This work was supported by the German Research Foundation (DFG, Collaborative Research Centre CRC 1181, subproject Z02; Priority Programme &#956;Bone, projects BA 4027/10-1 and BO 3811), including additional support for the scanning devices (INST&#160;410/77-1&#160;FUGG and INST&#160;410/93-1&#160;FUGG), and by the Emerging Fields Initiative (EFI) &#8220;Big Thera&#8221; of the Friedrich-Alexander-University Erlangen-N&#252;rnberg.

All care and experimental procedures were performed in accordance with national and regional legislation on animal protection, and all animal procedures were approved by the State Government of Franconia, Germany (reference number 55.2 DMS-2532-2-228).
1. Induction of breast cancer bone metastases in the right hind leg of nude rats
NOTE: A detailed description of the induction of breast cancer bone metastases in nude rats has been published elsewhere...

<ol>
	<li>D'Oronzo, S., Brown, J., Coleman, R. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=The+role+of+biomarkers+in+the+management+of+bone-homing+malignancies.">The role of biomarkers in the management of bone-homing malignancies.</a> Journal of Bone Oncology. 9, 1-9 (2017).</li><li>Ellmann, S., Beck, M., Kuwert, T., Uder, M., B&#228;uerle, T. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Multimodal+imaging+of+bone+metastases:+From+preclinical+to+clinical+applications.">Multimodal imaging of bone metastases: From preclinical to clinical applications.</a> Journal of Orthopaedic Translation. 3 (4), 166-177 (2015).</li><li>Braun, S., Pantel, K. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Clinical+significance+of+occult+metastatic+cells+in+bone+marrow+of+breast+cancer+patients.">Clinical significance of occult metastatic cells in bone marrow of breast cancer patients.</a> The Oncologist. 6 (2), 125-132 (2001).</li><li>Braun, S., Rosenberg, R., Thorban, S., Harbeck, N. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Implications+of+occult+metastatic+cells+for+systemic+cancer+treatment+in+patients+with+breast+or+gastrointestinal+cancer.">Implications of occult metastatic cells for systemic cancer treatment in patients with breast or gastrointestinal cancer.</a> Seminars in surgical oncology. 20 (4), 334-346 (2001).</li><li>Ellmann, S., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Prediction+of+early+metastatic+disease+in+experimental+breast+cancer+bone+metastasis+by+combining+PET/CT+and+MRI+parameters+to+a+Model-Averaged+Neural+Network.">Prediction of early metastatic disease in experimental breast cancer bone metastasis by combining PET/CT and MRI parameters to a Model-Averaged Neural Network.</a> Bone. 120, 254-261 (2018).</li><li>B&#228;uerle, T., Komljenovic, D., Berger, M. R., Semmler, W. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Multi-modal+imaging+of+angiogenesis+in+a+nude+rat+model+of+breast+cancer+bone+metastasis+using+magnetic+resonance+imaging,+volumetric+computed+tomography+and+ultrasound.">Multi-modal imaging of angiogenesis in a nude rat model of breast cancer bone metastasis using magnetic resonance imaging, volumetric computed tomography and ultrasound.</a> Journal of Visualized Experiments. (66), e4178 (2012).</li><li>Merz, M., Komljenovic, D., Semmler, W., B&#228;uerle, T. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Quantitative+contrast-enhanced+ultrasound+for+imaging+antiangiogenic+treatment+response+in+experimental+osteolytic+breast+cancer+bone+metastases.">Quantitative contrast-enhanced ultrasound for imaging antiangiogenic treatment response in experimental osteolytic breast cancer bone metastases.</a> Investigative Radiology. 47 (7), 422-429 (2012).</li><li>B&#228;uerle, T., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Characterization+of+a+rat+model+with+site-specific+bone+metastasis+induced+by+MDA-MB-231+breast+cancer+cells+and+its+application+to+the+effects+of+an+antibody+against+bone+sialoprotein.">Characterization of a rat model with site-specific bone metastasis induced by MDA-MB-231 breast cancer cells and its application to the effects of an antibody against bone sialoprotein.</a> International Journal of Cancer. 115 (2), 177-186 (2005).</li><li>Patel, J., Goyal, R. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Applications+of+Artificial+Neural+Networks+in+Medical+Science.">Applications of Artificial Neural Networks in Medical Science.</a> Current Clinical Pharmacology. 2 (3), 217-226 (2008).</li><li>Naftaly, U., Intrator, N., Horn, D. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Optimal+ensemble+averaging+of+neural+networks.">Optimal ensemble averaging of neural networks.</a> Network: Computation in Neural Systems. 8 (3), 283-296 (1997).</li><li>B&#228;uerle, T., Merz, M., Komljenovic, D., Zwick, S., Semmler, W. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Drug-induced+vessel+remodeling+in+bone+metastases+as+assessed+by+dynamic+contrast+enhanced+magnetic+resonance+imaging+and+vessel+size+imaging:+A+longitudinal+in+vivo+study.">Drug-induced vessel remodeling in bone metastases as assessed by dynamic contrast enhanced magnetic resonance imaging and vessel size imaging: A longitudinal in vivo study.</a> Clinical Cancer Research. 16 (12), 3215-3225 (2010).</li><li>Cheng, C., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Evaluation+of+treatment+response+of+cilengitide+in+an+experimental+model+of+breast+cancer+bone+metastasis+using+dynamic+PET+with+18F-FDG.">Evaluation of treatment response of cilengitide in an experimental model of breast cancer bone metastasis using dynamic PET with 18F-FDG.