Sib sib zog nqus Structured Residual Encoder-decoder Network Nrog Ib Qhov Kev Tsis Txaus Siab rau Nuclei Segmentation Ntawm Lub raum Thiab Mis Histopathology Duab

Abstract

Txhawm rau txhim kho cov txheej txheem ntawm kev kuaj mob thiab kho cov kab mob qog noj ntshav, tsis siv neeg segmentation ntawm hematoxylin thiab eosin (H & E) stained cell nuclei los ntawm cov duab histopathology yog thawj kauj ruam hauv cov kab mob digital. Lub tswv yim sib sib zog nqus cov txheej txheem residual encoder-decoder network (DSREDN) tsom rau ob yam: ua ntej, nws tau siv cov kev sib txuas ntawm cov seem hauv lub network thiab muab txoj hauv kev dav thiab sib sib zog nqus encoder-decoder, uas ua rau kom ntes cov ntsiab lus tseem ceeb thiab ntau qhov chaw hauv cheeb tsam. Qhov thib ob, txoj kev ploj mus ntawm qhov kuaj pom nuclei tau hais los ntawm kev tawm tswv yim kev ua haujlwm tsis zoo uas ua haujlwm zoo dua peb cov qauv npaj thiab txo qhov kev twv tsis tseeb uas tsis xav tau, tshwj xeeb tshaj yog hauv kev siv kho mob. Lub tswv yim tsim qauv tau sim ntawm peb qhov sib txawv muaj nyob rau hauv H&E stained histopathological datasets xws li: (I) raum (RCC) (II) Triple Negative Breast Cancer (TNBC) (III) MoNuSeg-2018. Peb tau txiav txim siab F1- cov qhab nia, Aggregated Jaccard Index (AJI), tag nrho cov tsis muaj, thiab FLOPs (Floating point operations), uas yog feem ntau nyiam kev ntsuas kev ntsuas rau kev sib piv ntawm nuclei segmentation. Cov qhab-nees ntsuam xyuas ntawm nuclei segmentation qhia tau hais tias cov tswv yim tsim qauv tau ua tiav cov txiaj ntsig zoo tshaj tsib lub xeev-ntawm-tus-kos duab kev kawm sib sib zog nqus ntawm peb qhov sib txawv histopathology datasets. Cov txiaj ntsig pom kev sib cais qhia tau tias cov qauv DSREDN tau thov kom raug faib cov cheeb tsam nuclear ntau dua li cov txheej txheem ntawm lub xeev-of-the-art.

Ntsiab lus

Kev kuaj mob raum mob qog noj ntshav thiab kuaj mob · Nuclei segmentation · Residual learning · Histopathology images.

