and priority. Here, we propose the use of deep learning methods to develop imaging biomarkers based on ultrasound Shear Wave Elastography (SWE) imaging. This biomarker will aim to predict the patient’s response to cancer therapy. Furthermore, we propose to develop and test in the clinical setting a software product that will communicate with ultrasound systems to be used as a commercial tool for the measurement of the deep learning-derived SWE biomarker. This marker will be able to separate tumors to responders and not responders and will assist clinicians in the decision making process for optimized, patient-specific therapy.
The main activities of the project are:
i) The development of deep learning methods for automatic segmentation of tumors based on ultrasound b-mode and shear wave elastography images.
ii) The implementation of in vivo experiments in preclinical tumors models to obtain data for the training and validation of the deep learning algorithms
iii) The execution of the phase II clinical trial for the use of ketotifen to improve chemotherapy in patients with sarcoma. More information for the ongoing clinical trial can be found at the EU clinical trials register here:
https://www.clinicaltrialsregister.eu/ctr-search/search?query=ketotifen(öffnet in neuem Fenster)
i) The development of deep learning methods for automatic segmentation of tumors based on ultrasound b-mode and shear wave elastography images.
ii) The implementation of in vivo experiments in preclinical tumors models to obtain data for the training and validation of the deep learning algorithms
iii) The execution of the phase II clinical trial for the use of ketotifen to improve chemotherapy in patients with sarcoma. More information for the ongoing clinical trial can be found at the EU clinical trials register here:
https://www.clinicaltrialsregister.eu/ctr-search/search?query=ketotifen(öffnet in neuem Fenster)
We developed a deep learning methodology for automatic segmentation of tumors based on ultrasound b-mode and shear wave elastography images. In addition we
performed experiments in preclinical tumors models to obtain data for the training and validation of the deep learning algorithms. The new data sets were added to the data pool from experiments that we performed in our laboratory to increase the dataset that was used for the training and the validation of the deep learning algorithm. We found that processing of ultrasound b-mode and elastography images with our deep learning algorithm can accurately be used for automatic segmentation of murine tumors and to predict the efficacy of chemotherapy, immunotherapy or the combination of the two. The results of this work have been reported in two articles.
The one article has been already published in the scientific, peer-reviewed journal, Translational Oncology. It has the title "Machine learning analysis reveals tumor stiffness and hypoperfusion as biomarkers predictive of cancer treatment efficacy" and can be found with open-access here: https://www.sciencedirect.com/science/article/pii/S1936523324000718(öffnet in neuem Fenster)
The second article is entitled ""A convolutional attention model for predicting response to chemo-immunotherapy from ultrasound elastography in mouse tumor models" is currently in revision in the open-access, peer-reviewed journal Nature Communications Medicine.
In addition, we are progressing with the execution of the phase II clinical trial for the use of ketotifen to improve chemotherapy in patients with sarcoma. Information for the ongoing clinical trial can be found at the EU clinical trials register here: https://www.clinicaltrialsregister.eu/ctr-search/search?query=ketotifen(öffnet in neuem Fenster)
Early data of this clinical trial have been included in an article that has been accepted for publication in the journal Clinical Cancer Research. The article has the title "Stabilizing tumor resident mast cells restores T cell infiltration and sensitizes sarcomas to PD-L1 inhibition" and it is planned to be published in May 2024. Out of the 15 patients that we expected to enroll in the clinical study, we have already enrolled 10 patients and we expect to reach the expected number in the next few months. The data so far show that ketotifen is capable to reduce tumor stiffness in human tumors, causing similar effects with those that we observed in preclinical tumor models.
In addition, the deep/machine learning algorithms that we developed have been integrated in a software and we performed a market research and explored the commercialization potential of such a software that reads ultrasound images to predict the efficacy of the cancer therapy prior the initiation of the treatment.
performed experiments in preclinical tumors models to obtain data for the training and validation of the deep learning algorithms. The new data sets were added to the data pool from experiments that we performed in our laboratory to increase the dataset that was used for the training and the validation of the deep learning algorithm. We found that processing of ultrasound b-mode and elastography images with our deep learning algorithm can accurately be used for automatic segmentation of murine tumors and to predict the efficacy of chemotherapy, immunotherapy or the combination of the two. The results of this work have been reported in two articles.
The one article has been already published in the scientific, peer-reviewed journal, Translational Oncology. It has the title "Machine learning analysis reveals tumor stiffness and hypoperfusion as biomarkers predictive of cancer treatment efficacy" and can be found with open-access here: https://www.sciencedirect.com/science/article/pii/S1936523324000718(öffnet in neuem Fenster)
The second article is entitled ""A convolutional attention model for predicting response to chemo-immunotherapy from ultrasound elastography in mouse tumor models" is currently in revision in the open-access, peer-reviewed journal Nature Communications Medicine.
In addition, we are progressing with the execution of the phase II clinical trial for the use of ketotifen to improve chemotherapy in patients with sarcoma. Information for the ongoing clinical trial can be found at the EU clinical trials register here: https://www.clinicaltrialsregister.eu/ctr-search/search?query=ketotifen(öffnet in neuem Fenster)
Early data of this clinical trial have been included in an article that has been accepted for publication in the journal Clinical Cancer Research. The article has the title "Stabilizing tumor resident mast cells restores T cell infiltration and sensitizes sarcomas to PD-L1 inhibition" and it is planned to be published in May 2024. Out of the 15 patients that we expected to enroll in the clinical study, we have already enrolled 10 patients and we expect to reach the expected number in the next few months. The data so far show that ketotifen is capable to reduce tumor stiffness in human tumors, causing similar effects with those that we observed in preclinical tumor models.
In addition, the deep/machine learning algorithms that we developed have been integrated in a software and we performed a market research and explored the commercialization potential of such a software that reads ultrasound images to predict the efficacy of the cancer therapy prior the initiation of the treatment.