</a> Hellenic Journal of Nuclear Medicine. 14 (1), 15-20 (2011).</li><li>Marturano-Kruik, A., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Human+bone+perivascular+niche-on-a-chip+for+studying+metastatic+colonization.">Human bone perivascular niche-on-a-chip for studying metastatic colonization.</a> Proceedings of the National Academy of Sciences of the United States of America. 115 (6), 1256-1261 (2018).</li><li>Sonntag, E., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=In+vivo+proof-of-concept+for+two+experimental+antiviral+drugs,+both+directed+to+cellular+targets,+using+a+murine+cytomegalovirus+model.">In vivo proof-of-concept for two experimental antiviral drugs, both directed to cellular targets, using a murine cytomegalovirus model.</a> Antiviral Research. 161, 63-69 (2019).</li><li>. Horos - Free DICOM Medical Image Viewer | Open-Source Available from: <a target="_blank" href="https://www.horosproject.org/">https://www.horosproject.org/</a> (2015)</li><li>. RStudio Team RStudio: Inteegrated Development for R Available from: <a target="_blank" href="https://rstudio.com">https://rstudio.com</a> (2015)</li><li>Dem&#353;ar, J., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Orange:+Data+Mining+Toolbox+in+Python.">Orange: Data Mining Toolbox in Python.</a> Journal of Machine Learning Research. 14, 2349-2353 (2013).</li><li>Saeys, Y., Inza, I., Larra&#241;aga, P. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+review+of+feature+selection+techniques+in+bioinformatics.">A review of feature selection techniques in bioinformatics.</a> Bioinformatics. 23 (19), 2507-2517 (2007).</li><li>. CRAN - Package caret Available from: <a target="_blank" href="https://cran.r-project.org/web/packages/caret/index.html">https://cran.r-project.org/web/packages/caret/index.html</a> (2016)</li><li>. CRAN: Package xgboost - Extreme Gradient Boosting Available from: <a target="_blank" href="https://cran.r-project.org/web/packages/xgboost/">https://cran.r-project.org/web/packages/xgboost/</a> (2019)</li><li>Fern&#225;ndez-Delgado, M., Cernadas, E., Barro, S., Amorim, D., Fern&#225;ndez-Delgado, A. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Do+we+Need+Hundreds+of+Classifiers+to+Solve+Real+World+Classification+Problems.">Do we Need Hundreds of Classifiers to Solve Real World Classification Problems.</a> Journal of Machine Learning Research. 15, 3133-3181 (2014).</li><li>Hira, Z. M., Gillies, D. F. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+Review+of+Feature+Selection+and+Feature+Extraction+Methods+Applied+on+Microarray+Data.">A Review of Feature Selection and Feature Extraction Methods Applied on Microarray Data.</a> Advances in Bioinformatics. 2015, 198363 (2015).</li><li>S&#225;nchez-Maro&#241;o, N., Alonso-Betanzos, A., Tombilla-Sanrom&#225;n, M. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Filter+methods+for+feature+selection+-+A+comparative+study.">Filter methods for feature selection - A comparative study.</a> Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 4881, 178-187 (2007).</li><li>Cawley, G. C., Talbot, N. L. C. C. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Fast+exact+leave-one-out+cross-validation+of+sparse+least-squares+support+vector+machines.">Fast exact leave-one-out cross-validation of sparse least-squares support vector machines.</a> Neural Network. 17 (10), 1467-1475 (2004).</li><li>Forghani, R., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Radiomics+and+Artificial+Intelligence+for+Biomarker+and+Prediction+Model+Development+in+Oncology.">Radiomics and Artificial Intelligence for Biomarker and Prediction Model Development in Oncology.</a> Computational and Structural Biotechnology Journal. 17, 995-1008 (2019).</li><li>Jaffe, C. C. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Measures+of+response:+RECIST,+WHO,+and+new+alternatives.">Measures of response: RECIST, WHO, and new alternatives.</a> Journal of Clinical Oncology Official Journal of the American Society of Clinical Oncology. 24 (20), 3245-3251 (2006).</li><li>Lambin, P., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Radiomics:+Extracting+more+information+from+medical+images+using+advanced+feature+analysis.">Radiomics: Extracting more information from medical images using advanced feature analysis.</a> European Journal of Cancer. 48 (4), 441-446 (2012).</li><li>Gillies, R. J., Kinahan, P. E., Hricak, H. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Radiomics:+Images+are+more+than+pictures,+they+are+data.">Radiomics: Images are more than pictures, they are data.</a> Radiology. 278 (2), 563-577 (2016).</li><li>Nioche, C., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Lifex:+A+freeware+for+radiomic+feature+calculation+in+multimodality+imaging+to+accelerate+advances+in+the+characterization+of+tumor+heterogeneity.">Lifex: A freeware for radiomic feature calculation in multimodality imaging to accelerate advances in the characterization of tumor heterogeneity.</a> Cancer Research. 78 (16), 4786-4789 (2018).</li><li>Ellmann, S., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Application+of+machine+learning+algorithms+for+multiparametric+MRI-based+evaluation+of+murine+colitis.">Application of machine learning algorithms for multiparametric MRI-based evaluation of murine colitis.</a> PLOS ONE. 13 (10), 0206576 (2018).</li></ol>