Nyem qhov no kom paub qhov cuam tshuam ntawm Cistanche

Taw qhia

Cov kev tshawb fawb tsis ntev los no qhia tau hais tias lub hauv paus kev kawm sib sib zog nqus ua tau zoo heev rau segmentation, nrhiav pom, thiab lwm yam haujlwm hauv computer pom. Nyob rau hauv kaum xyoo dhau los, nrog rau kev nce qib ntawm hom tshiab ntawm kev suav tshuab, cov tswv yim tsim nyog los daws cov teeb meem overfitting los cob qhia cov tes hauj lwm sib sib zog nqus, thiab ntau qhov kev hloov pauv uas tsim nyog rau kev kawm tob. Segmentation ntawm hematoxylin thiab eosin (H & E) los ntawm stained histopathology dluab yog thawj qhov yuav tsum tau ua ua ntej hauv kev tsim kab mob. Cov kev npaj kho mob histopathology tau tham los ntawm Slaoui M li al. hauv [27], los ntawm cov kauj ruam hauv qab no: (I) Cov ntaub so ntswg (II) Fixation (III) Embedding (IV) Sectioning (V) De-paraffining (VI) Staining (VII) Digitizing tus swb los ntawm tag nrho cov duab swb (WSI) . Muaj ntau ntau txoj kev sau cov ntaub so ntswg uas zoo koob aspiration, biopsy koob, excisional biopsy, thiab lwm yam. Ib tug loj biopsy muaj ntaub ntawv ntau tshaj ib tug me me koob biopsy vim nws khaws cia lub loj cellular ntsiab lus ntawm histopathology slides. Fixation ntawm cov ntaub so ntswg yog xav tau rau tshuaj thiab lub cev stabilization. Embedding yog yuav tsum tau muab ib qho zoo rau cov ntaub so ntswg xws li nws tuaj yeem txiav tau yooj yim los ntawm cov tshuab. Kev faib tawm yuav tsum tau txais tag nrho cov ntaub ntawv peb-dimensional cov ntaub ntawv nyob rau hauv daim ntawv ntawm ntau cov ntaub ntawv nyias nyias thiab cov ntaub ntawv ob-dimensional. Tshem tawm paraffin los ntawm cov ntaub so ntswg yog ib qho tseem ceeb, tsis muaj de-paraffining cov ntaub so ntswg yuav zoo li me ntsis qhov muag plooj nyob rau hauv qee qhov chaw. Staining ntawm cov ntaub so ntswg slides yuav tsum tau vim nws tsis pom los yog hom pob tshab nyob rau hauv lub ci teb microscopy. Feem ntau siv stains rau cov duab histopathology yog hematoxylin thiab eosin. Segmentation cov dej num tuaj yeem muab faib ua cov tsoos lossis xuas tes ua cov txheej txheem rho tawm thiab CNN-raws li kev kawm tob. Cov txheej txheem segmentation ib txwm yog feem ntau raws li kev sib luag-raws li txoj hauv kev, kev tsis sib txuas-raws li txoj hauv kev, cov txheej txheem dej ntws, cov txheej txheem ua kom nquag plias thiab lawv cov kev hloov pauv, superpixel, thiab cov txheej txheem sib koom ua ke, thiab lwm yam. Cov txheej txheem zoo sib xws tau tham los ntawm Gonzalez RC li al. nyob rau hauv [8], yog raws li nyob rau hauv lub zos thresholding, ntiaj teb no thresholding, adaptive thresholding, Otsu thresholding, cheeb tsam loj hlob, cheeb tsam splitting, thiab merging, qhov twg cov txheej txheem no sim pab pawg thiab segment zoo li pixels. Rau cov duab histograms uas muaj cov kwj ha, qhov zoo sib xws-raws li txoj hauv kev tsis ua haujlwm zoo thiab qhov kev xaiv tsis raug ntawm tus nqi pib yuav ua rau muaj ntau tshaj segmentation thiab under-segmentation hauv qhov no. Qhov kev tsis sib haum xeeb raws li txoj hauv kev sim seg cov pixels uas raug cais nyob rau hauv qee yam xws li cov ntsiab lus, kab, thiab cov npoo, thiab nws yog lub ntsej muag ua raws li txoj hauv kev. Txoj kev no yuav tsum muaj cov neeg ua haujlwm sib txawv ntawm ntau theem. Cousty J et al. tau npaj ib txoj hauv kev segmentation watershed nyob rau hauv [4], raws li kev sib cais, sib koom ua ke, thiab cim-tswj watershed. Cov ciam teb nyob rau hauv txoj kev watershed nyob ntawm seb cell complexity. Song T et al. thov kom active contour segmentation nyob rau hauv [28], qhov uas lawv xav txog kev siv cov ntaub ntawv thiab cov ntaub ntawv ntug hauv zos rau kev tshawb nrhiav cov khoom ciam teb. Txoj kev segmentation superpixel siv los ntawm Albayrak A li al. hauv [1], yog raws li pawg ntawm kev sib txuas pixels uas muaj cov yam ntxwv zoo ib yam. Nws txiav txim siab cov xim thiab tswj cov ntaub ntawv ntawm cov neeg nyob ze pixels. Cov txheej txheem no muab cov ntaub ntawv zoo dua hauv cheeb tsam tab sis tsis muaj txiaj ntsig zoo nyob rau hauv rooj plaub ntawm cell segmentation. Clustering-based segmentation tau npaj los ntawm Win KY li al. hauv [37], ua pab pawg raws li lawv qhov zoo sib xws. Hauv kev tshawb fawb ua haujlwm tsis ntev los no, feem ntau ntawm cov kws sau ntawv tau tshaj tawm tias cov txheej txheem segmentation raws li kev sib sib zog nqus convolutional neural network ua tau zoo dua li cov txheej txheem segmentation. Kev tshuaj xyuas cov ntsiab lus ntawm CNN raws li txoj hauv kev tau nthuav tawm hauv Tshooj 2. Cov kev kawm sib sib zog nqus segmentation kuj raug kev txom nyem los ntawm ntau yam teeb meem. Yog tias peb faib cov kev cov nyom no, nws yuav ua raws li cov hauv qab no.

1. Vim muaj kev hloov pauv loj ntawm cov ntaub so ntswg zoo li qub thiab muaj ntau qhov sib txawv ntawm cov chav kawm thiab sub-class ntawm cov ntaub so ntswg, nws yog qhov nyuaj rau kev paub.

2. Segmentation ntawm cov ciam teb nyuaj, overlapped ciam teb, thiab vanishing ciam teb tsis yog ib txoj hauj lwm yooj yim.

3. Kev npaj qhov tseeb hauv av nyob rau hauv rooj plaub ntawm kev saib xyuas kev kawm kuj yog ib qho nyuaj heev. Kev saib xyuas ntawm cov kws paub txog kab mob yog qhov tsim nyog vim tias qhov tseeb ntawm kev twv ua ntej yog nyob ntawm qhov tseeb hauv av.

Nyob rau hauv cov ntaub ntawv ntawm complex histopathology dluab, cov pa txoj kev raug kev txom nyem over-segmentation los yog under-segmentation. Txoj kev npaj tau tsom mus rau cais cov cheeb tsam sib tshooj thiab ploj mus los ntawm cov duab histopathology. Txhawm rau daws cov teeb meem hauv ntu ntawm nuclei los ntawm cov duab histopathology peb cov kev koom tes hauv daim ntawv no yog raws li hauv qab no.

1. Txhawm rau ntxiv dag zog rau ntau theem nruab nrab cov yam ntxwv, peb cov qauv DSREDN tau thov kom siv tau zoo ntawm kev kawm seem.

