²`«×¾Ç²ß©óÂå¾Ç¼v¹³¤ÀªR¡X Deep learning in
medical image analysis
Fall 2024½Òµ{ºõn
±Â½Ò±Ð®v |
¶À«aµØ±Ð±Â ¿ì¤½«Ç¡Gºî¦X¤@À]423«Ç ¹q¸Ü¡G03-5131334 ¹q¤l¶l¥ó¡Gghuang@nycu.edu.tw |
¤W½Ò®É¶¡»P¦aÂI |
¨C¬P´Á¤@13:20-16:30©óºî¦X¤@À]A304«Ç |
½Òµ{ºô¶ |
|
¶}½Ò³æ¦ì |
²ÎpºÓ |
¥Ã¤[½Ò¸¹ |
SCIS30031 |
¾Ç¤À¼Æ |
3 |
½Òµ{·§z»P¥Ø¼Ð
¾÷¾¹¾Ç²ß(machine learning)ªº³Ì·s¶i®i¡A¯S§O¬O¦b²`«×¾Ç²ß(deep
learning)¡A¹ï©óÂå¾Ç¼v¹³ªºÃѧO¡B¤ÀÃþ©M¯S¼x¿ï¨ú²£¥Í¤F¥¨¤jªº§U¯q¡C³o¨Ç¥ý¶iªº²`«×¾Ç²ß¤èªk¡A¤]±q¦Ó´£°ª¤FÂå¾Ç¼v¹³©ó¦UºØÁ{§ÉÂåÀø¤WªºÀ³¥Î¡C
¥»½Òµ{±N¥H¹ê»ÚªºÂå¾Ç¼v¹³¼Æ¾Ú¬°®Ö¤ß¡A·f°tPython³nÅ骺¨Ï¥Î¡A¤Þ¾É½Òµ{°Ñ»PªÌ±µÄ²¨Ã¾Ç²ß¡G(1)Âå¾Ç¼v¹³ì²z»P³B²zÀx¦s¨t²Î¡A(2)²`«×¾Ç²ß¤èªk»Pì²z¡A¥H¤Î(3)²`«×¾Ç²ß¤èªk©ó¯S©wÂå¾Ç¼v¹³¤ÀªR¤WªºÀ³¥Î¡C
½Òµ{¦@¤À¥|¤j¥DÃD¡A¤À§O¬°¡G
¥DÃD1¡X³æ¥ú¤l¹q¸£Â_¼h±½´y(single photon emission computed
tomography, SPECT)ªº¦hÃþ§O¤ÀÃþ(multiclass classification)¡F
¥DÃD2¡X®ÖºÏ¦@®¶³y¼v(magnetic resonance imaging, MRI)ªº3D¼v¹³¤ÀªR(3D image analysis)¡F
¥DÃD3¡X¶Wµªi±½´y(sonography)©óÀV°Ê¯ß¤À³Î(segmentation)»P¦å¬y³t«×µû¦ô¡F
¥DÃD4¡XX¥ú¼v¹³(X-ray)¤¤±`¨£¯Ý³¡¯e¯f¤§ª«¥ó°»´ú(object detection)¡C
ì«h¤W¡A¨C¤@¥DÃD±N¥H¤T©Pªº®É¶¡§¹¦¨±Â½Ò(°£¤F¥DÃD1¦]¬°·s¼W¤FRNN»PTransformer·|¥H¤©P§¹¦¨)¡G²Ä¤@©P¤¶²Ð¬ÛÃöªºÂå¾Ç¼v¹³ì²z»P³B²zÀx¦s¨t²Î¡A²Ä¤G©P»¡©úÀ³¥Îªº²`«×¾Ç²ß¤èªk»Pì²z¡A²Ä¤T©P«h°Q½×²`«×¾Ç²ß¤èªk©ó¸ÓÂå¾Ç¼v¹³¤ÀªR¥DÃD¤WªºÀ³¥Î¡C