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

ML algorithms are powerful tools used to integrate several predictive features into a combined model and obtain an accuracy that exceeds that of its separate constituents when used alone. Nonetheless, the actual result depends on several critical steps. First, the ML algorithm used is a crucial factor, because different ML algorithms yield different results. The algorithm used in this protocol is an avNNet, but other promising algorithms include Extreme Gradient Boosting21 or Random Forests. The c...

The purpose of this protocol is to integrate several functional imaging parameters from MRI and PET/CT into a model-averaged neural network (avNNet) ML algorithm. This algorithm predicts the growth of macrometastases in a rat model of breast cancer bone metastases at an early timepoint, when macroscopic changes within the bone are not yet visible.
Prior to the growth of macrometastases, a bone marrow invasion of disseminated tumor cells occurs, commonly referred to as micrometastatic disease1,2. This initial invasion can be considered an early step in metastatic disease, but is typically missed during conventional staging examinations3,4. Although the currently available imaging modalities cannot detect bone marrow microinvasion when used alone, a combination of imaging parameters yielding information on vascularization and metabolic activity has been shown to perform better5. This complementary benefit is achieved by combining different imaging parameters into an avNNet, which is an ML algorithm. Such an avNNet allows for the reliable prediction of bone macrometastases formation before any visible metastases are present. Therefore, integrating imaging biomarkers into an avNNet could serve as a surrogate parameter for bone marrow microinvasion and early metastatic disease.
To develop the protocol, a previously described model of breast cancer bone metastases in nude rats was used6,7,8. The advantage of this model is its site-specificity, meaning that the animals develop bony metastases exclusively in their right hind leg. However, the tumor-take rate of this approach is 60%&#8211;80%, so a considerable number of the animals do not develop any metastases during the study. Using imaging modalities such as MRI and PET/CT, the presence of metastases is detectable from day 30 postinjection (PI). At earlier time points (e.g., 10 PI) imaging does not distinguish between animals that will develop metastatic disease and those will not (Figure 1).
An avNNet trained on functional imaging parameters acquired on day 10 PI, as described in the following protocol, reliably predicts or excludes the growth of macrometastases within the following ~3 weeks. Neural Networks combine artificial nodes within different layers. In the study protocol, the functional imaging parameters for bone marrow blood supply and metabolic activity represent the bottom layer, while the prediction of malignancy represents the top layer. An additional intermediate layer contains hidden nodes that are connected to both the top and the bottom layer. The strength of the connections between the different nodes is updated during the training of the network to perform the respective classification task with high accuracy9. The accuracy of such a neural network can be further increased by averaging the outputs of several models, resulting in an avNNet10.

