Kuvandudza Kutarisirwa Kwevarwere Nekugadziriswa Kwekuderera Kwesimba reAI muPET Imaging

Chidzidzo chitsva chine musoro unoti "Utilizing Pix-2-Pix GAN for Deep Learning-Based Whole-Body PSMA PET/CT Attenuation Correction" chakaburitswa munguva pfupi yapfuura muVhoriyamu 15 yeOncotarget musi wa7 Chivabvu, 2024.

Kuonekwa kwemwaranzi kubva muongororo dzakatevedzana dzePET/CT mukutevera varwere vekenza kuri kukonzera kunetseka. Muongororo iyi yazvino, timu yevaongorori vanosanganisira Kevin C. Ma, Esther Mena, Liza Lindenberg, Nathan S. Lay, Phillip Eclarinal, Deborah E. Citrin, Peter A. Pinto, Bradford J. Wood, William L. Dahut, James L. Gulley, Ravi A. Madan, Peter L. Choyke, Ismail Baris Turkbey, naStephanie A. Harmon vekuNational Cancer Institute kuNational Institutes of Health vakaunza chishandiso che artificial intelligence (AI). Chishandiso ichi chine chinangwa chekugadzira mifananidzo ye attenuation-corrected PET (AC-PET) kubva kumifananidzo ye non-attenuation-corrected PET (NAC-PET), zvingangoderedza kudiwa kwe CT scans shoma.

"Mifananidzo yePET yakagadzirwa neAi ine mukana wekuderedza kudiwa kwekugadziriswa kweCT scans ukuwo ichichengetedza zviratidzo zvehuwandu uye hunhu hwemifananidzo yevarwere vekenza yeprostate."

Nzira: Algorithm yekudzidza zvakadzama yakavakirwa pa2D Pix-2-Pix generative adversarial network (GAN) yakagadzirwa zvichibva pamifananidzo yeAC-PET neNAC-PET. Chidzidzo che18F-DCFPyL PSMA (Prostate-specific membrane antigen) PET-CT chevarwere 302 vane kenza yeprostate chakakamurwa kuva mapoka ekudzidzisa, kusimbisa, uye ekuyedza (n 183, 60, uye 59, zvichiteerana). Modhi iyi yakadzidziswa uchishandisa nzira mbiri dzakajairwa: Standard Uptake Value (SUV) yakavakirwa uye SUV-NYUL yakavakirwa. Scanning horizontal performance yakaongororwa uchishandisa normalized mean square error (NMSE), mean absolute error (MAE), structural similarity index (SSIM) uye peak signal-to-noise ratio (PSNR). Chiremba weNuclear medicine akaita ongororo ye lesion level yenzvimbo yaidiwa. Zviratidzo zveSUV zvakaongororwa uchishandisa intra-group correlation coefficient (ICC), repeatability coefficient (RC), uye linear mixed effects models.

Mhedzisiro：Muboka rekuzviongorora rakazvimirira, avhareji yeNMSE, MAE, SSIM, nePSNR yaive 13.26%, 3.59%, 0.891, uye 26.82, zvichiteerana. ICC yeSUVmax neSUV mean yaive 0.88 ne0.89, zvichiratidza hukama hwakasimba pakati pezviratidzo zvekutanga neAI-generated quantitative imaging. Zvinhu zvakaita senzvimbo yekukuvara, huwandu (Hounsfield units), uye kutorwa kwekukuvara zvakawanikwa zvichikanganisa kukanganisa kwekuenzanisa muSUV metrics dzakagadzirwa (ese p < 0.05).

"AC-PET yakagadzirwa nePix-2-Pix GAN modhi inoratidza zviyero zveSUV zvinoenderana nemifananidzo yekutanga. Mifananidzo yePET yakagadzirwa neAI inoratidza mukana wekurapa unovimbisa wekuderedza kukosha kweCT scans yekugadzirisa attenuation uku uchichengetedza mamaki ehuwandu uye mhando yemufananidzo."

—— ...

Sezvatinoziva tese, kukura kweindasitiri yemifananidzo yezvokurapa hakuparadzaniswi nekugadzirwa kwemidziyo yekurapa yakatevedzana - majekiseni ekudzivirira emhando dzakasiyana-siyana uye zvinodyiwa zvinotsigira - zvinoshandiswa zvakanyanya mundima iyi. MuChina, iyo inozivikanwa neindasitiri yayo yekugadzira, kune vagadziri vazhinji vanozivikanwa mudzimba nekune dzimwe nyika nekugadzira michina yemifananidzo yekurapa, kusanganisiraLnkMedKubva payakavambwa, LnkMed yanga ichinyanya kutarisana nekushandisa majekiseni ekudzivirira kupisa kwemhepo ane simba guru. Chikwata cheinjiniya cheLnkMed chinotungamirirwa nePhD. chine ruzivo rwemakore anopfuura gumi uye chakabatanidzwa zvakanyanya mukutsvaga nekuvandudza. Vachitungamirirwa naye,jekiseni reCT remusoro mumwe chete,jekiseni reCT remusoro miviri,Injector yeMRI contrastuyeInjector yeAngiography ine pressure yakakwiraYakagadzirwa nezvinhu izvi: muviri wayo wakasimba uye mudiki, ine hungwaru uye inoshanda zvakanaka, ine mabasa akazara, yakachengeteka zvikuru, uye ine dhizaini yakasimba. Tinogonawo kupa masirinjisi nemachubhu anoenderana neaya majekiseni eCT, MRI, DSA ane mukurumbira. Nehunhu hwavo hwechokwadi uye simba ravo rehunyanzvi, vashandi vese veLnkMed vanokukokai nemwoyo wese kuti muuye kuzoongorora misika yakawanda pamwe chete.