Featured IGCT Research

Multi-disciplinary research in image-guided brain cancer has focused on integrating the imaging data throughout the multi-disciplinary treatment process. This includes validating the imaging data with the histopathology information to val­idate the imaging signals and translating this data to improve the surgical resection of the tumor; translating the surgical resection information to the radiation therapy treatment plan­ning process and enabling personalized adaptive treatment; fi­nally correlating the comprehensive treatment with the patient outcomes, including local control and normal tissue toxicities.

Optimizing Imaging for Diagnosis, Delineation, and Staging. An incredible amount of work has been done by IGCT members to harness the power of artificial intelligence and advanced imaging to diagnosis and delineate brain tumors and identify the heterogeneous grade throughout the tumor. A strong collaboration between Imaging Physics, Diagnostic Radiology, and Radiation Oncology, led by Drs. Jing Li and Jingfei Ma was estab­lished to develop a computer aided detection algorithm for brain metastases for stereotactic radiosurgery (1). Sara Thrower, Ph.D., the first “IGCT Fellow” worked jointly with Drs. Kristy Brock and Caroline Chung to establish evidence-based guidelines for the minimum imaging slice thickness required for accurate delineation of brain tumors (2). Evan Gates, a Gulf Coast Consortia Training Fellow, demonstrated that while local pathological grade can be predicted with a high accuracy using clinical images, advanced im­aging data can further improve this accuracy, in a study led by Drs. David Fuentes and Dawid Schellingerhout (3).

Improving Treatment Through Image Guidance. An extensive multi-disciplinary team of clinicians and scientists have engaged in exciting translational research to improve the acquisi­tion, analysis, and integration of imaging to guide the delivery of cancer treatment. A lively brainstorming session was hosted by the IGCT to capture the needs of clinicians to improve their ability to effectively treat brain cancer patients. This session led to several initiatives, some of which are highlighted here.

Sara Thrower, Ph.D., joined the IGCT program working jointly between surgery, radiation oncology, and diagnostic imaging. In her collaborative work with Drs. Rick Layman and Jeffrey Wein­berg, she optimized the acquisition of vascular imaging to improve the surgical planning of brain tumors.

Funding from the Marnie Rose Foundation and the MD Ander­son Institutional Research Grant has enabled Dr. Caroline Chung to investigate the use of serial multiparametric MRI to personalize radiation therapy to target the area of highest risk of tumor recur­rence while sparing critical normal tissues by leveraging adaptive techniques. Collaborations with the IGCT Program is facilitating the development of the automated segmentation and deformable registration processes necessary to achieve the precision required for this personalized treatment (4).

Generous funding from the Apache Foundation has enabled research focused on understanding how the brain shifts during neurosurgery and the development of biomechanical models to de­scribe this motion using an approach that is accurate and efficient enough to translate into the clinical environment.

Drs. Ho-Ling (Anthony) Liu, Kyle Noll, Sujit Prabhu and their team discovered that neurocognitive changes of patients following brain tumor resection are associated with changes in brain connec­tomics measured by functional MRI. This builds a strong founda­tion for our direction on developing a brain connectome-based virtual surgery system for predicting patient outcomes after neuro­surgery (5).

References for Image-Guidance for Brain Cancer:

1. Zhou Z, Sanders JW, Johnson JM, Gule-Monroe MK, Chen MM, Briere TM, Wang Y, Son JB, Pagel MD, Li J, Ma J. (2020). Computer-aided detection of brain metas­tases in T1-weighted MRI for stereotactic radiosurgery using deep learning sin­gle-shot detectors. Radiology; 295(2):407-415. PMID: 32181729. doi:10.1148/radiol.20201914792.
2. Thrower SL, Al Feghali KA, Luo D, Paddick I, Hou P, Briere T, Li J, McAleer MF, McGovern SL, Woodhouse KD, Yeboa DN, Brock KK. (2021). The impact of slice thickness on contours of brain metastases for stereotactic radiosurgery. Adv. Radiat Oncol; 6(4):100708. PMID: 34124413. doi:10.1016/j.adro.2021.100708.
3. Gates EDH, Lin JS, Weinberg JS, Prabhu SS, Hamilton J, Hazle JD, Fuller GN, Baladandayuthapani V, Fuentes DT, Schellingerhout, D. (2020). Imaging-based algorithm for the local grad­ing of glioma. AJNR Am J Neuroradiol; 41(3):400-407. PMID: 32029466. doi:10.3174/ajnr.A6405
4. McCulloch M, Anderson B, Buszek SM, Al Feghali K, Elhalawani H, Casoulat G, Chung C, Brock KK. (2018). Biomechanical model-based deformable image regis­tration for glioma patients. Presented at 2018 RSNA Annual Meeting, Chicago, IL. Abstract: SSG15-09. 
5. Noll KR, Chen HS, Wefel JS, Kumar VA, Hou P, Ferguson SD, Rao G, Johnson JM, Schomer DF, Suki D, Prabhu SS, Liu HL. (2021). Alterations in functional connectomics associated with neurocognitive changes following glioma resection. Neurosurgery; 88(3):544-551. PMID: 33080024. doi:10.1093/neuros/nyaa453.

