AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma

AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma View ORCID Profile Mark Schuiveling , View ORCID Profile Isabella A.J. van Duin , Laurens ter Maat , Janneke C. van der Weerd , View ORCID Profile Rik J. Verheijden , View ORCID Profile Franchette van den Berkmortel , View ORCID Profile Christian U. Blank , View ORCID Profile Gerben E. Breimer , View ORCID Profile Femke H. Burgers , View ORCID Profile Marye Boers-Sonderen , Alfons J. M. van den Eertwegh , View ORCID Profile Jan Willem B. de Groot , View ORCID Profile John B.A.G. Haanen , View ORCID Profile Geke A.P. Hospers , View ORCID Profile Ellen Kapiteijn , View ORCID Profile Djura Piersma , View ORCID Profile Gerard Vreugdenhil , View ORCID Profile Hans Westgeest , View ORCID Profile Anne M.R. Schrader , View ORCID Profile Josien Pluim , View ORCID Profile Paul J. van Diest , View ORCID Profile Mitko Veta , View ORCID Profile Karijn P.M. Suijkerbuijk , View ORCID Profile Willeke Blokx doi: https://doi.org/10.1101/2025.05.28.25328410 Mark Schuiveling 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Mark Schuiveling For correspondence: m.schuiveling{at}umcutrecht.nl Isabella A.J. van Duin 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Isabella A.J. van Duin Laurens ter Maat 2 Image Sciences Institute, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Janneke C. van der Weerd 3 Department of Pathology, University Medical Center Groningen, University of Groningen, Hanzeplein 1 , Groningen 9713 GZ , the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Rik J. Verheijden 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Rik J. Verheijden Franchette van den Berkmortel 4 Department of Medical Oncology, Zuyderland Medical Center Sittard, Dr. H. van der Hoffplein 1, Sittard-Geleen 6162 BG , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Franchette van den Berkmortel Christian U. Blank 5 Department of Medical Oncology, Netherlands Cancer Institute, Plesmanlaan 121, 1066CX Amsterdam, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Christian U. Blank Gerben E. Breimer 6 Department of Pathology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gerben E. Breimer Femke H. Burgers 5 Department of Medical Oncology, Netherlands Cancer Institute, Plesmanlaan 121, 1066CX Amsterdam, the Netherlands MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Femke H. Burgers Marye Boers-Sonderen 7 Department of Medical Oncology, Radboud University Medical Center, Geert Grooteplein Zuid 10 , Nijmegen 6525 GA , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Marye Boers-Sonderen Alfons J. M. van den Eertwegh 8 Department of Medical Oncology , Amsterdam UMC, VU University Medical Center, Cancer Center Amsterdam, De Boelelaan 1118 , Amsterdam 1081 Hz , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jan Willem B. de Groot 9 Isala Oncology Center, Isala, Dokter van Heesweg 2, Zwolle 8025 AB , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jan Willem B. de Groot John B.A.G. Haanen 5 Department of Medical Oncology, Netherlands Cancer Institute, Plesmanlaan 121, 1066CX Amsterdam, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for John B.A.G. Haanen Geke A.P. Hospers 10 Department of Medical Oncology, University Medical Center Groningen, University of Groningen, Hanzeplein 1 , Groningen 9713 GZ , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Geke A.P. Hospers Ellen Kapiteijn 11 Department of Medical Oncology, Leiden University Medical Center, Albinusdreef 2 , Leiden 2333ZA , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ellen Kapiteijn Djura Piersma 12 Department of Internal Medicine, Medisch Spectrum Twente, Koningsplein 1, Enschede 7512 KZ , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Djura Piersma Gerard Vreugdenhil 13 Department of Internal Medicine, Maxima Medical Center, De Run 4600, Eindhoven 5504 DB , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gerard Vreugdenhil Hans Westgeest 14 Department of Internal Medicine, Amphia Hospital, Molengracht 21 , Breda 4818 CK , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Hans Westgeest Anne M.R. Schrader 15 Department of Pathology, Leiden University Medical Center, Albinusdreef 2 , Leiden 2333 ZA , the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Anne M.R. Schrader Josien Pluim 16 Medical Image Analysis, Department of Biomedical Engineering, Eindhoven University of Technology , Eindhoven, the Netherlands. PO Box 513, 5600 MB Eindhoven PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Josien Pluim Paul J. van Diest 6 Department of Pathology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Paul J. van Diest Mitko Veta 16 Medical Image Analysis, Department of Biomedical Engineering, Eindhoven University of Technology , Eindhoven, the Netherlands. PO Box 513, 5600 MB Eindhoven PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Mitko Veta Karijn P.M. Suijkerbuijk 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Karijn P.M. Suijkerbuijk Willeke Blokx 6 Department of Pathology, University Medical Center Utrecht, Utrecht University, Heidelberglaan 100, 3584 CG Utrecht, the Netherlands MD PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Willeke Blokx Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Importance Easy and accessible biomarkers to predict response to immune checkpoint inhibition (ICI)-treated melanoma are limited. Objective To evaluate artificial intelligence (AI) detected tumor-infiltrating lymphocytes (TILs) on pretreatment melanoma metastases as a biomarker for response and survival in ICI-treated patients. Design Multicenter cohort study including patients with advanced