¤W½Ò¤º®e¡A±N¼sªx¥]§t©Ò¦³¬ÛÃöª¾ÃÑ¡A¤W½Ò®É°¼«Á¿z³o¨Çª¾ÃѪº°ò¥»Æ[©À»P¼Ò«¬¸ÑÄÀ¡C¹ï©ó²`¤Jªº²z½×»P¨ä¾l¸Ô²Ó¸ê°T¡A«h¶È§@«ÂI´£¥Ü©Î´£¨Ñ°Ñ¦Ò¤åÄm¡C½Ò°ó¤¤±N¥H¹ê»Úªº¨Ò¤l¨Ó¸É¥R¤W½Ò¤º®e¡A¨Ã°Q½×¬ÛÃö¤èªkªºPython³nÅ骺¹ê§@¡C
¾Ç´Á¦¨ÁZªºµû©w¡A«h¨Ì¾Úú¥æªº5¥÷§@·~¡C
½Òµ{²Õ¦¨³¡¤À
½Ò°óÁ¿¸Ñ
ì«h¤W¡A¨C¬P´Á¤@13:20-15:10 («e¨â¸`½Ò)¡A¥Ñ±Â½Ò±Ð®v(©ÎÁܽÐÁ¿ªÌ)Á¿±Â¬ÛÃöª¾ÃÑ»P¤èªk¡A15:30-16:20 (²Ä¤T¸`½Ò)¡A«h¥Ñ½Òµ{§U±Ð±a»â¦P¾Ç¶i¦æ¬ÛÃö¤èªkªºPython³nÅé¹ê§@¡C
¤W½Ò¤º®e¡A±N¼sªx¥]§t©Ò¦³¬ÛÃöª¾ÃÑ¡A¤W½Ò®É°¼«Á¿z³o¨Çª¾ÃѪº¥Í¦¨°Ê¾÷¡B°ò¥»Æ[©À»P¼Ò«¬¸ÑÄÀ¡C¹ï©ó²`¤Jªº²z½×»P¨ä¾l¸Ô²Ó¸ê°T¡A«h¶È§@«ÂI´£¥Ü©Î´£¨Ñ°Ñ¦Ò¤åÄm¡C´Á¬ß¤é«á·í¾Ç¥Í¿W¥ß¶i¦æÂå¾Ç¼v¹³¤ÀªR®É¡A³o¨Ç¼sªxªºª¾ÃÑ¡A¯à¼W¼s¥LÌ«ä¦Ò°ÝÃDªº¨¤«×¡A¨Ã¦¨¬°²³¦h¥LÌ¥i¿ï¾Üªº¸Ñ¨M¤è®×¡CYn¶i¦æ§ó²`¤Jªº¼Ò«¬¬ã¨s»P²z½×±À¾É®É¡A«hª¾¹Dn±q¦ó¤U¤â»P¨ì¦ó³B¥h§ä´M¬ÛÃöªº»²§U¸ê°T¡C
½Ò«e¡B½Ò«áªº¦Û¦æ¾\Ū¡B¦Û¦æ¾Ç²ß
½Ò°óÁ¿¸Ñ·|¼sªx¥]§t©Ò¦³¬ÛÃö¥DÃD¡A°¼«Æ[©ÀªºÁ¿z¡C¸É¥R»Pl¥Í¤º®e¡A«h·|´£¨Ñ¬ÛÃö¨Ó·½»Pºô¸ô³sµ²¡An¨D¾Ç¥Í©ó½Ò«e©Î½Ò«á¦Û¦æ¾\Ū¡C¤S¥Ñ©óÂå¾Ç¼v¹³¤ÀªR»â°ìªº½´«kµo®i¡A¬ÛÃö¶}©ñ½Òµ{¡B¤ÀªR¤èªk¡B¤ÀªR¤u¨ã¡B¦¨ªGÀ³¥Î¡B¶}©ñ¸ê®Æ¡B¡Kµ¥¹M§G©óºô¸ô¡A¦]¦¹¦P¾ÇÌ«h±`»Ýn(©Î¥i¥H)¦Û¦æ¾Ç²ß·sªº³nÅé¡B¤u¨ã¡A¨Ã§l¦¬·sªºª¾ÃÑ¡BÀ³¥Î¡Cª`·N¡A³\¦hºô¸ô³sµ²»P¤å¥ó¬O¥H^¤å¼¶¼g¡A^¤å¾\Ūªº¯à¤O±N·|«D±`«n¡C