The primary focus of the multi-disciplinary team of Head and Neck Cancer clinicians and scientists has been to improve the detection, treatment, and prevention of therapy- and dis­ease-related injury to normal tissue, improving outcomes by reducing sequelae while preserving or enhancing rates of can­cer cure. This research has spanned from preclinical research to innovative clinical trials and brings together researchers from numerous divisions including diagnostic imaging, radia­tion oncology, surgery, and biostatistics.

Optimizing Imaging for Diagnosis, Delineation, and Staging. The multi-disciplinary IGCT team has successfully developed, optimized, and demonstrated the impact of advanced imaging for the detection of toxicity in head and neck cancer. Drs. Clifton Fuller and Stephen Lai led the investigation into the charac­terization and quantification of dynamic contrast enhanced (DCE)-MRI parameters associated with advanced mandibular osteora­dionecrosis (ORN). Their recently published results confirmed a significantly higher degree of leakiness in the mandibular vascula­ture as measured using DCE-MRI parameters in of advanced ORN, compared to healthy mandible (1).

Dr. Carlos Cardenas, a recent graduate from the MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences Medical Physics program and new Assistant Professor and member of the IGCT, demonstrated the power of convolutional neural networks to automatically segment the clinical target volumes for oropha­ryngeal cancer, under the mentorship of Dr. Laurence Court. The evaluation of their model on an independent test set demonstrated a high degree of overlap and a close surface distance agreement compared to the ground truth contours (2).

Improving Treatment Through Image Guidance. Through a multi-disciplinary and multi-institutional collaboration, Dr. Molly McCulloch, a post-doctoral fellow under the mentorship of Dr. Kristy Brock, investigated the deviations between the planned and accumulated doses in head and neck cancers treated with radi­ation therapy to predict clinically significant dosimetric deviations during treatment to evaluate the potential need to adapt treatment. This study was the first to provide clinical guidance, validated through multi-institutional data, of this dose deviation threshold (3).

As one of the first radiation oncology departments to have an MR-guided LINAC installed, Dr. Abdallah Mohamed investigated the feasibility and dosimetric advantages of MRI-guided dose adap­tation for human papillomavirus positive oropharyngeal cancer pa­tients compared with standard intensity modulated radiation thera­py, under the mentorship of Dr. Clifton Fuller and in collaboration with the MD Anderson MRLinAc Development Working Group. This in silico study demonstrated that the proposed MRI-guided adaptive approach is technically feasible and advantageous in reduc­ing dose to organs at risk, especially swallowing. This work has now been translated into a clinical trial (4).

Predicting Toxicity Using Artificial Intelligence. Exciting early results are demonstrating use of convolutional layers joined by residual connection to extract features and feed into densely connected layers along with clinical factors to predict occurrence of ORN. This work, a collaboration between Drs. Kristy Brock, Clif­ton Fuller, and Stephen Lai, is being performed by MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences graduate student Brandon Reber.