2. Los ntawm cov pov thawj empirical thiab ua tib zoo sim thiab tsom xam, peb tau npaj qhov kev poob haujlwm tshiab. Cov txiaj ntsig pom thiab kev ua tau zoo matrices qhia tau hais tias peb qhov kev poob haujlwm zoo dua qhia tus qauv thiab raug faib cov cheeb tsam nuclear piv rau cov txheej txheem hauv xeev.

Cistanche tubulosa

Lwm yam haujlwm

Feem ntau ntawm CNN architecture rau cov haujlwm ntawm tes segmentation muaj txoj hauv kev encoder-decoder rau feature extraction. Ntau qhov kev tshawb fawb tsis ntev los no siv ntau lub sijhawm muaj peev xwm xws li kev txhim kho cov tswv yim kev cob qhia, daws teeb meem overfitting, txoj kev ua kom zoo dua qub, thiab ntau lwm cov tswv yim kom tau txais kev kwv yees zoo dua. Txawm li cas los xij, ntau tus kws sau ntawv tau tshaj tawm lawv cov txiaj ntsig uas muaj txiaj ntsig zoo tab sis qhov tseeb thiab muaj txiaj ntsig segmentation algorithm tseem qhib kev tshawb fawb vim qhov nyuaj ntawm cov duab histopathology. Ib qho tseem ceeb ntawm kev koom tes los ntawm Ronneberger li al. nyob rau hauv [26], hu ua UNet, muab ib tug zoo heev kev taw qhia thiab ib tug ua heev breakthrough nyob rau hauv lub teb ntawm biomedical duab segmentation. UNet yog ib tug symmetric encoder-decoder convolutional network thiab muaj ib tug loj tus naj npawb ntawm feature channels uas tso cai rau rho tawm nta mus rau lub siab dua txheej nyob rau hauv ib tug sib sib zog nqus network. Rov ua dua daim ntawv thov ntawm (3 x 3) convolution kernel tom qab los ntawm ReLU activation, (2 x 2) max-pooling thiab (2 x 2) up-sampling nrog stride loj ntawm 2, thiab (1 x 1) convolution ua raws li sigmoid activation ntawm txheej kawg, tag nrho ntawm 23 txheej hauv lub network. Hauv [36], Veit A et al. paub los ntawm lawv cov kev sim hais tias yog ib lub network muaj ib qho kev sib sau ua ke ces txoj kev luv luv yog txaus thaum lub sij hawm kev cob qhia, los yog txoj kev tob heev tsis tas yuav tsum tau thaum lub sij hawm kev cob qhia. Cov ntau txoj hauv kev no tsis muaj zog nyob ntawm ib leeg thiab lawv txoj kev sib raug zoo nrog ntau txoj hauv kev siv tau ua rau kev ua haujlwm ntawm lub network. Hauv [22], Milletari F et al. tau npaj ib qho encoder-decoder convolutional network rau cov ntaub ntawv peb-dimensional los ntawm kev siv cov tsuav poob raws li kev poob haujlwm. Lawv qhov kev ntsuas empirical ua tiav kev ua tau zoo dua ntawm cov ntaub ntawv tsis sib xws. Hauv [24], Nogues I et al. tau npaj ib qho kev tsim qauv rau kev tshawb pom cov qog ntshav los ntawm ob lub nested saib xyuas kev sib txuas lus sib txuas lus thiab cov txheej txheem ua kom zoo tshaj plaws hauv kev ua haujlwm zoo. Kev txo qis ntawm cov ntaub ntawv hauv lub network sib sib zog nqus tau hais los ntawm Kaiming He li al. nyob rau hauv [9], los ntawm kev qhia ib tug sib sib zog nqus residual network uas yog yooj yim rau cob qhia thiab optimize. Qhov kev sib txuas seem yog pom tau los ntawm kev hla ib lossis ntau txheej kom rov qab tau cov ntaub ntawv ntws hauv lub network sib sib zog nqus. Rau segmentation thiab nrhiav pom cov khoom histological, Chen H li al. hauv [5], tau qhia txog tus qauv contour-paub uas rho tawm cov ntaub ntawv ntau qib hauv kev saib xyuas. Hauv [10], Huang G et al. npaj ib lub network convolutional, uas ntxiv dag zog rau tag nrho cov ntws ntawm cov tswv yim feature daim ntawv qhia los ntawm kev pub cov txheej txheem ua ntej thiab cov tswv yim qub. Lawv qhov kev sim kuj qhia tau hais tias vim yog kev sib koom ua ke ntawm kev qhia txog tus kheej, tus qauv kawm paub ntau dua nta thiab txo qhov teeb meem ploj mus. Nyob rau hauv cov ntaub ntawv ntawm cov ntaub ntawv tsis sib npaug, kev kwv yees yog kev tsis ncaj ncees rau qhov siab thiab qis rov qab uas tsis tuaj yeem zam tau, tshwj xeeb tshaj yog hauv