§@·~
²Ä¤@¥÷§@·~±N¤ÀªR¤@µ§¡u¤â¼g¶l»¼°Ï¸¹¼Æ¦r¡v¸ê®Æ¡AÅý¦P¾Ç¥HPython³nÅé¨Ó¹ê§@¶Ç²ÎªºÃþ¯«¸gºô¸ô(neural networks)¼Ò«¬¡A¤]ÂǦ¹¼ô±xPython³nÅ骺¹B§@¡C±µµÛ¨C¤@¥DÃD½Òµ{§¹¦¨«á¡A±N·|´£¨Ñ¸Ó¥DÃDªºÂå¾Ç¼v¹³¼Æ¾Ú¡AÅý¾Ç¥Í½m²ß¼v¹³ªº²M²z¡B¦s¨ú¡F¸ê®Æ¡Bµ²ªGªºµøı¤Æ¡F¦p¦ó¹B¥Î¥¿½T¡B·s¿oªº²`«×¾Ç²ß¤èªk¡C
¨C¥÷§@·~¬ÛÃöªºPythonµ{¦¡½d¨Ò¡A·|¦b¨C¶g§U±Ð³nÅé¹ê§@½Ò(²Ä¤T¸`½Ò)®ÉÁ¿¸Ñ¡C
¥Ñ©ó¤j³¡¥÷ªº§@·~°ÝÃD¡A·|¶·n¥HPython³nÅé¨Ó¶i¦æ¹ê§@¡B¤ÀªR¡A¦]¦¹n¨D¦P¾Ç̪º§@·~n¥Hjupyter (https://jupyter.org/)ªº®æ¦¡¨Ó¼¶¼g¡Cjupyter¯à±N§Aªº¤å¦r»¡©ú¡B¼Æ¾Ç¦¡¤l¡BPythonµ{¦¡¡BPython°õ¦æµ²ªG¡B¡Kµ¥¡Aµ²¦X¦¨¤@Ó¤å¥ó¡A¦p¦¹±N©ö©ó¥L¤H¾\Ū»P«»s(reproduce)§Aªº¤ÀªR¡C¬Y´XÓ§@·~±N·|¥H¡uKaggle¸ê®ÆÄvÁÉ¡vªº§Î¦¡¶i¦æ¡A·|n¨D¦P¾Ç̱N§@·~µ²ªG¤W¶Ç¦ÜKaggle¥¥x(https://www.kaggle.com/)¡A³Ì«á±N¥H¦P¾Ç̦bÄvÁɤWªº¦¨ÁZ©M±Æ¦W§@¬°µû¤Àªº¨Ì¾Ú¡C
§@·~ªº¥Øªº¦b¾Ç²ß¹ê§@²`«×¾Ç²ß¤èªk¡A¨Ã¥B´ú¸Õ§A¹ï½Ò°ó¤º®eªº²z¸Ñµ{«×¡C§â¼g§@·~µø¬°¤@ӾDzߪº¾÷·|¡A¦Ó¤£¬O¬°¤FnÁȨú¤À¼Æ¡C§A¥i»P¨ä¥L¦P¾Ç°Q½×§@·~¡A¥HÀ°§U²z¸Ñ©Ò°Ýªº°ÝÃD¡BÂç²M½Òµ{·§©À¡C¦ý¬O§A¥²¶·¿W¥ß§¹¦¨©Òú¥æªº§@·~¡A§@·~¤¤n¨D¼gªº¹q¸£µ{¦¡¡B¶]ªº¸ê®Æ¤ÀªR¡B¸ÑÄÀªº¤ÀªRµ²ªG¡A³£¤£¥i»P¥L¤H¦@¦P¦X§@¡C
¥ý׬ì¥Ø©Î¥ý³Æ¯à¤O
1.
¦³¼g¹q¸£µ{¦¡ªº¸gÅç
l ¹³¡GC, C++, Java,
Python, R,¡K
2.
³Ì¦n×¹L°ò¦²Îp¾Ç
l ª¾¹D¡GÀH¾÷ÅܼơB«H¿à°Ï¶¡¡B°²³]ÀË©w¡B¡K
3.