References for Image-Guidance for Head & Neck Cancer:

1. Joint Head and Neck Radiation Therapy-MRI Development Cooperative, Mohamed ASR, He R, Ding Y, Wang J, Fahim J, Elgohari B, Elhalawani H, Kim AD, Ahmed H, Garcia JA, Johnson JM, Stafford RJ, Bankson JA, Chambers MS, Sandulache VC, Fuller CD, Lai SY. (2020).  Quantitative dynamic contrast-enhanced MRI identify radiation-induced vascular damage in patients with advanced osteora­dionecrosis: Results of a prospective study. Int J Radiat Oncol Biol Phys; 108(5):1319-1328. PMID: 32712257. doi:10.1016/j.ijrobp.2020.07.029.
2. Cardenas CE, Anderson BM, Aristophanous M, Yang J, Rhee DJ, McCarroll RE, Mohammed ASR, Kamal M, Elgohari BA, Elhalawani HM, Fuller CD, Rao A, Garden AS, Court LE. (2018). Auto-delineation of oropha­ryngeal clinical target volumes using 3D convolutional neural networks. Phys Med Biol; 63(21):215026. PMID: 30403188. doi:10.1088/1361-6560/aae8a9.
3. McCulloch MM, Lee C, Rosen BS, Kamp JD, Lockhart CM, Lee JY, Polan DF, Hawkins PG, Anderson CJR, Heukelom J, Sonke JJ, Fuller CD, Balter JM, Ten Haken RK, Eisbruch A, Brock KK. (2019). Predictive models to determine clinically relevant deviations in delivered dose for head and neck cancer. Pract Radiat Oncol; 9(4):e422-e431. PMID: 30836190. doi:10.1016/j.prro.2019.02.014.
4. Mohamed ASR, Bahig H, Aristophanous M, Blanchard P, Kamal M, Ding Y, Cardenas CE, Brock KK, Lai SY, Hutcheson KA, Phan J, Wang J, Ibbott G, Gabr RE, Narayana PA, Garden AS, Rosenthal DI, Gunn GB, Fuller CD, MD Anderson MRLinac Development Working Group. (2018). Prospective in silico study of the feasibility and dosimetric advantages of MRI-guided dose adaptation for human papillomavirus positive oropharyngeal cancer patients compared with standard IMRT. Clin Transl Radiat Oncol; 11:11-18. PMID: 30014042. doi:10.1016/j.ctro.2018.04.005.

Multi-disciplinary research in image-guided lung cancer has focused on expanding the use of minimally invasive surgery and improving our understanding of the delivered dose in radiation oncology, and the correlation of delivered dose with radiographic indications of normal lung toxicity. We have been fortunate to receive a generous donation to fund the pilot work in image-guided lung surgery. This work has also been leveraged to assist in the iD3CODE initiative for evaluating the CT images of COVID-19 positive patients.

Optimizing Imaging for Diagnosis, Delineation, and Staging. A multi-disciplinary team of clinicians and scientists have partnered in the IGCT to advance the segmentation of the lung, especially in the presence of pathologies, the vascular struc­ture within the lung, and the tumor. The importance of this work was amplified with the onset of the COVID-19 pandemic. The figure below demonstrates the automatic segmentation of lung vasculature (left) by Dr. Guillaume Cazoulat, the development of a convolutional neural network to segment the lung in the presence of complex pathologies (middle), by Dr. Bastien Rigaud and Brian Anderson under the mentor of Dr. Kristy Brock, and the identifica­tion of ground glass and consolidation (right) by Yulun He (The University of Texas MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences graduate student under the mentorship of Dr. Brock), in collabo­ration with Drs. Myrna Godoy and Carol Wu and the iD3CODE team at MD Anderson.

Improving Treatment Through Image Guidance. Drs. Guillaume Cazoulat and Ravi Rajaram are pioneering an innovative strategy to expand the use of minimally invasive surgery. Using patients who presented with large pneumothorax following lung biopsy, shown on the right, they demonstrated the accuracy of biomechanical models to describe the deflation of the lung, as demonstrated through CT im­ages. In these cases, large volume changes and deformation of the lung were observed, similar to the surgical scenario. In this initial evalua­tion, an accuracy of better than 1 cm could be achieved. Work is ongoing to further improve these models in preparation of the early phase clinical trial (1).