melanoma treated with first-line anti-PD1 ± anti-CTLA4 between 2016 and 2023. Median follow-up was 36.3 months. Setting 11 melanoma treatment centers in the Netherlands. Participants 1,202 patients with advanced cutaneous melanoma. Exposure All patients received first-line anti-PD1 ± anti-CTLA4. Main Outcome(s) and Measure(s) The percentage of TILs inside manually annotated tumor area in H&E stained pretreatment metastases was determined using the Hover-NeXt model trained and evaluated on an independent melanoma dataset containing 166,718 pathologist-verified manually annotated cells. The primary outcome was objective response rate (ORR); secondary outcomes were progression-free survival (PFS) and overall survival (OS). Correlation with manual TILs, scored according to the guidelines stated by the immune-oncology working group, was evaluated with Spearman correlation coefficients. Logistic regression and Cox proportional regression were conducted, adjusted for age, sex, disease stage, ICI type, BRAF status, brain metastases, LDH level, and performance status. Results Metastatic melanoma specimens were available for 1,202 patients, of whom 423 received combination therapy. Median TIL percentage was 9.9% (range 0.3% - 69.4%). A 10% increase in TILs was associated with increased ORR (adjusted OR 1.40 [95% 1.23-1.59]), PFS (adjusted HR 0.85 [95% CI 0.79 – 0.92]) and OS (adjusted HR 0.83 [95% CI 0.76 – 0.91]. Results were consistent for both patients treated with anti-PD1 monotherapy and combination treatment with anti-PD1 plus anti-CTLA4. When comparing manual TIL scoring with AI-detected TILs, associations with response and survival were consistently stronger for AI-detected TILs. Conclusions and Relevance In patients with advanced melanoma, higher levels of AI-detected TILs on pre-treatment H&E slides were independently associated with improved ICI response and survival. Given the accessibility of TIL scoring on routine histology, TILs may serve as a predictive biomarker for ICI outcomes. To facilitate broader validation, the Hover-NeXt architecture and model weights are publicly available. Key points Question: What is the predictive value of artificial intelligence-detected tumor-infiltrating lymphocytes (TILs) for clinical outcomes in patients with advanced melanoma receiving first-line immune checkpoint inhibition? Findings: In this multicenter cohort of 1202 patients, TILs in pretreatment metastases were quantified using a melanoma-specific publicly available AI model trained on an independent dataset. A 10% increase was associated with response (aOR 1.40 [95% CI 1.23–1.59]), progression-free survival (aHR 0.85 [95% CI 0.79–0.92]), and overall survival (aHR 0.83 [95% CI 0.76–0.91]). Associations were independent of clinical predictors. Meaning: AI-detected TILs in pretreatment melanoma metastases independently correlate with response and survival. Introduction Immune checkpoint inhibition (ICI) has significantly improved the prognosis for patients with advanced cutaneous melanoma. In the Checkmate 067 trial, ten-year overall survival (OS) rates of 37% for anti-PD1 monotherapy and 43% for combination therapy with anti-CTLA4 have been reported. 1 However, significant challenges remain. Despite not all patients responding to therapy, many still experience side effects, with severe toxicity occurring in up to 60% of patients receiving combination treatment 2 , 3 . Furthermore, ICIs place a substantial financial burden on the healthcare system, with average medication costs per patient reaching €60,000 for monotherapy and €117,000 for combination therapy in The Netherlands. 4 , 5 Therefore, there is a need for predictive biomarkers that can identify patients most likely to benefit from ICI treatment. Our previous research identified tumor-infiltrating lymphocytes (TILs) as such a potential biomarker. In a cohort of 650 patients, we showed that the percentage of intra-tumoral stroma occupied by TILs, which is the scoring method defined by the International TILs Working Group 6 , was associated with both treatment response and survival outcomes. 7 However, it also became clear that manual scoring is hindered by substantial interobserver variability. 7 – 9 Artificial intelligence (AI) offers a scalable solution by enabling reproducible and quantitative assessment of TILs on routine H&E slides. While AI-based TIL quantification has demonstrated prognostic value in multiple tumor types 10 – 12 , evidence in melanoma remains inconsistent. Prior studies in advanced melanoma cohorts reported conflicting associations between AI-detected TILs and clinical outcomes, likely due to methodological variability, small cohort sizes, and limited model generalizability 13 – 16 . To address these challenges, we developed an open-source, melanoma-specific AI model to detect TILs on H&E-stained slides 16 . In this study, we applied this model to pre-treatment metastatic melanoma samples to evaluate whether AI-detected TILs are associated with ICI response and survival in a real-world cohort of patients with advanced disease. Methods TIL model and developmental cohort The TIL detection model used in this study was trained and evaluated on an independent developmental cohort consisting of 155 unique primary and 197 unique metastatic melanoma samples. This dataset included 161,835 nuclei annotations verified by a dermatopathologist (author W.A.M.B.). Part of the developmental dataset, along with the cell detection model and the melanoma-specific model weights, are publicly available. 17 , 18 Further details on the construction of the developmental cohort and the model training process are provided in the data supplements. 