Ä@·N¾Ç²ß·sªº³nÅé¡B¤u¨ã
l ±`·|«D±`ªá®É¶¡
l n¤j¶q¾\Ūºô¸ô¤Wªº¤å¥ó
l ¾\Ū³\¦h^¤å¤å¥ó
½Òµ{¹ê§@³nÅé»P±Ð¬ì®Ñ
¥»ªù½Ò±N·|¥HPython³nÅé(https://www.python.org/)¡A¨Ó·í§@¤ÀªR¹ê§@ªº¤u¨ã¡C¦]¦¹¤£½×ºt²ß½Ò§U±ÐÁ¿¸Ñ»P§@·~°ÝÃD¡A¬Ò·|¥HPython³nÅ骺¾Þ§@»P¼¶¼g¬°°ò¦¡C¦P¾Ç̪º§@·~n¥Hjupyter (https://jupyter.org/)ªº®æ¦¡¨Ó¼¶¼g¡A¥H§Q©ó±N§Aªº¤å¦r»¡©ú¡B¼Æ¾Ç¦¡¤l¡BPythonµ{¦¡¡BPython°õ¦æµ²ªG¡B¡Kµ¥¡Aµ²¦X¦¨¤@Ó¤å¥ó¡A¤è«K¥L¤H¾\Ū»P«»s§Aªº¤ÀªR¡C
¥»ªù½ÒÁöµL«ü©w¡B¥²¶·ÁʶRªº±Ð¬ì®Ñ¡AµM¬ÛÃöªº¦Û¦æ¾\Ū¡B¸É¥R±Ð§÷¤º®e¡A±N¥X¦Û¥H¤U´X¥»°Ñ¦Ò®ÑÄy¡G
1.
Goodfellow I, Bengio Y, Courville A (2016): Deep learning. The MIT Press.
(DL) ³o¥»®Ñªº¬ÛÃö°T®§¡A¥i±q¥H¤U³sµ²Àò±o¡Ghttps://www.deeplearningbook.org/
2.
Zhou SK, Greenspan
H, Shen D (editors) (2017): Deep Learning for Medical Image Analysis. Academic Press. (DLMIA)
3.
Hastie T, Tibshirani R, Friedman J (2009): The Elements of
Statistical Learning (2nd edition). Springer. (ESL) ³o¥»®Ñ¬O¡u¾÷¾¹¾Ç²ß¤èªk¡vªº¥Dn°Ñ¦Ò®Ñ¥Ø¡C
¥»½Òµ{©Ò¦³¤W½Ò§ë¼v¤ù»P¬ÛÃö¸É¥R¸ê®Æ¡AÁÙ¦³¥Î¥H°õ¦æ¹ê»Ú¤ÀªR¨Ò¤lªºPythonµ{¦¡¡A³£±N·|¤½§G©ó½Òµ{ºô¶¡C
¾Ç´Á¦¨ÁZµû¤À¤è¦¡
¾Ç´Á¦¨ÁZªºpºâ¡A±N¨Ì¾Úú¥æªº5¥÷§@·~¡G
1.
²Ä¤@¥÷§@·~(Ãþ¯«¸gºô¸ô¼Ò«¬¹ê§@)¡G10%
2.
²Ä¤G¥÷§@·~(³æ¥ú¤l¹q¸£Â_¼h±½´yªº¦hÃþ§O¤ÀÃþ)¡G20%
3.
²Ä¤T¥÷§@·~(®ÖºÏ¦@®¶³y¼vªº3D¼v¹³¤ÀªR)¡G20%
4.
²Ä¥|¥÷§@·~(¶Wµªi±½´y©óÀV°Ê¯ß¤À³Î»P¦å¬y³t«×µû¦ô)¡G20%
5.
²Ä¤¥÷§@·~(X¥ú¼v¹³¤¤±`¨£¯Ý³¡¯e¯f¤§ª«¥ó°»´ú)¡G30%
¡X
³o¥÷§@·~±N¥H¡up¹º®Ñ±³ø§i¡vªº§Î¦¡Ãº¥æ¡A¤º®e¸Ôz¤ÀªRªº¡G°ÝÃD(¥Øªº¬°¦ó¡H·Q¹w´ú©Î¦ôp¤°»ò¡H«n©Ê¡H)¡B¸ê®Æ¨Ó·½(¤½¥Î¸ê®Æ®w¤U¸ü¡B¹êÅç«Ç¸ê®Æ©Î¦Û¦æ·j¶°¡H)»PI´º¡B¤ÀªR¬yµ{(¤ÀªR¤èªk¡B¼Ò«¬¡Bµøı¤Æ¤âªk¡B¡K)¡Bµ²ªG(·sµo²{¡B»PÆ[²³·¾³q¡Bµøı¤Æ)¡C