kev kho mob. Qhov teeb meem no yog hais los ntawm Salehi SM thiab al. nyob rau hauv [29], uas tau cob qhia lub sib sib zog nqus network, txawm nrog ib tug heev imbalanced dataset, thiab tuav tau zoo qhov twg tsis muaj tseeb kev kwv yees yog ntau txaus ntshai tshaj qhov tsis zoo zoo. Tus cwj pwm ntawm kev poob haujlwm xws li qhov hnyav hla-entropy thiab cov tsuav poob nrog cov kev kawm sib txawv uas tau soj ntsuam los ntawm Sudre CH li al. hauv [30], ntawm cov duab kho mob thiab cov ntaub ntawv hauv tsev. Lawv qhov kev sim pom tau hais tias raws li qib ntawm qhov tsis sib xws nce qhov sib tshooj ntsuas-raws li kev ua haujlwm poob zoo dua. Ua tau zoo heev ntawm kev nco thiab lub sijhawm rau kev sib cais ntawm txoj kev thiab sab hauv tsev, ib qho kev kos duab-decoder architecture hu ua SegNet los ntawm Badrinarayanan V li al. hauv [3]. SegNet tsim ib qho kev txiav txim siab me me uas txhawb nqa nrog lub pas dej hloov pauv thiab nws cov lus daws teeb meem qis dua los ntawm nws cov encoder. Txhawm rau kom raug ntu ze ntawm thaj tsam ciam teb Zhou S li al. [38], siv ib qho chaw seem nrog lub dilated convolution thaiv. Lawv siv ntau cov hierarchical blocks nyob rau hauv parallel los muab cov ntaub ntawv muaj txiaj ntsig semantic. Txhawm rau daws cov teeb meem tsis txaus ntseeg hauv chav kawm lossis txo cov kev twv tsis tseeb hauv kev kho mob, Hashemi SR hauv [11] tawm tswv yim 3D-dense CNN nrog Tversky index-raws li asymmetric zoo sib xws poob uas cob qhia lub network nrog qhov qis tshaj qhov deb. Qhov teeb meem nyuaj ntawm thaj tsam cuam tshuam txog kev faib tawm los ntawm Naylor P li al. nyob rau hauv [25], los ntawm kev tsim ib tug poob muaj nuj nqi raws li intra- nuclear nrug. Lawv tus qauv encoder-decoder ua tau zoo tshaj FCN, FCN ntxiv rau PP, Mask R-CNN, U-Net, thiab U-Net ntxiv rau PP tau sim nrog TNBC thiab MoNuSeg datasets. Lub ntsiab lus txuas ntxiv nyob rau hauv tus qauv encoder-decoder los ntawm kev sib txuas ntxiv ib qho ntxiv hu ua rooj vag saib xyuas los ntawm Schlemper J li al. nyob rau hauv [31], thiab saib raws li qhov seem mechanism los ntawm Lal S li al. nyob rau hauv [20], qhov twg lub network tau kawm nyob rau hauv xws li ib tug txoj kev uas nws suppresses irrelevance nta thaum highlighting lub ntsiab feature. Rau txoj kev scene segmentation Malekijoo A li al. nyob rau hauv [23], siv tus qauv autoencoder-raws li qhov twg convolution, deconvolution, thiab pyramid pooling tau thov rau kev txhawb zog hauv zos. Rau segmentation ntawm microscopic, MR, thiab CT cov duab encoder-decoder architecture los ntawm Zhou S li al. nyob rau hauv [39], txuas cov ntsiab lus sib txuas kom meej nrhiav cov ciam teb nyuaj. Rau segmentation ntawm nuclei hauv cov duab pathology, Lal S li al. qauv [21], muaj kev hloov xim deconvolution, multiscale thresholding ua raws li kev ua haujlwm morphological, thiab lwm cov kauj ruam tom qab ua tiav. Rau qhov segmentation ntawm cov duab kho mob, ib qho kev poob haujlwm tshiab los ntawm Karimi D li al. hauv [16], kwv yees Hausdorff nrug siv txoj kev ua haujlwm morphological, txoj kev hloov pauv mus deb, thiab cov kab sib txuas sib txuas ntawm cov radii sib txawv. Siv cov kev txo qis ntawm Hausdorff nrug, lawv cob qhia CNN rau ntau yam duab thaij duab thiab sib piv lawv cov txiaj ntsig nrog kev siv feem ntau poob haujlwm. Hanif MS et al. nyob rau hauv [12], thov kom muaj kev sib tw residual network los ntawm stacking ntau residual units hu ua ib tug dav network. Lawv txoj kev tshawb fawb tau xaus lus tias kev ua tau zoo ntawm lub network dav no zoo dua li qhov sib sib zog nqus thiab nyias. Chanchal AK et al. thiab Aatresh AA et al. nyob rau hauv [2, 6], siv sib cais convolution pyramid pooling thiab dimension-wise pyramid pooling rau nuclei segmentation hauj lwm.

