Tolkach-ESCA¶
9,000 tiles of oesophageal tissue across six classes — tumour, regression, adventitia,
muscularis propria, oesophageal and gastric mucosa — from three centers (UKK, WNS and CHA),
scored at k = 61. The TCGA cohort of the original Tolkach
dataset is held out, following PathoROB, so like Camelyon this is scored
outside TCGA.
Model |
bio bacc |
conf bacc |
|
|
|
F(0) |
LTM₁₀ |
support |
|---|---|---|---|---|---|---|---|---|
Midnight-12k |
0.976 |
0.728 |
0.943 |
0.941 |
0.58 |
0.051 |
-0.08 |
99.0% |
Mascaret |
0.977 |
0.445 |
0.972 |
0.973 |
0.51 |
0.030 |
0.01 |
100.0% |
RudolfV-2-S |
0.984 |
0.599 |
0.967 |
0.975 |
0.49 |
0.032 |
-0.00 |
99.8% |
CONCH |
0.973 |
0.654 |
0.951 |
0.957 |
0.44 |
0.045 |
-0.04 |
99.8% |
RudolfV-2 |
0.986 |
0.637 |
0.966 |
0.969 |
0.41 |
0.032 |
-0.01 |
99.4% |
RudolfV-2-B |
0.984 |
0.664 |
0.960 |
0.968 |
0.41 |
0.036 |
-0.02 |
99.0% |
CONCHv1.5 |
0.973 |
0.633 |
0.952 |
0.964 |
0.39 |
0.043 |
-0.03 |
99.9% |
GenBio-PathFM |
0.981 |
0.598 |
0.960 |
0.964 |
0.39 |
0.038 |
-0.02 |
99.9% |
H0-mini |
0.967 |
0.642 |
0.935 |
0.946 |
0.38 |
0.058 |
-0.07 |
99.7% |
Virchow |
0.970 |
0.703 |
0.935 |
0.943 |
0.37 |
0.053 |
-0.05 |
99.6% |
Virchow2 |
0.978 |
0.613 |
0.954 |
0.957 |
0.35 |
0.040 |
-0.04 |
99.6% |
MUSK |
0.969 |
0.739 |
0.924 |
0.931 |
0.29 |
0.063 |
-0.07 |
99.2% |
H-optimus-1 |
0.977 |
0.680 |
0.940 |
0.949 |
0.26 |
0.049 |
-0.04 |
99.1% |
Phaet |
0.967 |
0.619 |
0.935 |
0.946 |
0.25 |
0.050 |
-0.03 |
100.0% |
GPFM |
0.969 |
0.841 |
0.883 |
0.902 |
0.24 |
0.095 |
-0.10 |
97.6% |
H-optimus-0 |
0.971 |
0.731 |
0.911 |
0.919 |
0.23 |
0.076 |
-0.08 |
98.5% |
UNI2-h |
0.976 |
0.767 |
0.916 |
0.927 |
0.22 |
0.064 |
-0.06 |
97.7% |
mSTAR |
0.970 |
0.818 |
0.881 |
0.898 |
0.19 |
0.087 |
-0.09 |
97.7% |
DINOv2-B † |
0.905 |
0.535 |
0.876 |
0.867 |
0.18 |
0.099 |
-0.07 |
100.0% |
Phikon |
0.962 |
0.898 |
0.772 |
0.782 |
0.17 |
0.179 |
-0.19 |
81.5% |
UNI |
0.973 |
0.835 |
0.880 |
0.891 |
0.17 |
0.086 |
-0.08 |
95.3% |
Hibou-B |
0.967 |
0.916 |
0.774 |
0.775 |
0.13 |
0.188 |
-0.17 |
85.7% |
Prov-GigaPath |
0.962 |
0.929 |
0.707 |
0.746 |
0.13 |
0.236 |
-0.16 |
79.1% |
Prost40M |
0.911 |
0.838 |
0.701 |
0.721 |
0.13 |
0.277 |
-0.24 |
96.2% |
Phikon-v2 |
0.956 |
0.896 |
0.741 |
0.746 |
0.12 |
0.225 |
-0.17 |
83.3% |
Hibou-L |
0.955 |
0.960 |
0.624 |
0.586 |
0.11 |
0.315 |
-0.29 |
67.0% |
One provenance caveat: the RudolfV-2 family’s disclosed Charité/LMU institutional
corpus creates a possible institutional/source-domain overlap with this cohort’s CHA
center. Exact patient or slide overlap is unknown, so this does not establish leakage —
but read the family’s scores here with it in mind.
The mildest of the three cohorts — 0 encoders fall
below zero — and the one where the count-based indices run out of room:
16 of the 25 ranked
encoders score above 0.90 on RI, so RI and MaRI have largely stopped
separating models here while CRoMa still spreads the panel.
The two rankings, on this cohort alone:
Median CRoMa against tail severity LTM₁₀ on Tolkach-ESCA. Better is up and to the
right; ringed points are undominated on both axes and named, and the shaded region is
dominated on both. Hover or tab to any point to name it with its two values. The
natural-image control is excluded — the frontier is a pathology-only claim.
The distribution explorer¶
The same explorer as the aggregate page’s, pinned to Tolkach-ESCA. Click a row to move the detail, drag across the detail curve to count the samples in any range, and pick a second encoder under Compare with to overlay its shape.
Several encoders are visibly bimodal — one population of neighbourhoods comfortably biology-dominant, another close to the line. A median reports where the middle of that lands and says nothing about the split, which is the case tail reporting exists for.