<table border="0" cellpadding="0" cellspacing="0" style="width:801px;" width="800">
	<colgroup>
		<col />
		<col />
		<col />
		<col />
	</colgroup>
	<tbody>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Binocular Operating Microscope</td>
			<td style="width:239px;">Leica</td>
			<td style="width:154px;">NA</td>
			<td style="width:167px;"></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">ClinScan MR System</td>
			<td style="width:239px;">Bruker</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">DICOM Viewer</td>
			<td style="width:239px;">Horos</td>
			<td style="width:154px;">NA</td>
			<td style="width:167px;">www.horosproject.org</td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Excel: Spreadsheet</td>
			<td style="width:239px;">Microsoft</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">FCS</td>
			<td style="width:239px;">Sigma</td>
			<td style="width:154px;">F2442-500ML</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Gadovist</td>
			<td style="width:239px;">Bayer-Schering</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Inveon PET/CT</td>
			<td style="width:239px;">Siemens</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="42">
			<td height="42" style="height:42px;width:241px;">Inveon Research Workplace Software</td>
			<td style="width:239px;">Siemens Healthcare GmbH</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">IVIS Spectrum</td>
			<td style="width:239px;">PerkinElmer</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
		<tr height="42">
			<td height="42" style="height:42px;width:241px;">MDA-MB-231 human breast cancer cells</td>
			<td style="width:239px;">American Type Culture Collection</td>
			<td style="width:154px;">N/A</td>
			<td></td>
		</tr>
		<tr height="63">
			<td height="63" style="height:63px;width:241px;">Open-source data visualization, machine learning and data mining toolkit.</td>
			<td style="width:239px;">Orange3, University of Ljubljana</td>
			<td style="width:154px;">NA</td>
			<td>https://orange.biolab.si/</td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">RPMI-1640</td>
			<td>Invitrogen/ThermoFisher</td>
			<td align="right" style="width:154px;">11875093</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Trypsin</td>
			<td style="width:239px;">Sigma</td>
			<td style="width:154px;">9002-07-7</td>
			<td></td>
		</tr>
		<tr height="21">
			<td height="21" style="height:21px;width:241px;">Vevo 3100</td>
			<td style="width:239px;">VisualSonics</td>
			<td style="width:154px;">NA</td>
			<td></td>
		</tr>
	</tbody>
</table>

machine learning algorithms for early detection of bone metastases in an experimental rat model

Machine learning (ML) algorithms permit the integration of different features into a model to perform classification or regression tasks with an accuracy exceeding its constituents. This protocol describes the development of an ML algorithm to predict the growth of breast cancer bone macrometastases in a rat model before any abnormalities are observable with standard imaging methods. Such an algorithm can facilitate the detection of early metastatic disease (i.e., micrometastasis) that is regularly missed during staging examinations.
The applied metastasis model is site-specific, meaning that the rats develop metastases exclusively in their right hind leg. The model&#8217;s tumor-take rate is 60%&#8211;80%, with macrometastases becoming visible in magnetic resonance imaging (MRI) and positron emission tomography/computed tomography (PET/CT) in a subset of animals 30 days after induction, whereas a second subset of animals exhibit no tumor growth.
Starting from image examinations acquired at an earlier time point, this protocol describes the extraction of features that indicate tissue vascularization detected by MRI, glucose metabolism by PET/CT, and the subsequent determination of the most relevant features for the prediction of macrometastatic disease. These features are then fed into a model-averaged neural network (avNNet) to classify the animals into one of two groups: one that will develop metastases and the other that will not develop any tumors. The protocol also describes the calculation of standard diagnostic parameters, such as overall accuracy, sensitivity, specificity, negative/positive predictive values, likelihood ratios, and the development of a receiver operating characteristic. An advantage of the proposed protocol is its flexibility, as it can be easily adapted to train a plethora of different ML algorithms with adjustable combinations of an unlimited number of features. Moreover, it can be used to analyze different problems in oncology, infection, and inflammation.

The rats recovered quickly from the surgery and injection of the MDA-MB-231 breast cancer cells and were then subjected to MR- and PET/CT imaging on days 10 and 30 PI (Figure 1). A representative DCE analysis of a rat&#8217;s right proximal tibia is presented in Figure 2A. The DCE raw measurements were saved by selecting the &#8220;Export&#8221; button and choosing &#8220;DCEraw.txt&#8221; as the file name.
Subsequent...

Watch this Scientific Journal Video about Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model at JoVE.com

Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model

This protocol was designed to train a machine learning algorithm to use a combination of imaging parameters derived from magnetic resonance imaging (MRI) and positron emission tomography/computed tomography (PET/CT) in a rat model of breast cancer bone metastases to detect early metastatic disease and predict subsequent progression to macrometastases.

Machine learning (ML) algorithms permit the integration of different features into a model to perform classification or regression tasks with an accuracy ...