16 , 19 , 20 Clinical cohort Patients were retrospectively identified from eleven melanoma treatment centers using high quality prospectively collected registry data. 21 Patients were included if above 18 years of age and treated with first-line anti-PD1 ± anti-CTLA4 for unresectable stage IIIC or stage IV melanoma after January 1, 2016 until the 1 st of January 2023. The patient’s stage of disease was based on the 8 th edition of the AJCC melanoma staging system. 22 For application of the model to the patient cohort, the tumor area excluding large areas of stroma and necrosis was delineated on the digitized specimen. This was performed manually (by authors J.v.d.W. and M.S., blinded for the outcome) under the supervision of a dermatopathologist (author W.A.M.B.). 23 – 25 The TIL score was defined as the percentage of AI-detected TILs among all cells detected within the annotated tumor area. A visual representation of the TIL scoring is shown in Figure 1 and an example output of the model is visualized in Supplementary Figure 1. Download figure Open in new tab Figure 1. Shown in the panels are a digitized whole slide image of a melanoma biopsy taken from a metastasis in a lymph node (A), the manually annotated tumor area (B), the output of the cell detection model filtered to the annotated tumor area (C), and a zoomed-in view of the cell detection results (D) with one zoomed in region with annotated tumor-infiltrating lymphocytes and one zoomed in region with tumor cells and apoptotic cells segmented as other. From all cells present, the percentage of tumor-infiltrating lymphocytes for the tumor area is calculated as: Slide selection Slides were collected from 30 different hospital archives through PALGA, the Dutch nationwide registry of histo- and cytopathology. 26 For each patient, a single representative H&E-stained slide was selected. In cases with multiple specimens, the largest available specimen before therapy initiation was used. Curation and selection were done by four authors (M.S., I.A.J.v.D., L.t.M., and J.v.d.W.) under supervision of two experienced pathologists (W.A.M.B. and P.J.v.D.). All selected slides were scanned with a Nanozoomer XR C12000-21/-22 (Hamamatsu Photonics, Hamamatsu, Shizuoka, Japan) at 40× magnification with a resolution of 0.23 µm per pixel. The anatomic location of the metastatic lesion was derived from the pathology report if described or assessed visually on the slide when possible (in case of presence of pre-existing normal tissue of the organ). Determination of the nature of the specimen type (resection or biopsy) was visually assessed. Manual TIL classification Manual TIL scoring was performed by authors I.A.J.v.D. and M.S., under the supervision of two experienced pathologists (W.A.M.B. and P.J.v.D.). TILs were scored using the semi-quantitative method proposed by the International Immuno-Oncology Biomarker Working Group. This method estimates the percentage of TILs present within the intra-tumoral stromal area. 6 Clinical variables Clinical data included age at treatment initiation, sex, ICI type, World Health Organization (WHO) performance status, BRAF V600E mutation status, serum lactate dehydrogenase (LDH) level, AJCC disease stage, and presence of symptomatic brain metastases. We incorporated the variable ‘presence of symptomatic brain metastases’ in the variable ‘stage of disease’: M1D stage was subcategorized either as ‘M1D-non-symptomatic’ or ‘M1D-symptomatic’. LDH levels were categorized as normal, 1-2 times the upper limit of normal (ULN) or twice the upper limit of normal. WHO performance status was categorized as 0, 1, or 2 and above. Outcome measures Response evaluation was determined by the treating physician and was based on the Response Evaluation Criteria in Solid Tumors, version 1.1, with melanoma-related death before first response assessment classified as progressive disease. 27 Primary outcome was objective response, defined as the best overall response (partial or complete response) within 6 months. Secondary outcomes were progression-free survival (PFS) and overall survival (OS). Missing data To account for missing covariate data, we used multiple imputation by chained equations (MICE) under the assumption that the data were Missing At Random (MAR). The mice package (version 3.15) in R was used to generate 100 imputed datasets over 20 iterations, which is by default more than the proposed minimum of datasets (the percentage of incomplete cases). Convergence of the imputation model was verified visually using trace plots. Statistical analysis Continuous variables were summarized using median and range or interquartile range (IQR), and categorical variables with frequencies and percentages. Associations between continuous variables and response were assessed using the Mann-Whitney U test, assuming non-normal distribution. Correlations were evaluated with Spearman correlation coefficients and visualized using scatter plots. Median follow-up was estimated using the reversed Kaplan-Meier method. PFS and OS were visualized with Kaplan-Meier curves, and differences between groups were assessed using the logrank test. Univariable and multivariable analyses were conducted using logistic regression and Cox proportional hazards regression. Model assumptions were visually checked and showed no violations. Multivariable analyses were performed on imputed datasets, with results pooled using Rubin’s rules. Unless stated otherwise, all estimates are based on these pooled analyses. Statistical analyses were performed using R (version 4.2.2; survival package version 3.5.0). Results AI-detected TILs model performance Our developed cell detection model achieved an F 1 score of 0.79 [95% CI 0.77 – 0.80] for TIL detection, closely approaching the interobserver agreement on a subset of the developmental dataset (F 1 score of 0.83 [95% CI 0.79 – 0.88]). Visual inspection revealed no remaining major misclassifications. Examples of the model’s output on the clinical dataset are displayed in Supplementary Figure 1. Patient characteristics and outcomes Of the 1935 eligible patients, 44 were excluded due to missing survival outcome data. From the remaining 1891 patients, metastatic specimens with viable tumor area were available for 1202 patients (Supplementary Figure 2). Patient characteristics are shown in Table 1 and compared well to those of excluded patients (Supplementary Table 1). Most patients were male, had non-elevated LDH levels, and were above 65 years of age. Lymph nodes and skin were identified as the most common site of origin for the metastatic specimens. In patients treated with combination therapy, the analyzed specimen was more often a brain or liver metastasis. View this table: View inline View popup Download powerpoint Table 1. Patient characteristics of 1202 ICI-treated patients with pre-treatment metastatic specimen available. Abbreviations: IQI, interquartile interval; LDH, Lactate dehydrogenase; TIL, tumor-infiltrating lymphocytes; ULN, upper limit of normal; WHO, World Health Organization. The median follow-up was 36.3 months with a median PFS of 9.0 months and a median OS of 35.0 months. The objective response rate (ORR) to ICI was 56.7%. The median average TIL percentage was 9.9 [range 0.3% - 69.4%]. In patients receiving anti-PD1 therapy, the median average percentage was 10.5 [range 0.3% - 69.4%], while in those treated with combination therapy, it was 8.6% [range 0.4% - 66.3%]. Higher TIL percentages were observed in lymph node (12.8% [range 1.2% - 69.4%]) and lung (11.3% [range 0.4% - 53.5%]) metastases, while liver metastases had lower TIL levels (7.3% [range 0.8% - 26.9%]; Supplementary Table 2). AI-detected TILs and response Patients who responded to therapy had significantly higher median average TIL levels (10.9% [IQR 6.1% - 18.6%] when compared to non-responders 8.3% [IQR 4.6% - 14.8%] P <.001). When stratified to treatment type, this effect remained significant in both patients treated with anti-PD1 monotherapy (12.4% [IQR 6.9% - 20.7%] vs 8.6% [IQR 5.0% - 16.3%] P <,001), and in patients treated with anti-PD1 and anti-CTLA4 combination therapy (9.5% [IQR 5.4% - 14.9%]%] vs 7.6% [IQR 4.2% - 13.6%]] P =.049, Supplementary Figure 3). Upon dividing TIL levels into tertiles, ORR was 46.4% in the lowest tertile, increasing to 56.4% in the middle and 63.7% in the highest tertile. In the anti-PD1 monotherapy subgroup, ORRs increased from 43.8% to 55.8% and 64.9% across the tertiles, whereas in the combination therapy group, the rates were 50.4%, 61.7%, and 58.2% in the low, middle, and highest tertiles, respectively ( Table 2 ). Cut-offs for the tertiles are 6.9% and 13.1% for all patients, 7.3% and 14.3% for the anti-PD1 treated patients and 6.1% and 11.7% for the patients treated with anti-PD1 combined with anti-CTLA4. Among the 5% of patients with the lowest TIL levels (cut-off1≤12.3%), the ORR was 40.7%, compared to 74.6% in the highest 5% (cut-off1≥135.4%). View this table: View inline View popup Download powerpoint Table 2. Association of AI-detected TILs with response, progression free survival and overall survival in 1202 patients treated with immune checkpoint inhibition. Results are also shown stratified for 779 patients treated with anti-PD1 monotherapy and 423 patients treated with anti-PD1 + anti-CTLA4. Abbreviations: ICI, immune checkpoint inhibition; CI, confidence interval; OR, odds ratio; HR, hazard ratio; ORR, objective response rate; PFS, progression free survival; OS, overall survival In univariable analysis, a 10% increase in baseline TIL percentage was associated with a greater likelihood of response (OR 1.37 [95% CI 1.21–1.54]). This association remained significant in multivariable analysis after adjusting for established clinical predictors: age, sex, disease stage, WHO performance status, LDH levels, BRAF V600 mutation status, presence of symptomatic brain metastases, and type of therapy (adjusted OR [aOR] 1.40 [95% CI 1.23–1.59]; Table 2 and Supplementary Table 3). When stratified by therapy type, TILs were predictive of response in both anti-PD1 monotherapy (aOR 1.44 [95% CI 1.23–1.67]) and combination therapy with anti-PD1 and anti-CTLA4 (aOR 1.36 [95% CI 1.07–1.73]; Table 2 ). AI-detected TILs and survival TILs were associated with an improved PFS and OS. Patients in the highest tertile had a median PFS of 15.6 months, while those in the middle and lowest tertiles had PFS of 9.7 and 5.3 months, respectively ( P <.001). OS was 49.6 months in the highest tertile, compared to 35.9 and 21.4 months in the middle and lowest tertiles ( P <.001). Results were similar for patients treated with anti-PD1 monotherapy, however in patients treated with anti-PD1 + anti-CTLA4 combination therapy there was no clear separation between the middle and highest tertile in both the OS and PFS analysis ( P =0.02 for OS and P =0.01 for PFS, Figure 2 ). Download figure Open in new tab Figure 2. Kaplan-Meier curves for progression-free survival (PFS) and overall survival (OS) grouped by low, intermediate, and high levels of AI-detected tumor-infiltrating lymphocytes in pre-treatment metastatic specimens of patients with advanced melanoma. Curves are shown for all patients (n = 1202, panels A and B), those treated with anti-PD1 monotherapy (n = 779, panels C and D), and those receiving combination therapy (n = 423, panels E and F). PFS data were missing for 7 