Shortcut susceptibility¶
Shortcut susceptibility for every encoder on this cohort, in domain (ID) and out of
domain (OOD); the natural-image control sits last. Rows are ranked by Change at V = 1,
the normalized change at maximum confounding, because nIPD averages over the whole
range: an early gain there can pay for a late collapse, so a curve ending at chance can
outrank one that never moved. Rows ending at or below -0.900 are marked ≈ chance,
where none of the above-chance margin survives.
Bold marks the leading ranked encoder in each column where higher is better, so a
column that disagrees with the ranking shows it at a glance. Each caption reports
Spearman ρ, the rank correlation between the CRoMa and nIPD columns: how closely
the two order the encoders the same way.
Shortcut susceptibility defines the measure and holds the interactive explorer.
Model |
Median CRoMa (m=5) |
Change at |
|
Baseline balanced accuracy |
|---|---|---|---|---|
Mascaret |
0.507 |
0.011 |
0.006 |
0.974 |
RudolfV-2-S |
0.485 |
0.011 |
0.006 |
0.976 |
GenBio-PathFM |
0.390 |
0.005 |
0.003 |
0.972 |
RudolfV-2 |
0.412 |
0.004 |
0.000 |
0.983 |
RudolfV-2-B |
0.407 |
0.003 |
0.000 |
0.979 |
Midnight-12k |
0.584 |
-0.002 |
0.002 |
0.974 |
CONCH |
0.439 |
-0.002 |
-0.000 |
0.973 |
Virchow2 |
0.351 |
-0.002 |
-0.002 |
0.984 |
H0-mini |
0.380 |
-0.004 |
-0.002 |
0.970 |
MUSK |
0.293 |
-0.006 |
-0.001 |
0.968 |
CONCHv1.5 |
0.392 |
-0.006 |
-0.001 |
0.973 |
Virchow |
0.371 |
-0.010 |
-0.005 |
0.977 |
H-optimus-1 |
0.260 |
-0.010 |
-0.003 |
0.973 |
Phaet |
0.249 |
-0.016 |
-0.003 |
0.966 |
UNI2-h |
0.221 |
-0.020 |
-0.008 |
0.972 |
mSTAR |
0.187 |
-0.025 |
-0.005 |
0.977 |
H-optimus-0 |
0.232 |
-0.031 |
-0.007 |
0.968 |
UNI |
0.175 |
-0.053 |
-0.011 |
0.980 |
GPFM |
0.241 |
-0.054 |
-0.014 |
0.972 |
Phikon |
0.175 |
-0.116 |
-0.024 |
0.969 |
Hibou-B |
0.134 |
-0.144 |
-0.024 |
0.973 |
Prost40M |
0.127 |
-0.184 |
-0.049 |
0.929 |
Phikon-v2 |
0.122 |
-0.194 |
-0.041 |
0.973 |
Prov-GigaPath |
0.132 |
-0.209 |
-0.042 |
0.969 |
Hibou-L |
0.111 |
-0.215 |
-0.039 |
0.973 |
DINOv2-B † |
0.175 |
-0.034 |
-0.008 |
0.942 |
Model |
Median CRoMa (m=5) |
Change at |
|
Baseline balanced accuracy |
|---|---|---|---|---|
RudolfV-2-S |
0.485 |
0.004 |
0.002 |
0.971 |
Virchow2 |
0.351 |
0.002 |
0.001 |
0.975 |
Mascaret |
0.507 |
0.002 |
0.001 |
0.969 |
Midnight-12k |
0.584 |
0.002 |
0.004 |
0.966 |
RudolfV-2 |
0.412 |
0.002 |
-0.000 |
0.974 |
CONCH |
0.439 |
0.001 |
-0.000 |
0.964 |
RudolfV-2-B |
0.407 |
0.000 |
-0.000 |
0.974 |
CONCHv1.5 |
0.392 |
-0.000 |
0.001 |
0.962 |
GenBio-PathFM |
0.390 |
-0.001 |
0.000 |
0.970 |
H-optimus-1 |
0.260 |
-0.003 |
-0.000 |
0.971 |
H0-mini |
0.380 |
-0.003 |
-0.003 |
0.966 |
Virchow |
0.371 |
-0.004 |
-0.002 |
0.967 |
UNI2-h |
0.221 |
-0.004 |
-0.002 |
0.972 |
MUSK |
0.293 |
-0.004 |
-0.000 |
0.959 |
Phaet |
0.249 |
-0.006 |
-0.000 |
0.958 |
H-optimus-0 |
0.232 |
-0.008 |
-0.002 |
0.962 |
UNI |
0.175 |
-0.011 |
-0.001 |
0.968 |
mSTAR |
0.187 |
-0.012 |
-0.003 |
0.968 |
Hibou-B |
0.134 |
-0.019 |
-0.000 |
0.965 |
GPFM |
0.241 |
-0.020 |
-0.003 |
0.960 |
Prov-GigaPath |
0.132 |
-0.039 |
-0.007 |
0.968 |
Phikon |
0.175 |
-0.041 |
-0.006 |
0.958 |
Phikon-v2 |
0.122 |
-0.068 |
-0.013 |
0.961 |
Hibou-L |
0.111 |
-0.081 |
-0.012 |
0.964 |
Prost40M |
0.127 |
-0.096 |
-0.028 |
0.922 |
DINOv2-B † |
0.175 |
-0.028 |
-0.006 |
0.923 |