This protocol facilitates the development of a machine learning algorithm for predicting the growth of bone metastases in an experimental model at the stage of early organ colonization. The main advantage of this technique is that it combines several imaging parameters into a machine learning algorithm that significantly outperforms the productive ability of each individual parameter. Though, this protocol aims for the early diagnosis of metastases in bone, it can be adapted to different organs or areas of multimodal and multiparametric imaging research.10 days post-surgery and injection in a DICOM viewer with a DCE plugin. Pick import to load the DCE sequence in 4D mode and select the DICOM folder containing the MRI images. Place a circular 1.5 square millimeter to dimensional region of interest in the proximal tibial shaft bone marrow of the right hind leg within the fourth or fifth MRI image and select relative enhancement in the plot type.Enter the baseline range from time points, one to five into the respective fields and export the analysis as a dot TXT file named DCE raw dot TXT. Open RStudio and load the provided DCE script dot R file. Select code, run region and run all to run the entire script and copy the output to the provided imaging features template file.In the DICOM viewer, place a second region of interest within the back muscle of the animal and repeat the DCE measurements as demonstrated. In the imaging feature spreadsheet the respective bone measurements will be automatically divided by the respective muscle measurements for data normalization. To analyze the PET/CT images open the PET/CT analysis software and import the PET/CT images.Click general analysis and select region of interest quantification, create and create a region of interest from a template. Place a four by six millimeter, to dimensional region of interest within the bone marrow of the proximal tibial shaft of the right hind leg and select regions of interest target one overlay. Note the mean, minimum and maximum values in becquerels per milliliter.Then, divide the maximum value by the injected activity and multiply the result by the animal weight in grams to calculate the standardized uptake value. To diagnose the tumor growth rate in the injected hind leg. After obtaining MRI and PET/CT images on day 30 post-injection, analyze the images as demonstrated and add a tumor column to the imaging features spreadsheet.Then enter a one or zero for animals with metastasis or without a visible tumor burden respectively. To determine the most relevant features for the prediction of future tumor growth, import the spreadsheet into an open source data visualization machine learning and data mining toolkit. Drag the file sub-routine from the data menu into the workspace and double-click the file.Click the folder icon to load the spreadsheet and select the imaging features spreadsheet. Select the export worksheet and assign the target attribute to the variable tumor. Assign the skip function to the animal number.Drag the rank sub-routine into the workspace and draw a line to connect to the file and rank sub-routines. Then, double click to open the rank sub-routine and select the information gain algorithm. For machine learning analysis, open RStudio 3.4.1 and load the provided train model R script.Select lines three through six within the script to load the required libraries. Click code to run the selected code and click run selected lines to run the selected lines. To train a model average to neural network algorithm, select lines eight through 39 from the train model R script and click code and run selected lines.Then to assess the standard parameters of diagnostic accuracy, select lines 41 through 50 from the train model R script and click code and run selected lines. On day 10 after surgery and cancer cell injection, MRI and PET/CT images can be acquired. DCE analysis allows the measurement of muscle and bone tissue areas of interest.These values can be normalized by dividing the bone measurements by the muscle measurements. On day 30 post-injection, all of animals are evaluated to determine whether they have developed metasticies with a one indicating a positive tumor burden and a zero indicating healthy animals without visible tumors. Running the train model R script allows the optimal hyper parameter combination to be determined and the final model to be calculated using the optimal hyper parameter combination.These data allow a set of standard diagnostic parameters to be calculated and the receiver operating characteristic curve of the model to be plotted. For example, in this analysis of 28 samples the model performs significantly better than all of its three constituents. Several machine learning algorithms tend to perform better when the input data are normalized.In this protocol normalization is achieved using the Box Cox function. This protocol uses a model average during network as a machine learning algorithm. However, the provided framework can easily be adapted to other algorithms such as random forests or support vector machines.Extracting numerical information from image material has become essential. Algorithms such as these may facilitate the integration of large amounts of data to allow patient stratification.

machine-learning-algorithms-for-early-detection-bone-metastases-an

Research

JoVE Journal

Cancer Research

6.5K visualizações.  Friedrich-Alexander Universität Erlangen-Nürnberg. Este protocolo facilita o desenvolvimento de um algoritmo de aprendizagem de máquina para prever o crescimento de metástases ósseas em um modelo experimental na fase de colonização precoce de órgãos. A principal vantagem dessa técnica é que ela combina vários parâmetros de imagem em um algoritmo de aprendizagem de máquina que supera significativamente a capacidade produtiva de cada parâmetro individual. Embora este protocolo visa o diagnóstico precoce de metástases no osso, pod...