patients. Differences between groups were assessed using log-rank tests. When using TILs as a continuous variable, a 10% increase in baseline TIL percentage was associated with a decreased risk of progression (HR 0.85, 95% CI [0.79–0.91]) and death (HR 0.81, 95% CI [0.74– 0.88]). This effect remained significant in multivariable analysis (adjusted HR 0.85 [95% CI 0.79–0.92] for PFS and 0.83 [95% CI 0.76–0.91] for OS; Table 2 and Supplementary Table 3). When stratified based on treatment, similar results were seen for both patients treated with anti-PD1 monotherapy and anti-PD1 combined with anti-CTLA4 ( Table 2 ). Biopsy versus surgical excision TILs remained predictive in both surgical resections and biopsies, with an odds ratio (OR) of 1.32 [95% CI 1.14–1.52] for resections and 1.54 [95% CI 1.24–1.91] for biopsies. Median average AI-detected TIL percentage was 9.4% [IQR 5.3% - 15.2%] in biopsies, compared to 10.4% [IQR 5.6% - 18.6%] in resections. The median annotated tumor area was larger in resections (96.3 mm² [IQR 35.3 – 197.4]) than in biopsies (11.3 mm² [IQR 5.4 – 24.0]). Correlation with manual TIL scoring AI-detected TILs showed a moderate correlation with manually scored intra-tumoral stromal TILs in 529 patients for which both were available (Spearman’s ρ = 0.45, P <.001; Supplementary Figure 4). When comparing the effect estimates of manually scored TILs with AI-detected TILs for prediction of ORR, AI-detected TILs showed a stronger association per 10% increase (OR 1.48 [95% CI 1.21–1.81] vs. 1.14 [95% CI 1.05–1.23]), indicating a higher predictive value ( Table 3 ). View this table: View inline View popup Download powerpoint Table 3. Association of AI-detected TILs and manually scored TILs with response, progression free survival and overall survival in 529 patients treated with immune checkpoint inhibition. Results are also shown stratified for 317 patients treated with anti-PD1 monotherapy and 208 patients treated with anti-PD1 + anti-CTLA4. Abbreviations: CI, confidence interval; OR, odds ratio; HR, hazard ratio; ORR, objective response rate; PFS, progression free survival; OS, overall survival Discussion In this multicenter study with a broad and representative cohort, we demonstrate that in patients with metastatic cutaneous melanoma, the presence of AI-detected TILs in pre-treatment metastatic specimens is associated with an improved likelihood of responding to ICI and a lower risk of disease progression and improved overall survival. Importantly, these correlations were independent of known clinical predictors. The results of our study are in line with the findings of Chatziioannou et al., who reported an association between electronic TILs score and prolonged PFS and melanoma-specific survival in 101 ICI-treated patients with advanced melanoma. In their study, a threshold of 12.2% electronic TILs within manually annotated tumor regions was identified as the optimal prognostic cut-off for PFS. This threshold was significantly associated with prolonged melanoma-specific survival in a Cox regression model adjusted for sex and LDH. 16 Our study included a substantially larger cohort of 1,202 patients, enabling the assessment of TILs as a continuous variable, which demonstrated an association between higher TIL levels and improved clinical outcomes. In addition, we adjusted for clinical variables known to be associated with response, providing an assessment of the independent predictive value of TILs. The results of our study and the study performed by Chatziioannou et al. contradict the study performed by Shen et al. who did not find a correlation between TILs and PFS in 144 advanced melanoma patients. 13 , 14 However, it is unclear what proportion of samples analyzed in their study were primary tumors versus metastases, as both were included. In our previous work, we found no association between TILs in primary melanomas and response to ICIs, which may partly explain the absence of a correlation. When considering the work of Chatziioannou et al. and our previous work in manual TILs, both of which focused on metastatic samples, there is increasing evidence to support a relationship between TILs in pre-treatment metastasis and treatment outcomes. This is further supported by multiple studies showing associations between intratumoral CD8 + and CD4 + lymphocyte presence and response to ICIs. 9 , 28 , 29 A major advantage of AI-based TIL quantification is its consistency and reproducibility, whereas interobserver agreement for manual scoring among pathologists is often only moderate. 7 , 8 Furthermore, TILs can be quantified on routinely available H&E-stained slides, eliminating the need for additional immune histochemical staining and making it a practical and accessible biomarker in clinical practice. In our study, AI-detected TILs demonstrated stronger associations with clinical outcomes compared to manually scored TILs, further highlighting their potential. To support transparency and reproducibility, our AI model, as well as a part of the developmental dataset, have been made publicly available, enabling further external validation and future clinical implementation. 17 , 30 Our study has several strengths, including its size, multicenter design, and the use of histopathology images from 30 institutions with varying staining protocols, which enhances the generalizability and applicability of our findings. Furthermore, we evaluated TILs as a continuous variable rather than relying on an optimal threshold, allowing for a more nuanced analysis of their predictive value. This study is also subject to several limitations. Some patients were excluded from the analysis due to the absence of available histopathology samples, however no significant differences were observed between included and excluded patients. AI-based segmentation may also occasionally misclassify TILs; however, no major misclassifications of tumor cells or TILs were identified upon visual inspection. In addition, we assessed only a single slide of a metastatic lesion per patient, despite known intra- and interlesional heterogeneity, which may limit the generalizability of findings across all tumor sites. 