Advancing Image-Based Outcomes Assessment. Under­standing the relationship between a delivered focal therapy and any potential normal tissue toxicity is critical for the development and validation of novel methods to reduce or eliminate them. Yulun He, under the mentorship of Drs. Kristy Brock and Radhe Mohan, in collaboration with Drs. Zhongxing Liao and Carol Wu, is devel­oping a novel approach to perform voxel-level dose accumulation of the delivered radiation therapy dose and precisely link this dose with the radiographic response of the normal lung. To date, his work has demonstrated that the deformation of average 4DCT im­ages for dose accumulation should not be used when the breathing motion exceeds 10 mm, as the uncertainties exceed the recommen­dations set by the American Association of Physicists in Medicine. Building on this knowledge to accurately accumulate the delivered dose, his early work has indicated that there may be a meaningful difference between the planned and accumulated dose when eval­uating the correlation between dose and radiographic evidence of lung toxicity.

References for Image-Guidance for Lung Cancer:

1. Lesage AC, Rajaram R, Tam AL, Rigaud B, Brock KK, Rice DC, Cazoulat G. (2020). Preliminary evaluation of biomechanical modeling of lung deflation during minimally invasive surgery using pneumothorax computed tomography scans. Phys Med Biol; 65(22):225010. PMID:32906090. doi.10.1088/1361-6560/abb6ba.

Research by the IGCT on image-guidance for liver cancer is an excellent demonstration of the collaborative nature of the program and the strong record of IGCT-technology translat­ing directly to patient care. The IGCT liver team has focused on developing advanced image analysis techniques, advancing image guidance techniques to improve local control, under­standing therapeutic response to focal therapy, and improving quality in image acquisition.

Optimizing Imaging for Diagnosis, Delineation, and Staging. An AI algorithm was developed by Brian Anderson, a graduate student mentored by Dr. Kristy Brock, in collaboration with Drs. Eugene Koay and Bruno Odisio. Three innovative algorithms automatically segment the liver, disease, and focal ablation region after therapy. The liver segmentation algorithm has been clinically commissioned and deployed in Radiation Oncology, where it has been used nearly 1000 times, and is currently being expanded to include segmentation of the liver lobes, which will enable its use for liver surgery (1).

These automated segmentation technologies have enabled Dr. Moiz Ahmad to investigate the development of a system to auto­matically score the image quality of every diagnostic imaging exam performed. By using automatically extracted image quality metrics of image noise, blood vessel enhancement, blur due to patient motion, and the presence or artifact validated against observer assessments, this goal can be systematically applied in the clinic, eliminating the need for patient recalls and additional imaging.

Innovative research led by Dr. Eugene Koay in collaboration with Drs. Peter Park, Jingfei Ma, Eric Tamm, and Sam Beddar, is devel­oping a novel enhancement pattern mapping technique to improve detectability of hepatobiliary tumors on CT (2).

Improving Treatment Through Image Guidance. Current rates of local recurrence at lesions treated with liver ablation are reported to be around 25%. One of the main factors associated with this high recurrence rates is the inability to properly access intra-pro­cedurally whether sufficient ablation tumor coverage was achieved. The use of advanced intra-procedure image guidance (Morfeus, a biomechanical model-based registration algorithm developed in Dr. Kristy Brock’s lab) will allow the interventional radiologist to assess proper tumor targeting and whether sufficient ablation margins were achieved, potentially providing improved tumor coverage and reducing local recurrence rates. This hypothesis will be tested in a randomized Phase II clinical trial, which recently opened at MD Anderson. Dr. Brock and Dr. Odisio will work in close collaboration with other key personnel including Dr. Kyle Jones and Dr. Guillaume Cazoulat of Imaging Physics. The grant is an NIH funded Academic Industry Partnership with RaySearch Laboratories, which will help to ensure clinical translation.

Advancing Image-Based Outcomes Assessment. The development of a robust and accurate automatic segmentation of the liver has enabled the expansion of biomechanical model-based registration for evaluation of therapeutic response. In a collaboration between Drs. Kristy Brock and Eugene Koay, the evaluation of cur­rent, clinically available deformable registration algorithms was per­formed, demonstrating the need for improved algorithms. (3) These algorithms are critical to evaluate the correlation of delivered dose with image-based toxicity, such as changes in DCE-MRI parameters. A collaboration between Drs. Brock, Cazoulat, and Koay pursued the development of a vasculature-driven biomechanical model to accu­rately provide this correlation. This advanced algorithm provides im­proved results, even in cases are large and complex deformation (4).