Cistanche capsules

Proposed architecture

Rau cov segmentation ntawm microscopy dluab, ib tug encoder-decoder architecture yog qhov zoo tshaj plaws suited vim hais tias yog ib tug encoder muaj ib txwm convolution khaubncaws sab nraud povtseg thiab max-pooling txheej, ces nws captures lub ntsiab lus teb nyob rau hauv daim duab zoo heev. Txoj kev decoder nthuav tawm cov zis los ntawm kev maj mam siv cov qauv, sau cov yam ntxwv cuam tshuam los ntawm tus encoder, thiab ua kom muaj qhov tseeb hauv cheeb tsam. Txhua lub lim dej nyob rau sab encoder ntawm DSREDN network qhia hauv daim duab 1, lees txais cov tswv yim ntawm qhov hloov pauv loj. peb tau thov tsis tu ncua (3 x 3) 2D tus qauv convolution, batch normalization, thiab max-pooling. Txhawm rau kom tsis txhob muaj teeb meem saturation thiab poob ntawm cov ntaub ntawv thaum nkag mus tob rau hauv lub network, peb tau rov qab cov ntaub ntawv qis dua los ntawm kev tsim ib txoj hauv kev ntxiv mus rau txoj hauv kev tseem ceeb ntawm lub network. Ob txoj hauv kev no tsis muaj kev sib raug zoo rau ib leeg thiab nws tsis txhob cuam tshuam cov teeb meem gradient. Rau txhua qhov ntau thiab tsawg lim, tag nrho cov encoder sab ntawm DSREDN network muaj peb lub convolution khaubncaws sab nraud povtseg nyob rau hauv parallel nrog ib tug convoluted txoj kev uas tsom mus rau ntws lub contextual feature nyob rau hauv lub network. Txij li thaum txoj kev ua tau zoo ntawm txoj kev decoder los tsim cov txiaj ntsig kawg yog nyob ntawm kev sau cov ntsiab lus ntawm cov ntsiab lus ntawm sab encoder, peb muaj ib txoj hauv kev sib txawv me ntsis ntawm sab decoder, rau kev ua kom zoo ntawm cov khoom sau. Los ntawm cov txheej txheem no, peb lub network DSREDN dhau los ua qhov dav thiab tob es tsis txhob nyias thiab tob. DSREDN network tau kawm nrog RGB cov duab loj (512 x 512 x 3). Tsib theem ntawm txoj kev encoder muaj tsib qhov sib txawv lim qhov ntau thiab tsawg thiab sib thooj decoder txoj kev muaj xws li (a) 2D convolution ntawm kernel loj (3 x 3) nrog ReLU activation (b) A high-resolution txheej (c) (2 x 2) max -pooling txheej nyob rau hauv encoder txoj kev kom txo cov spatial loj ntawm daim duab thiab coj (2 x 2) up-sampling txheej nyob rau hauv lub decoder sab mus sau contextual feature ntawm encoder sab los ntawm concatenation ua hauj lwm (d) nyob rau theem kawg a (1 x 1 ) convolution yog siv los ua daim ntawv qhia qhov loj (512 x 512 x 16) rau (512 x 512 x 1) nrog sigmoid activation.

Xaus

Daim ntawv no tau hais txog CNN-based architecture hu ua qhov sib sib zog nqus cov txheej txheem encoder-decoder network (DSREDN), uas hais txog ob qhov kev txhawj xeeb loj hauv kev tsis siv neeg nuclei segmentation. Thawj qhov kev txhawj xeeb loj yog txhawm rau txheeb xyuas cov nuclei los ntawm cov duab histopathology uas muaj ntau yam sib txawv nrog ntau cov khoom cuav. Qhov teeb meem no tau hais los ntawm kev qhia lub zog encoder-decoder uas muaj ob txoj hauv kev uas muaj peev xwm cais tawm ntau dua thiab muaj peev xwm rov qab tau cov ntaub ntawv tseem ceeb thiab cov ntaub ntawv compact. Cov kev siv tes hauj lwm siv tau zoo siv zog ntawm cov kev kawm seem ntxiv nrog rau cov txheej txheem encoder-decoder architecture los ntawm kev koom nrog txoj hauv kev dav thiab sib sib zog nqus network uas ntxiv dag zog rau cov yam ntxwv nruab nrab. Peb tau thov kom muaj kev poob haujlwm zoo los ntawm kev ua tib zoo sim thiab tsom mus rau ntu ntu ntawm cov nuclei uas muaj cov ciam teb nyuaj lossis ploj mus uas yog qhov teeb meem thib ob hauv txoj haujlwm segmentation. Peb tau siv qhov kev ua tau zoo tshaj plaws matrices F1- cov qhab nia thiab AJI cov qhab nia los ntawm kev ua thwmsim ntawm peb qhov sib txawv uas muaj rau pej xeem H&E stained histopathological datasets. Cov kev ntsuas tau zoo thiab kev kwv yees thaj tsam nuclear ntawm cov txheej txheem npaj tau zoo dua piv rau cov qauv ntawm lub xeev-of-the-art.