31 However, in the single selected sample before therapy initiation there was already predictive value supporting the predictive capability of TILs in advanced melanoma. Lastly, the observed ORR in our cohort was relatively high. This may, in part, be explained by the use of investigator-assessed treatment response in the Dutch Melanoma Treatment Registry. The response rate is similar to earlier studies published from this registry data. 2 Whereas this study shows that TILs can be used to identify patients with a favorable prognosis, reaching a response rate of 74.6% in patients with high TIL scores (above 34.5%), a major limitation remains that it is not sufficiently possible to identify non-responders as demonstrated by the ORR of 40.7% in patients with less than 3.54% TILs. Future research should therefore focus on improving the negative predictive value of AI-assisted histopathology analysis. Previous studies suggest that not only the quantity but also the spatial context of TILs, such as their presence within the tumor center or at the invasive margin and distance between TILs and tumor cells, may be prognostically relevant 28 , 32 . Investigating whether foundation models, capable of directly analyzing whole-slide histopathological images, can outperform current AI-based methods is another promising direction 33 , 34 . These models may better capture tumor–immune interactions which are associated with clinical outcomes. However, a downside is their limited interpretability when compared to TIL analysis, as they are based on abstract feature representations of the histopathology image data. Regarding the current TIL detection model, future efforts should focus on integrating AI-detected TILs into prediction models that can be used in clinical practice, not only to guide the decision to start treatment, but also to de-escalate treatment. TILs appear to be particularly valuable in identifying patients with a high likelihood of response. In this light, the algorithm might be valuable in the neoadjuvant setting, identifying for whom anti-PD1 monotherapy suffices. Conclusion In this large multicenter study, AI-detected TILs in pre-treatment metastatic melanoma specimens are predictive of response to ICI and associated with improved PFS and OS. TILs provide a standardized, reproducible, and potentially more predictive alternative to manual scoring and can be derived from routine H&E-stained slides, making them a practical and accessible biomarker. These findings support the potential of AI-driven histopathological analysis as a predictive tool in melanoma. Further research is needed to improve the predictive value of AI-detected TILs and its use in treatment guidance and de-escalation. Funding This research was funded by The Netherlands Organization for Health Research and Development (ZonMW, project number 848101007), by an unrestricted grant of Stichting Hanarth Fonds, The Netherlands, and by Philips. Data Availability Part of the developmental dataset used for model training is publicly available on Zenodo 1. The model architecture is available on Github2 and model specific weights are automatically downloaded upon usage. The whole slide images and associated patient outcome data used in this study cannot be shared due to confidentiality agreements and data sharing restrictions related to patient privacy. 1. Schuiveling M. (2025) Melanoma Histopathology Dataset with Tissue and Nuclei Annotations. Published online March 19, 2025. https://zenodo.org/records/15050523 2. Schuiveling M. (2025) HoVer-NeXt inference for TIL quantification in melanoma: Fork of HoVer-NeXt inference by Baumann et al. with additional pipelines and dataset support for melanoma research [Computer software]. Published online May 27, 2025. Available at: https://github.com/mschuiveling/hover-next-inference-tils-melanoma Forked from: Baumann E, Dislich B, Rumberger JL, Nagtegaal ID, Martinez MR, Zlobec I. HoVer-NeXt inference. [Original software]. Available at: https://github.com/digitalpathologybern/hover_next_inference https://zenodo.org/records/15050523 https://github.com/mschuiveling/hover-next-inference-tils-melanoma Data Sharing Statement Part of the developmental dataset used for model training is publicly available on Zenodo 17 . The model architecture is available on Github 18 and model specific weights are automatically downloaded upon usage. The whole slide images and associated patient outcome data used in this study cannot be shared due to confidentiality agreements and data sharing restrictions related to patient privacy. Acknowledgements The authors would like to thank PALGA, the nationwide network and registry of histo- and cytopathology in the Netherlands, for providing histopathological data and for their help in the collection of FFPE tissue. Abbreviations AI Artificial Intelligence AJCC American Joint Committee on Cancer CI Confidence Interval F 1 score Harmonic mean of precision and recall in model evaluation H&E Hematoxylin and Eosin ICI Immune Checkpoint Inhibition IQR Interquartile Range LDH Lactate Dehydrogenase MICE Multiple Imputation by Chained Equations OR Odds Ratio ORR Objective Response Rate OS Overall Survival PFS Progression-Free Survival TILs Tumor-Infiltrating Lymphocytes WHO World Health Organization References 1. ↵ Wolchok JD , Chiarion-Sileni V , Rutkowski P , et al. Final, 10-Year Outcomes with Nivolumab plus Ipilimumab in Advanced Melanoma . New England Journal of Medicine . 