JFIFPhotoshop 3.08BIM?Explainable Biomedical Foundation Model Learns Concepts Through720260817PSCIENMAGUAuthorgExplainable Biomedical Foundation Model Learns Concepts Through Large-Scale Vision-Language Pretraining Science@https://scienmag.com/explainable-biomedical-foundation-model-lea© 2026 Sciencehttps://scienmag.com/Medical artificial intelligence has reached a point where being right is no longer enough. In hospitals, clinicians must also understand why an algorithm produced a diagnosis, particularly when the decision could influence surgery, cancer treatment, or emergency care. A new study introduces ConceptCLIP, a biomedical foundation model designed to combine high diagnostic performance with explanations […]ScienceMedMedicineMedicine7combining diagnostic accuracy with explainability in AI.ConceptCLIP for medical image interpretability.deep learning for clinical concept recognition,enhancing transparency in medical AI systems@explainable AI for surgical and cancer treatment decision suppor(Explainable biomedical foundation models1interpretable artificial intelligence in medicine5large-scale vision-language pretraining in healthcare+linking medical images to clinical conceptsmultimodal biomedical AI models'visual evidence-based medical diagnosis'visual-textual medical data integrationC    !'"#%%%),($+!$%$C   $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$"=!1AQa"2q#BR34Sbr$6c8!1AQ"2aqBRr#35s ?m0 eҚ:P]=O.hWPQIu!p 8fV ۭ.(y,.RKjY4TI*>:jZlnԀ}*^R%H DYqI$oLZj+IH%amȅǯ_ I-`8}Q7dчeɶ{h~wg lYHp7H{3P.McքQؐQ{,|5XXbt5Iba=t!st!E|z\8~jc-]Lj7u? (pwP[SZެn3Q\uupQmN )g VuռM0 ׵▘J*䃃8Rq҆謈i-ZP2jE[JAP% Pȫ,WJ'Sc@i`vu袒YDQY 9Gxdv7  D HiJg8ACMMCZՒʕUn<+Kn#JF8N>$_;cd{7sĞCq_Lba 2Up8Ijj*F#tZǏIGDu+d Wmt;T0{ɰ_nJcЏja7Hs.4%_toi'E73%|f^/ / 3lfĦX[I(*v6rP|7Yۇ )`'Qq)s#6V˽#nBwZ艬l#]2KxznhǮwI.J%M:G܊FPNr0JItdR;evRa%:3~} er4 Ēq׌xpْ欤:=t"œ\#X-=WF>/q &^rҴ8H'G,u1KTN6˭uf0__ajmcSe5=-mVQnu\дKj;+;juƄ$=lk13 [YN&w` &^ٖb9-*Si)RPIhPyµ!ĕ6@;ήD~Z>]HPP |pMUΊ2!칾3,5%ͱ}$_C&l r6Y @Po'$NJ~44Ϧ@ hHJ8cVK&Lxd!MIOjMnuY>)?҄f.\Y})QG=캠V jw; (ae=IV\*;(W8Zxy@VѸ.zFlk[| \BD~'*v_64 +V䎛Vw'BmčC?ʯ";ɖ􇛟'K =>q ԬmWX|rtn _] +}e¹ۡ (c=z^ۑv#+Sҝ!Ԥqބ87O9i%(P35CEP.uKЖ BT dk) '1f+#33u"y[WRA^yfZHu Nr;~?v)Xa$?J|4۶Vq'|kV)[M#!ࡂOʙA- ߰&R$'GCbSKq'=iM<%+Yb5[*U/%Ҡm,TRaUOCR!J3Qep M@^*DBx^hx5k^\\HTE:C R6Jz戌ԢYx%&fi Aѽe<ǚ췦]"Lrֲ%Z;dtcj>\a{NzT|!hv91mJPKW5J2HҢW )rm}vՆmui,kj,AËfKn!9 רgSȔ2Xd$xz=>,d1v낟1\ϔ\qHv.iO6ӥXγJ\U+u9PQ䘲CN+P_ DTzC<=ԥ||V :yFߝ ~S)o6NFGu6qDPBR[ĶʒOfD5;*q.RHGڭb O|/s`&ß2tRßroKL7Q*V+iNBnp9,u|"mi|ܟF|jFsNw8YcX,53EKGyMrITeJmӍG*oi#RPOZ;m= VzmXoJFզ@og?%)kΥe(bbt#xiq\Υ^b0OvJI"pRAVT}kOߪw5)T_MUiamʖ+r0i%]*E$Ҡd]a֐G )k;T2; W\UrbKq+Jլwuo6Jӿֲja2SNi soWɇ1tUָJx1ףii J;(8c!5uI᫣sG2.@q]!zCl#)!j?@XH#{q_BkMqrI%ArTDӊ.#p|ȦiZP} <ZD 䚻q{8ZrRʒ1&,ұmM Ma0qAranDKG)c'>Z-lJ ~c{VN8z:fy[q)Ze IHN+9_W%.pV.h!1(U|Gi Yujlj胅Ϗ7J֕;x$O)} l) 4F<-}\$b@^ZL_,23]Hq*iN jo>r^D\VDuW(UVNhG(։/BiP#EE$\GZ(SVROTJ**/`hs J׷]*-Lj,GLxӛ 4@Xa7(W|nuAI$~U`9 QvLB^w_޺66z +{046>&1d )$)AV6~,BY[MtPwܮ- 9ʳj;%;Q B=Bԅ%ZTr(sӽ݊ЋeH2^b=Jqb|7NE)W5Q8HTQ}UXϻJ ʏJT@tTۺ[]?GKwb=}ROU; #EWz7wwǧJs4PK9X2hөX1 •*#n `SwJ7M=6iRF