References for Image-Guidance for Liver Cancer:

1. Anderson BM, Lin EY, Cardenas CE, Gress DA, Erwin WD, Odisio BC, Koay EJ, Brock KK. (2020). Automated contouring of contrast and noncontrast computed tomography liver images with fully convolutional networks. Adv Radiat Oncol; 6(1):1000464. PMID: 33490720. doi:10.1016/j.adro.2020.04.0232.
2. Park PC, Choi GW, Zaid MM, Elganainy D, Smani DA, Tomich J, Samaniego R, Ma J, Tamm EP, Beddar S, Koay EJ. (2020). Enhancement pattern mapping technique for improving contrast-to-noise ratios and detectability of hepatobiliary tumors on multiphase computed tomography. Med Phys; 47(1):64-74. PMID: 31449684. doi:10.1002/mp.137693.              
3. Sen A, Anderson BM, Cazoulat G, McCulloch MM, Elganainy D, McDonald BA,  He Y, Mohamed ASR, Elgohari BA, Zaid M, Koay EJ, Brock KK. (2020). Accuracy of deformable image registration techniques for alignment of longitudinal cholangiocarcinoma CT images. Med Phys; 47(4):1670-1679. PMID: 31958147. doi:10.1002/mp.14029.
4. Cazoulat G, Elganainy D, Anderson BM, Zaid M, Park PC, Koay EJ, Brock KK. (2019). Vasculature-driven biomechanical deformable image registra­tion of longitudinal liver cholangiocarcinoma computed tomographic scans. Adv Radiat Oncol; 5(2):269-278. PMID: 32280827. doi: 10.1016/j.adro.2019.10.002.

Optimizing Imaging for Diagnosis, Delineation, and Staging. The emergence of deep learning for segmentation and automation in treatment planning has been a focus of research in cervix cancer. The applications of this work focus on two critical ar­eas: facilitating the treatment of cervical cancer patients in low- and middle-income countries with limited resources and enabling the investigation of personalized, high precision adaptive radiotherapy to reduce toxicities. Drs. Kristy Brock and Laurence Court have partnered in these initiatives.

DJ Rhee, a graduate student under the mentorship of Dr. Court, has developed a convolutional neural network for the auto­matic contouring of clinical treatment volumes and normal tissues (1). These contours can then be used to automatically optimize beam weights for 3D conformal plans.

Dr. Bastien Rigaud, a postdoctoral fellow under the mentorship of Dr. Brock, developed a convolutional neural network to segment the normal tissues and primary target volume and integrated this into an automated process to perform online adaptive planning. Online adaptive radiotherapy can enable reduced uncertainty mar­gins, which we hypothesize will lead to significantly lower rates of toxicity while maintaining local control. A randomized clinical trial is being planned to investigate (2).

Improving Treatment Through Image Guidance. Drs. Ann Klopp and Aradhana Venkatesan have led a highly multi-disciplinary team in the development of an intraoperative 3T MRI-guided brachytherapy program within a diagnostic imaging suite. Careful development of the methods, process workflow, and value-based analysis demonstrates the feasibility of this approach, where clinical trials are now ongoing to evaluate the improvement in the therapeutic ratio for gynecologic cancers. (3)

Modeling the complex changes in anatomy in the pelvis, espe­cially the sigmoid colon, between external beam radiotherapy and brachytherapy has limited the ability to quantify the total dose de­livered to critical normal tissues. In a study by Dr. Bastien Rigaud, an innovative method was developed to preserve the topology of the sigmoid colon while accurately correlating the geometric location in both instances. The dosimetric analysis suggests that correctly accumulating the dose using these models could have a substantial impact on our understanding of the dose-toxicity rela­tionship (4).

Advancing Image-Based Outcomes Assessment. Dr. Jennifer Ho, a resident under the mentorship of Dr. Klopp, investigated the utility of volumetric diffusion weighted MR imag­ing, relative to other clinical factors, in the ability to predict recur­rence and survival following definitive chemoradiation in cervical cancer patients. In this retrospective study, apparent diffusion coefficient, determined from the MRI, was a significant predictor of both profession free survival and distant metastasis free survival, independent of standard clinical factors. These results, which must be confirmed in prospective trials, have the potential to identify patients which require treatment intensification (5).