2025 ; 392 ( 1 ): 11 – 22 . doi: 10.1056/NEJMoa2407417 OpenUrl CrossRef 2. ↵ van Not OJ , Blokx WAM , van den Eertwegh AJM , et al. BRAF and NRAS Mutation Status and Response to Checkpoint Inhibition in Advanced Melanoma . JCO Precis Oncol . 2022 ; 6 : e2200018 . doi: 10.1200/PO.22.00018 OpenUrl CrossRef 3. ↵ Suijkerbuijk KPM , van Eijs MJM , van Wijk F , Eggermont AMM . Clinical and translational attributes of immune-related adverse events . Nature Cancer . 2024 ; 5 ( 4 ): 557 – 571 . doi: 10.1038/s43018-024-00730-3 OpenUrl CrossRef 4. ↵ Franken MG , Leeneman B , Aarts MJB , et al. Trends in survival and costs in metastatic melanoma in the era of novel targeted and immunotherapeutic drugs . ESMO Open . 2021 ; 6 ( 6 ): 100320 . doi: 10.1016/j.esmoop.2021.100320 OpenUrl CrossRef 5. ↵ Blommestein HM , de Groot S , Leeneman B , et al. Cost-effectiveness of treatment sequences for BRAF-mutant advanced melanoma in the Netherlands using a health economic model . European Journal of Cancer . 2025 ; 218 : 115071 . doi: 10.1016/j.ejca.2024.115071 OpenUrl CrossRef 6. ↵ Hendry S , Salgado R , Gevaert T , et al. Assessing Tumor-Infiltrating Lymphocytes in Solid Tumors: A Practical Review for Pathologists and Proposal for a Standardized Method from the International Immuno-Oncology Biomarkers Working Group: Part 2: TILs in Melanoma, Gastrointestinal Tract Carcinomas, Non-Small Cell Lung Carcinoma and Mesothelioma, Endometrial and Ovarian Carcinomas, Squamous Cell Carcinoma of the Head and Neck, Genitourinary Carcinomas, and Primary Brain Tumors . Adv Anat Pathol . 2017 ; 24 ( 6 ): 311 – 335 . doi: 10.1097/PAP.0000000000000161 OpenUrl CrossRef PubMed 7. ↵ Duin IAJ van , Schuiveling M , Maat LS ter , et al. Baseline tumor-infiltrating lymphocyte patterns and response to immune checkpoint inhibition in metastatic cutaneous melanoma . European Journal of Cancer . 2024 ; 208 . doi: 10.1016/j.ejca.2024.114190 OpenUrl CrossRef 8. ↵ Swisher SK , Wu Y , Castaneda CA , et al. Interobserver Agreement Between Pathologists Assessing Tumor-Infiltrating Lymphocytes (TILs) in Breast Cancer Using Methodology Proposed by the International TILs Working Group . Ann Surg Oncol . 2016 ; 23 ( 7 ): 2242 – 2248 . doi: 10.1245/s10434-016-5173-8 OpenUrl CrossRef 9. ↵ Lopez de Rodas M , Villalba-Esparza M , Sanmamed MF , Chen L , Rimm DL , Schalper KA . Biological and clinical significance of tumour-infiltrating lymphocytes in the era of immunotherapy: a multidimensional approach . Nat Rev Clin Oncol . 2025 ; 22 ( 3 ): 163 – 181 . doi: 10.1038/s41571-024-00984-x OpenUrl CrossRef PubMed 10. ↵ Artificial Intelligence–Powered Spatial Analysis of Tumor-Infiltrating Lymphocytes as Complementary Biomarker for Immune Checkpoint Inhibition in Non–Small-Cell Lung Cancer | Journal of Clinical Oncology . Accessed January 27, 2025 . https://ascopubs.org/doi/full/10.1200/JCO.21.02010 11. Amgad M , Salgado R , Cooper LAD . MuTILs: A multiresolution deep-learning model for interpretable scoring of Tumor-Infiltrating Lymphocytes in breast carcinomas using clinical guidelines . Published online May 15 , 2023 :2022.01.08.22268814. doi: 10.1101/2022.01.08.22268814 OpenUrl Abstract / FREE Full Text 12. ↵ Loi S , Salgado R , Schmid P , et al. Association Between Biomarkers and Clinical Outcomes of Pembrolizumab Monotherapy in Patients With Metastatic Triple-Negative Breast Cancer: KEYNOTE-086 Exploratory Analysis . JCO Precis Oncol . 2023 ;( 7 ): e2200317 . doi: 10.1200/PO.22.00317 OpenUrl CrossRef PubMed 13. ↵ Chatziioannou E , Roßner J , Aung TN , et al. Deep learning-based scoring of tumour-infiltrating lymphocytes is prognostic in primary melanoma and predictive to PD-1 checkpoint inhibition in melanoma metastases . eBioMedicine . 2023 ; 93 . doi: 10.1016/j.ebiom.2023.104644 OpenUrl CrossRef 14. ↵ Shen J , Choi YL , Lee T , et al. Inflamed immune phenotype predicts favorable clinical outcomes of immune checkpoint inhibitor therapy across multiple cancer types . J Immunother Cancer . 2024 ; 12 ( 2 ): e008339 . doi: 10.1136/jitc-2023-008339 OpenUrl Abstract / FREE Full Text 15. Acs B , Ahmed FS , Gupta S , et al. An open source automated tumor infiltrating lymphocyte algorithm for prognosis in melanoma . Nat Commun . 2019 ; 10 : 5440 . doi: 10.1038/s41467-019-13043-2 OpenUrl CrossRef 16. ↵ Schuiveling M , Liu H , Eek D , et al. A novel dataset for nuclei and tissue segmentation in melanoma with baseline nuclei segmentation and tissue segmentation benchmarks . GigaScience . 2025 ; 14 : giaf011 . doi: 10.1093/gigascience/giaf011 OpenUrl CrossRef PubMed 17. ↵ Schuiveling M . ( 2025 ) Melanoma Histopathology Dataset with Tissue and Nuclei Annotations . Published online March 19, 2025. https://zenodo.org/records/15050523 18. ↵ Schuiveling M. ( 2025 ) HoVer-NeXt inference for TIL quantification in melanoma: Fork of HoVer-NeXt inference by Baumann et al. with additional pipelines and dataset support for melanoma research [Computer software] . Published online May 27, 2025. Available at: https://github.com/mschuiveling/hover-next-inference-tils-melanoma Forked from: Baumann E, Dislich B, Rumberger JL, Nagtegaal ID, Martínez MR, Zlobec I. HoVer-NeXt inference. [Original software]. Available at: https://github.com/digitalpathologybern/hover_next_inference 19. ↵ Baumann E , Dislich B , Rumberger JL , Nagtegaal ID , Martınez MR , Zlobec I . HoVer-NeXt: A Fast Nuclei Segmentation and Classification Pipeline for Next Generation Histopathology . Medical Imaging with Deep Learning. Published online July 3 , 2024 . 