Cistanche ntsiav tshuaj

Txawm hais tias tus qauv npaj tau tsim cov txiaj ntsig zoo, qhov chaw tshwj xeeb tuaj yeem ua kom muaj txiaj ntsig ntxiv los ntawm kev sib koom ua ke ntawm kev ua haujlwm siab ntawm cov khoom siv rho tawm. Tsis tas li ntawd, txoj kev npaj tuaj yeem ua kom dav dav rau kev ua haujlwm ntawm ntau cov duab qauv. Txoj kev tshawb no yog binary segmentation ntawm histopathology dluab, ntawm no peb tsuas segment lub nuclear cheeb tsam. Nyob rau hauv lub neej yav tom ntej, peb tuaj yeem muab cov cheeb tsam nuclear no rau hauv lawv cov sub-types. Ob peb lub tswv yim tshiab ntawm cov duab sib txawv tau tshaj tawm los ntawm Shoeibi A li al. hauv [32, 33], nyob rau hauv uas generative adversarial networks (GANs), recurrent neural networks (RNNs), autoencoders (AEs), convolutional neural networks (CNNs), sib sib zog nqus neural networks (DNNs), thiab lwm yam hybrid tes hauj lwm tau tsim rau automated detection of COVID-19 thiab ntau yam sclerosis. Hauv [18, 34], Khodatars M et al. thiab Sadeghi D et al. piav qhia txog kev siv tau ntawm kev kawm tob rau kev kuaj mob ntawm autism spectrum disorder thiab kuaj pom tus kab mob schizophrenia. Cov piv txwv no qhia txog yuav ua li cas thaj chaw ntawm kev kuaj mob hauv computer tau hloov pauv sai, thiab tseem yuav muaj ntau daim ntawv thov uas tseem tsis tau tsom mus rau.

Yuav Ua Li Cas Cistanchis Txhim Kho Raum Ua Haujlwm

Cistanche yog ib hom tshuaj ntsuab uas tau ntev tau siv hauv cov tshuaj suav tshuaj los txhim kho lub raum ua haujlwm. Nws ntseeg tau tias muaj ntau yam txiaj ntsig vim nws cov khoom nquag, xws li phenylethanoid glycosides thiab iridoids.

Cov kev tshawb fawb qhia tias Cistanche tuaj yeem txhawb nqa lub raum ua haujlwm los ntawm kev txhim kho lub raum cov ntshav khiav, txo cov kev ntxhov siab oxidative, thiab txhim kho cov khoom loj hlob uas txhawb rau lub raum noj qab haus huv. Tsis tas li ntawd, nws tuaj yeem pab tswj cov ntshav siab thiab txo qhov mob, uas yog yam tseem ceeb rau lub raum noj qab haus huv.

Ntxiv mus, Cistanche tau qhia txog qhov muaj peev xwm tiv thaiv lub raum puas tsuaj los ntawm qee yam tshuaj lossis co toxins. Nws tuaj yeem muaj kev tiv thaiv ntawm lub raum los ntawm inhibiting inflammatory teb thiab txo cov cell tuag.

Txawm li cas los xij, nws yog ib qho tseem ceeb uas yuav tsum nco ntsoov tias kev tshawb fawb ntxiv yog xav tau kom nkag siab txog cov txheej txheem thiab kev ua tau zoo ntawm Cistanche hauv kev txhim kho lub raum. Ib yam li txhua yam tshuaj ntsuab, nws raug nquahu kom sab laj nrog tus kws kho mob tsim nyog ua ntej siv nws rau kev kho mob.

Cov ntaub ntawv

1. Albayrak A, Bilgin G (2019) Tsis siv neeg cell segmentation hauv cov duab histopathological ntawm ob-theem superpixel-raws li algorithms. Med Biol Eng Comput 57(3): 653–665

2. Aatresh AA, Yatgiri RP, Chanchal AK, Kumar A, Ravi A, Das D, Raghavendra BS, Lal S, Kini J (2021) Kev kawm paub tob tob nrog qhov dav-ntse pyramid pooling rau nuclei segmentation ntawm histopathology dluab. Comput Med Imaging Graph 93:101975. https://doi.org/10.1016/j.compmedimag.2021.101975

3. Badrinarayanan V, Kendall A, Cipolla R (2017) Segnet: Ib tug sib sib zog nqus convolutional encoder-decoder architecture rau duab segmentation. IEEE Trans Pattern Anal Mach Intell 39(12): 2481-2495

4. Cousty J, Bertrand G, Najman L, Couprie M (2010) Watershed cuts: Thinnings, shortest path forests, and topological watersheds. IEEE Trans Pattern Anal Mach Intell 32(5): 925–939

5. Chen H, Qi X, Yu L, Heng PA (2016) DCAN: sib sib zog nqus contour-paub tes hauj lwm rau lub caj pas segmentation. Computer Vision thiab Pattern Recognition. arXiv:1604.02677v1 [cs.CV]

6. Chanchal AK, Kumar A, Lal S, Kini J (2021) Ua tau zoo thiab muaj zog kev kawm sib sib zog nqus architecture rau segmentation ntawm lub raum thiab lub mis histopathology duab. Computer Electr Eng 92:107177. https://doi.org/10.1016/j.compeleceng.2021.107177

7. Chanchal AK, Lal S, Kini J (2021) High-resolution sib sib zog nqus pauv ASPPU-Net rau nuclei segmentation ntawm histopathology dluab. Int J Comput Assist Radiol Surg. https://doi.org/10.1007/s11548-021-02497-9

