{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,11]],"date-time":"2026-10-11T00:25:24Z","timestamp":1791678324758,"version":"4.3.4"},"reference-count":26,"publisher":"Wiley","issue":"9","license":[{"start":{"date-parts":[[2001,9,13]],"date-time":"2001-09-13T00:00:00Z","timestamp":1000339200000},"content-version":"vor","delay-in-days":12,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2001,9]]},"abstract":"<jats:p>Many computer\u2010aided diagnosis (CAD) systems use neural networks (NNs) for either detection or classification of abnormalities. Currently, most NNs are \u201coptimized\u201d by manual search in a very limited parameter space. In this work, we evaluated the use of automated optimization methods for selecting an optimal convolution neural network (CNN) architecture. Three automated methods, the steepest descent (SD), the simulated annealing (SA), and the genetic algorithm (GA), were compared. We used as an example the CNN that classifies true and false microcalcifications detected on digitized mammograms by a prescreening algorithm. Four parameters of the CNN architecture were considered for optimization, the numbers of node groups and the filter kernel sizes in the first and second hidden layers, resulting in a search space of 432 possible architectures. The area <jats:inline-graphic xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"graphic\/mp5036-math-0001.png\" xlink:title=\"urn:x-wiley:00942405:media:mp5036:mp5036-math-0001\"\/> under the receiver operating characteristic (ROC) curve was used to design a cost function. The SA experiments were conducted with four different annealing schedules. Three different parent selection methods were compared for the GA experiments. An available data set was split into two groups with approximately equal number of samples. By using the two groups alternately for training and testing, two different cost surfaces were evaluated. For the first cost surface, the SD method was trapped in a local minimum 91% (392\/432) of the time. The SA using the Boltzman schedule selected the best architecture after evaluating, on average, 167 architectures. The GA achieved its best performance with linearly scaled roulette\u2010wheel parent selection; however, it evaluated 391 different architectures, on average, to find the best one. The second cost surface contained no local minimum. For this surface, a simple SD algorithm could quickly find the global minimum, but the SA with the very fast reannealing schedule was still the most efficient. The same SA scheme, however, was trapped in a local minimum on the first cost surface. Our CNN study demonstrated that, if optimization is to be performed on a cost surface whose characteristics are not known <jats:italic>a priori<\/jats:italic>, it is advisable that a moderately fast algorithm such as a SA using a Boltzman cooling schedule be used to conduct an efficient and thorough search, which may offer a better chance of reaching the global minimum.<\/jats:p>","DOI":"10.1118\/1.1395036","type":"journal-article","created":{"date-parts":[[2002,7,26]],"date-time":"2002-07-26T14:01:01Z","timestamp":1027692061000},"page":"1937-1948","source":"Crossref","is-referenced-by-count":21,"title":["Selection of an optimal neural network architecture for computer\u2010aided detection of microcalcifications\u2014Comparison of automated optimization techniques"],"prefix":"10.1002","volume":"28","author":[{"given":"Metin N.","family":"Gurcan","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Berkman","family":"Sahiner","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Heang\u2010Ping","family":"Chan","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lubomir","family":"Hadjiiski","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicholas","family":"Petrick","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2001,9,13]]},"reference":[{"key":"e_1_2_1_2_1","doi-asserted-by":"publisher","DOI":"10.1118\/1.597428"},{"key":"e_1_2_1_3_1","doi-asserted-by":"publisher","DOI":"10.1016\/0893-6080(95)00061-5"},{"key":"e_1_2_1_4_1","doi-asserted-by":"publisher","DOI":"10.1109\/42.538937"},{"key":"e_1_2_1_5_1","doi-asserted-by":"publisher","DOI":"10.1016\/S0304-3800(99)00096-4"},{"key":"e_1_2_1_6_1","doi-asserted-by":"publisher","DOI":"10.1109\/72.572102"},{"key":"e_1_2_1_7_1","doi-asserted-by":"publisher","DOI":"10.1109\/72.248452"},{"key":"e_1_2_1_8_1","unstructured":"J. Moody and J. Utans \u201cArchitecture selection strategies for neural networks: Application to corporate bond rating prediction \u201d inNeural Networks in Capital Markets(Wiley New York 1994)."