References for Image-Guidance for Cervix Cancer:

1. Rhee DJ, Jhingran A, Rigaud B, Netherton T, Cardenas CE, Zhang L, Vedam S, Kry S, Brock KK, Shaw W, O'Reilly F, Parkes J, Burger H, Fakie N, Trauernicht C, Simonds H, Court LE. (2020). Automatic contouring system for cervical cancer using convolutional neural networks. Med Phys; 47(11):5648-5658. PMID: 32964477. doi: 10.1002/mp.14467.
2. Rigaud B, Anderson BM, Yu ZH, Gobeli M, Cazoulat G, Soderberg J, Samuelsson E, Lidberg D, Ward C, Taku N, Cardenas C, Rhee DJ, Venkatesan AM, Peterson CB, Court L, Svensson S, Lofman F, Klopp AH, Brock KK. (2021). Automatic segmentation using deep learning for online dose optimization during adaptive radiotherapy of cervical cancer. Int J Radiat Oncol Biol Phys; 109(4):1096-1110. PMID: 33181248. doi:10.1016/j.ijrobp.2020.10.038.
3. Ning MS, Venkatesan AM, Stafford RJ, Bui TP, Carlson III R, Bailard NS, Vedam S, Davis R, Olivieri ND, Guzman AB, Incalcaterra JR, McKelvey FA, Thaker NG, Rauch GM, Tang C, Frank SJ, Joyner MM, Lin LL, Jhingran A, Eifel PJ, Klopp AH. (2020). Developing an intraoperative 3T MRI-guided brachytherapy program within a diagnostic imaging suite: Methods, process workflow, and value-based analysis. Brachytherapy; 19(4):427-437. PMID: 31786169. doi:10.1016/j.brachy.2019.09.010. 
4. Rigaud B, Cazoulat G, Vedam S, Venkatesan AM, Peterson CB, Taku N, Klopp AH, Brock KK. (2020). Modeling complex deformations of the sigmoid colon between external beam radiation therapy and brachytherapy images of cervical cancer. Int J Radiat Oncol Biol Phys; 106(5):1084-1094. PMID: 32029345. doi:10.1016/j.ijrobp.2019.12.028.
5. Ho JC, Fang P, Cardenas CE, Mohamed ASR, Fuller CD, Allen PK, Bhosale PR, Frumovitz MM, Jhingran A, Klopp AH. (2019). Volumetric assessment of apparent diffusion coefficient predicts outcome following chemoradiation for cervical cancer. Radiother Oncol; 135:58-64. PMID: 31015171. doi:10.1016/j.radonc.2019.02.012.

One of the primary motivations in the creation of the Image Guided Cancer Therapy Research Program is to ensure that scientific innovations get translated into clinical use, clinical problems get identified and described to scientists to develop multidisciplinary teams to address them, and technology bar­riers get broken down through strong collaborations between clinicians, scientists, and industry partners. Several industry collaborations have been advanced by the IGCT as well as supported by the IGCT faculty through institutional strategic alliances. Here are some highlights of the technology barriers being addressed to tackle clinical challenges in surgery, radia­tion oncology, and interventional radiology.

One of the IGCT’s strategic goals is to demonstrate a signif­icant improvement in the local control and quality of life of cancer patients through image guided cancer therapy. This requires innovative, multi-disciplinary clinical trials. Many of the IGCT faculty are actively involved in the 210 clinical protocols that are currently ongoing. Here, we highlight four clinical trials that were actively developed through the IGCT Research Program.

Regardless of focal therapy intervention — radiation, surgical excision, or focal ablation — precisely identifying the target, developing a personalized treatment plan, and accurately ex­ecuting the best possible treatment is crucial to ensuring local control and overall survival. Collaborations between imaging scientists and clinicians drive the development of innovative imaging techniques to identify the heterogeneity within the tumor and the precise boundaries of its extent. Multidisci­plinary alliances between clinicians, scientists, and, in many instances industry partners, enable the translation of the information obtained on these high-precision targets into the treatment room through image guidance.

For the treatment of many tumors, avoidance of critical nor­mal tissue is just as challenging — and important — as targeting the tumor. The development of novel imaging techniques to characterize the function of formal tissue is a key strategic aim of the IGCT Program. In addition, during image guided treatment planning and execution, the efficient and accurate identification of normal tissues is being accelerated through the development of 2D and 3D deep learning tools. Here are some examples.