20. ↵ Rumberger JL , Baumann E , Hirsch P , Janowczyk A , Zlobec I , Kainmueller D . Panoptic segmentation with highly imbalanced semantic labels . Published online April 19 , 2022 . doi: 10.48550/arXiv.2203.11692 OpenUrl CrossRef 21. ↵ Jochems A , Schouwenburg MG , Leeneman B , et al. Dutch Melanoma Treatment Registry: Quality assurance in the care of patients with metastatic melanoma in the Netherlands . Eur J Cancer . 2017 ; 72 : 156 – 165 . doi: 10.1016/j.ejca.2016.11.021 OpenUrl CrossRef PubMed 22. ↵ Gershenwald JE , Scolyer RA , Hess KR , et al. Melanoma staging: Evidence-based changes in the American Joint Committee on Cancer eighth edition cancer staging manual . CA: A Cancer Journal for Clinicians . 2017 ; 67 ( 6 ): 472 – 492 . doi: 10.3322/caac.21409 OpenUrl CrossRef PubMed 23. ↵ Bankhead P , Loughrey MB , Fernández JA , et al. ( 2025 ) QuPath: Open source software for digital pathology image analysis (Version 0.5.0) [Computer software) . Software https://github.com/qupath/qupath/releases/tag/v0.5.0 . Published online December 4, 2017. Accessed January 24, 2025 . https://www.nature.com/articles/s41598-017-17204-5 24. Thagaard J , Broeckx G , Page DB , et al. Pitfalls in machine learning-based assessment of tumor-infiltrating lymphocytes in breast cancer: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer . J Pathol . 2023 ; 260 ( 5 ): 498 – 513 . doi: 10.1002/path.6155 OpenUrl CrossRef 25. ↵ Bankhead P , Loughrey MB , Fernández JA , et al. QuPath: Open source software for digital pathology image analysis . Sci Rep . 2017 ; 7 ( 1 ): 16878 . doi: 10.1038/s41598-017-17204-5 OpenUrl CrossRef PubMed 26. ↵ Casparie M , Tiebosch ATMG , Burger G , et al. Pathology databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide histopathology and cytopathology data network and archive . Cell Oncol . 2007 ; 29 ( 1 ): 19 – 24 . doi: 10.1155/2007/971816 OpenUrl CrossRef PubMed Web of Science 27. ↵ Eisenhauer EA , Therasse P , Bogaerts J , et al. New response evaluation criteria in solid tumours: Revised RECIST guideline (version 1.1) . European Journal of Cancer . 2009 ; 45 ( 2 ): 228 – 247 . doi: 10.1016/j.ejca.2008.10.026 OpenUrl CrossRef PubMed Web of Science 28. ↵ Gide TN , Silva , Ines P. , Quek , Camelia , et al. Close proximity of immune and tumor cells underlies response to anti-PD-1 based therapies in metastatic melanoma patients . OncoImmunology . 2020 ; 9 ( 1 ): 1659093 . doi: 10.1080/2162402X.2019.1659093 OpenUrl CrossRef 29. ↵ Tumeh PC , Harview CL , Yearley JH , et al. PD-1 blockade induces responses by inhibiting adaptive immune resistance . Nature . 2014 ; 515 ( 7528 ): 568 – 571 . doi: 10.1038/nature13954 OpenUrl CrossRef PubMed Web of Science 30. ↵ Schuiveling M . ( 2025 ) Weights for PUMA challenge . Zenodo . doi: 10.5281/zenodo.14651631 . OpenUrl CrossRef 31. ↵ Madore J , Vilain RE , Menzies AM , et al. PD-L1 expression in melanoma shows marked heterogeneity within and between patients: implications for anti-PD-1/PD-L1 clinical trials . Pigment Cell & Melanoma Research . 2015 ; 28 ( 3 ): 245 – 253 . doi: 10.1111/pcmr.12340 OpenUrl CrossRef PubMed 32. ↵ AbdulJabbar K , Raza SEA , Rosenthal R , et al. Geospatial immune variability illuminates differential evolution of lung adenocarcinoma . Nat Med . 2020 ; 26 ( 7 ): 1054 – 1062 . doi: 10.1038/s41591-020-0900-x OpenUrl CrossRef 33. ↵ Zimmermann E , Vorontsov E , Viret J , et al. Virchow2: Scaling Self-Supervised Mixed Magnification Models in Pathology . Published online November 6, 2024. doi: 10.48550/arXiv.2408.00738 OpenUrl CrossRef 34. ↵ Chen RJ , Ding T , Lu MY , et al. Towards a general-purpose foundation model for computational pathology . Nat Med . 2024 ; 30 ( 3 ): 850 – 862 . doi: 10.1038/s41591-024-02857-3 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted May 29, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma Mark Schuiveling , Isabella A.J. van Duin , Laurens ter Maat , Janneke C. van der Weerd , Rik J. Verheijden , Franchette van den Berkmortel , Christian U. Blank , Gerben E. Breimer , Femke H. Burgers , Marye Boers-Sonderen , Alfons J. M. van den Eertwegh , Jan Willem B. de Groot , John B.A.G. Haanen , Geke A.P. Hospers , Ellen Kapiteijn , Djura Piersma , Gerard Vreugdenhil , Hans Westgeest , Anne M.R. Schrader , Josien Pluim , Paul J. van Diest , Mitko Veta , Karijn P.M. Suijkerbuijk , Willeke Blokx medRxiv 2025.05.28.25328410; doi: https://doi.org/10.1101/2025.05.28.25328410 Share This Article: Copy Citation Tools AI-detected tumor-infiltrating lymphocytes for predicting outcomes in anti-PD1 based treated melanoma Mark Schuiveling , Isabella A.J. van Duin , Laurens ter Maat , Janneke C. van der Weerd , Rik J. Verheijden , Franchette van den Berkmortel , Christian U. Blank , Gerben E. Breimer , Femke H. Burgers , Marye Boers-Sonderen , Alfons J. M. van den Eertwegh , Jan Willem B. de Groot , John B.A.G. Haanen , Geke A.P. Hospers , Ellen Kapiteijn , Djura Piersma , Gerard Vreugdenhil , Hans Westgeest , Anne M.R. Schrader , Josien Pluim , Paul J. van Diest , Mitko Veta , Karijn P.M. Suijkerbuijk , Willeke Blokx medRxiv 2025.05.28.25328410; doi: https://doi.org/10.1101/2025.05.28.25328410 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Oncology Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15227) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6597) Geriatric Medicine (668) Health Economics (997) Health Informatics (4534) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9230) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a0032e6dce51593a',t:'MTc3OTUzMDA4OA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();