},{"key":"e_1_2_1_9_1","doi-asserted-by":"crossref","unstructured":"J. Moody \u201cPrediction Risk and Architecture Selection for Neural Networks \u201dNATO ASI Series F From Statistics to Neural Networks: Theory and Pattern Recognition Applications(Springer\u2010Verlag New York 1994).","DOI":"10.1007\/978-3-642-79119-2_7"},{"key":"e_1_2_1_10_1","doi-asserted-by":"publisher","DOI":"10.1109\/72.265959"},{"key":"e_1_2_1_11_1","doi-asserted-by":"publisher","DOI":"10.1109\/72.265960"},{"key":"e_1_2_1_12_1","doi-asserted-by":"publisher","DOI":"10.1109\/TSMC.1983.6313076"},{"key":"e_1_2_1_13_1","doi-asserted-by":"publisher","DOI":"10.1097\/00004424-197903000-00002"},{"key":"e_1_2_1_14_1","unstructured":"D. M. Green and J. A. Swets Signal Detection Theory and Psychophysics(Wiley New York 1966)."},{"key":"e_1_2_1_15_1","doi-asserted-by":"publisher","DOI":"10.1002\/(SICI)1097-0258(19980515)17:9<1033::AID-SIM784>3.0.CO;2-Z"},{"key":"e_1_2_1_16_1","unstructured":"W. H. Press B. P. Flanner S. A. Teukolsky and W. T. Vetterling Numerical Recipes in C(Cambridge University Press New York 1988)."},{"key":"e_1_2_1_17_1","doi-asserted-by":"crossref","unstructured":"S. Kirkpatrick C. D. Gelatt and M. P. Vecchi \u201cOptimization by simulated annealing \u201d Science 671\u2013680 (1983).","DOI":"10.1126\/science.220.4598.671"},{"key":"e_1_2_1_18_1","doi-asserted-by":"crossref","unstructured":"D. F. Wong H. W. Leong and C. L. Liu Simulated Annealing for VLSI Design(Kluwer Academic Boston 1988).","DOI":"10.1007\/978-1-4613-1677-0"},{"key":"e_1_2_1_19_1","doi-asserted-by":"crossref","unstructured":"P. J. M. van Laarhoven and E. H. L. Aarts Simulated Annealing: Theory and Applications(Kluwer Academic Norwell MA 1987).","DOI":"10.1007\/978-94-015-7744-1"},{"key":"e_1_2_1_20_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.1984.4767596"},{"key":"e_1_2_1_21_1","doi-asserted-by":"publisher","DOI":"10.1016\/0375-9601(87)90796-1"},{"key":"e_1_2_1_22_1","doi-asserted-by":"publisher","DOI":"10.1016\/0895-7177(89)90202-1"},{"key":"e_1_2_1_23_1","unstructured":"D. E. Goldberg Genetic Algorithms in Search Optimization and Machine Learning(Addison\u2010Wesley New York 1989)."},{"key":"e_1_2_1_24_1","doi-asserted-by":"publisher","DOI":"10.1119\/1.17516"},{"key":"e_1_2_1_25_1","unstructured":"L. B. Booker \u201cIntelligent behavior as an adaptation to the task environment \u201d Ph.D. dissertation University of Michigan 1982."},{"key":"e_1_2_1_26_1","doi-asserted-by":"publisher","DOI":"10.1016\/0895-7177(92)90108-W"},{"key":"e_1_2_1_27_1","unstructured":"M. Mitchell J. H. Holland and S. Forrest \u201cWhen will a genetic algorithm outperform hill climbing? inNeural Information Processing Systems 6 edited by J. D. Cowan G. Tesauro and J. Alpspector (Morgan Kaufman San Franscisco CA 1994)."}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.wiley.com\/onlinelibrary\/tdm\/v1\/articles\/10.1118%2F1.1395036","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.wiley.com\/onlinelibrary\/tdm\/v1\/articles\/10.1118%2F1.1395036","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1118\/1.1395036","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,9,16]],"date-time":"2023-09-16T02:54:26Z","timestamp":1694832866000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1118\/1.1395036"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2001,9]]},"references-count":26,"journal-issue":{"issue":"9","published-print":{"date-parts":[[2001,9]]}},"alternative-id":["10.1118\/1.1395036"],"URL":"https:\/\/doi.org\/10.1118\/1.1395036","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2001,9]]}}}