8. Gonzalez RC, Woods RE (2006) Digital image processing, 3rd ed. Prentice Hall, New York, Tebchaws USA. ISBN -013168728X

9. He K, Zhang X, Ren S, Sun J (2016) Kev kawm tob tob rau kev paub txog duab. 2016 IEEE Conference on computer vision and pattern recognition (CVPR), Las Vegas, NV. Ib., 770–778. https://doi.org/10.1109/CVPR.2016.90

10. Huang G, Liu Z, Maaten L, Weinberger KQ (2017) Densely txuas convolutional tes hauj lwm. IEEE Conference on computer vision and pattern recognition (CVPR), Honolulu. Ib., 2261–2269. https://doi.org/10.1109/CVPR.2017.243

11. Hashemi SR, Salehi SM, Erdogmus D, Prabhu SP, Warfield SK, Gholipour A (2019) Asymmetric poob kev ua haujlwm thiab sib sib zog nqus sib sib zog nqus sib txuas rau cov duab kho mob tsis txaus ntseeg segmentation: daim ntawv thov rau ntau yam sclerosis lesion detection. In: IEEE Access, vol 7, pp 1721–1735. https://doi.org/10.1109/ACCESS.2018.2886371

12. Hanif MS, Bilal M (2020) Competitive residual neural network rau kev faib cov duab. ICT Express 6(1):28–37. https://doi.org/10.1016/j.icte.2019.06.001

13. Ioffe S, Szegedy C (2015) Batch normalization: accelerating sib sib zog nqus kev cob qhia los ntawm kev txo cov kev hloov hauv covariate. Machine Learning. arXiv: 1502.03167

14. Irshad H, Kouhsari LM, Waltz G, Bucur O, Nowak JA, Dong F, Knoblauch NW, Beck AH (2015) Crowdsourcing image annotation for nucleus detection and segmentation in computational pathology: assessment experts, automated method, and the crowd. Hauv: Pacific Symposium on biocomputing (PSB), pp 294–305. https://doi.org/10.13140/2.1.4067.0721

15. Jadon S (2020) Daim ntawv ntsuam xyuas ntawm kev poob haujlwm rau semantic segmentation. [Online]. Available: arXiv: 2006.14822

16. Karimi D, Salcudean SE (2020) Txo qhov Hausdorff nyob deb ntawm kev kho mob segmentation nrog convolutional neural networks. IEEE Trans Med Imaging 39(2): 499–513

17. Kumar N, Verma R, Sharma S, Bhargava S, Vahadane A, Sethi A (2017) A dataset and a technique for generalized nuclear segmentation for computational pathology. IEEE Trans Med Imaging 36(7): 1550–1560

18. Khodatars M, Shoeibi A, Sadeghi D, Ghaasem N, Jafari M, Meridian P, Khadem A, Alizadehsani R, Zare A, Kong Y, Khosravi A, Nahavandi S, Hussain S, Acharya UR, Berk M (2021) Deep Kev kawm rau kev kuaj mob neuroimaging thiab kho dua tshiab ntawm Autism Spectrum Disorder: Kev tshuaj xyuas. Comput Biol Med 139:104949. https://doi.org/10.1016/j.compbiomed.2021.104949

19. Kingma DP, Ba J (2015) Adas: Ib txoj hauv kev rau kev ua kom zoo dua qub. Hauv: Lub rooj sablaj thoob ntiaj teb ntawm kev sawv cev rau kev kawm, vol 9. arXiv:1412.6980v9 [cs.LG]

20. Lal S, Das D, Alabhya K, Kanfade A, Kumar A, Kini J (2021) NucleiSegNet: Sib sib zog nqus kawm architecture rau nuclei segmentation ntawm daim siab mob qog noj ntshav histopathology. Comput Biol Med 128:104075

21. Lal S, Kanfade A, Alabhya K, Dsouza R, Kumar A, Chanchal AK, Maneesh M, Peryail G, Kini J (2020) Ib txoj hauv kev muaj zog rau nuclei segmentation ntawm H&E stained histopathology dluab. 7th IEEE lub rooj sib tham thoob ntiaj teb ntawm kev teeb tsa thiab kev sib koom ua ke (SPIN2020), Amity University Delhi NCR, Noida, UP

22. Milletari F, Navab N, Ahmadi SA (2016) V-Net: tag nrho convolutional neural networks rau volumetric kho mob duab segmentation, plaub lub rooj sib tham thoob ntiaj teb ntawm 3D vision (3DV). Stanford, CA. Ib., 565–571. https://doi.org/10.1109/3DV.2016.79

23. Malekijoo A, Fadaeieslam MJ (2019) Convolution-deconvolution architecture nrog lub pyramid pooling module rau semantic segmentation. Multimed Tools Appl 78:32379–32392. https://doi.org/10.1007/s11042-019-07990-7

24. Nogues I et al (2016) Automatic lymph node cluster segmentation using holistically-nested neural networks and structured optimization in CT images. Nyob rau hauv: Kho mob duab xam thiab computer pab cuam – MICCAI 2016. Lecture notes in computer science, vol 9901. Springer, Cham. https://doi.org/10.1007/978-3-319-46723-845

25. Naylor P, Lae M, Reyal F, Walter T (2019) Segmentation of nuclei in histopathology images by deep regression of the distance map. IEEE Trans Med Images 38(2): 448–459

26. Ronneberger O, Fischer P, Brox T (2015) U-Net: Convolutional tes hauj lwm rau biomedical duab segmentation. Hauv: Proc. MICCAI. Springer, Munich, Lub teb chaws Yelemees, pp 234-241

27. Slaoui M, Fiette L (2011) Histopathology cov txheej txheem: los ntawm cov ntaub so ntswg mus rau kev ntsuam xyuas histopathological. Methods Mol Biol (Methods Protoc) 691:69–82

28. Nkauj T, Sanchez V, EIDaly H, Rajpoot NM (2017) Dual-channel active contour qauv rau megakaryocytic cell segmentation nyob rau hauv cov pob txha pob txha trephine histology dluab. IEEE Trans Biomed Eng 64(12): 2913–2923

29. Salehi SM, Erdogmus D, Gholipour A (2017) Tversky poob muaj nuj nqi rau cov duab segmentation siv 3D tag nrho convolutional sib sib zog nqus tes hauj lwm. Hauv: Proc, Int Rhiav Mach Kawm Med Imag. Springer, Cham, Switzerland, pp 379–387

30. Sudre CH, Li W, Vercauteren T, Ourselin S, Cardoso MJ (2017) Generalized dice overlap as a deep learning poob function for highly unbalanced segmentations. Nyob rau hauv: Sib sib zog nqus kev kawm nyob rau hauv kev kho mob duab tsom xam thiab multimodal kev kawm rau kev kho mob txiav txim txhawb. Ib., pp 240–248

31. Schlemper J, Oktay O, Schaap M, Heinrich M, Kainz B, Glocker B, Rueckert D (2019) Kev saib xyuas cov tes hauj lwm: Kawm paub siv cov cheeb tsam tseem ceeb hauv cov duab kho mob. Med Image Anal 53(ISSN 1361- 8415): 197–207

32. Shoeibi A, Khodatars M, Alizadehsani R, Ghassemi N, Jafari M, Meridian P, Khadem A, Sadeghi D, Hussain S, Zare A, Sani ZA, Bazeli J, Khozeimeh F, Khosravi A, Nahavandi S, Acharya UR, Shi P (2020) Automated detection and forecasting of covid-19 siv cov kev kawm tob: kev tshuaj xyuas. Machine Learning. arXiv: 2007.10785 [cs.LG]

33. Shoeibi A, Khodatars M, Jafari M, Meridian P, Rezaei M, Alizadehsani R, Khozeimeh F, Gorriz JM, Heras J, Panahiazar M (2021) Kev siv cov txheej txheem kev kawm tob rau automated multiple sclerosis detection using magnetic resonance imaging: A tshuaj xyuas, duab, thiab ua yeeb yaj kiab. arXiv: 2105.04881

34. Sadeghi D, Shoeibi A, Ghassemi N, Meridian P, Khadem A, Alizadehsani R, Teshnehlab M, Gorriz JM, Nahavandi S (2021) Ib qho kev piav qhia ntawm cov tswv yim txawj ntse rau kev kuaj mob schizophrenia raws li kev sib nqus resonance imaging modalities: , kev nyuaj siab, thiab tej hauj lwm yav tom ntej. Machine Learning. arXiv: 2103.03081

35. Sugino T, Kawase T et al (2021) Poob qhov hnyav rau kev txhim kho lub hlwb tsis zoo segmentation siv cov kev sib txuas lus sib txuas, kev kho mob. PAB 9(8):938

36. Veit A, Wilber M, Belongie S (2016) Residual networks behave like ensembles of also shallow networks. Neural Inf Process Syst pp 550–558. arXiv: 1605.06431

37. Win KY, Choomchuay S, Hamamoto K (2017) K txhais tau hais tias kev sib koom ua ke raws li kev siv cov segmentation ntawm overlapping cell nuclei hauv pleural effusion cytology dluab. Lub rooj sib tham thoob ntiaj teb ntawm cov thev naus laus zis rau kev sib txuas lus (ATC). Ib., 265–269. https://doi.org/10.1109/ATC.2017.8167630

38. Zhou S, Nie D, Adeli E, Gao Y, Wang L, Yin J, Shen D (2018) Fine-grained segmentation using hierarchical dilated neural networks. Nyob rau hauv: Kho duab duab xam thiab computer assisted intervention, vol 11073. Springer, Cham, pp 488-496

39. Zhou S, Nie D, Adeli E, Yin J, Lian J, Shen D (2020). Hauv: IEEE Transactions on Image Processing, vol 29, pp 461–475. https://doi.org/10.1109/TIP.2019.2919937

Amit Kumar Chanchal 1 · Shyam Lal 1 · Jyoti Kini 2

1 Department of Electronics thiab Communication Engineering, National Institute of Technology Karnataka, Surathkal, Mangaluru-575025, Karnataka, India

2 Department of Pathology, Kasturba Medical College Mangalore, Manipal Academy ntawm Kev Kawm Qib Siab, Manipal, India