{"id":"d53d90a1-a542-43c4-9ca4-0d136b0f449e","arxiv_id":"2412.08358","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"An isotropic depolarization filter, implemented by subtracting the perfect-depolarizer term of the characteristic decomposition, amplifies anisotropic polarimetric contrast and reveals structures in heart and brain tissue images.","lead":"This paper presents a polarization-based image filter that removes the isotropic, structureless part of depolarization and amplifies the anisotropic part, making hidden tissue structures more visible. Applied to slices of lamb heart and cattle brain, the filter sharpens boundaries between tissue types and reveals white-matter fiber tracts that ordinary polarimetric images do not show.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Division by noisy P3 estimates as low as 0.03 makes the claimed structure-unveiling potentially an amplification artifact; no noise analysis or flat-field control is provided.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing concern: pixel-wise division by small P3 amplifies noise and may produce artifacts. My analysis agrees and adds two specifics: (i) P3 is a derived quantity from the same noisy measurement, so the denominator noise is multiplicative and can dominate when P3 is as low as 0.03; (ii) the filtered Mueller matrix Ma can violate physical realizability, which weakens the physical interpretation of the filtered observables. These points do not overturn the paper's mathematical derivation, but they do make the central 'structure unveiling' claim empirically unsubstantiated. The suggested flat-field control is a direct and feasible experiment that would distinguish genuine signal enhancement from noise amplification. Since the paper's data are not available and no noise analysis is included, the conditional verdict is appropriate; no change is required.","tokens_in":21123,"tokens_out":6435,"duration_ms":67683,"concrete_test":"Acquire a Mueller matrix image of a homogeneous low-P3 depolarizing sample (e.g., a thick scattering suspension with P3 ≈ 0.03), apply the IDF, and compute the spatial standard deviation of D' and P1' across the image. If the resulting flat-field contrast is comparable to the contrast between the claimed structures in Fig. 3(b) or Fig. 2(b), the filter generates pseudo-structures from noise. This control would settle whether the 1/P3 amplification preserves genuine anisotropic signal or merely magnifies measurement noise.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The filter's central operation Ma = M - (1-P3) m00 M3 reduces, for every normalized observable, to dividing the original observable by P3(x,y) (Eqs. (8), (13), (14)). For the heart and brain samples, Table S1 lists P3 values of 0.032 and 0.125, so the scale factor is 8-31. The key problem is that P3 is itself estimated from the same noisy Mueller matrix, via the eigenvalues of H (Eq. (12)). In low-P3 regions the eigenvalue spectrum of H is nearly degenerate, so small measurement noise in M can produce large relative errors in P3. The filtered observable D' = D/P3 therefore inherits a multiplicative noise contribution from the denominator. The paper contains no repeated-measurement noise characterization, no flat-field control, and no comparison against histology or other independent ground truth, and the data are not publicly available. Consequently, the fiber tracts and myocardial boundaries in Figs. 2 and 3 could be spatial fluctuations of the measurement noise amplified by 1/P3 rather than genuine anisotropic depolarization. A secondary concern is that H(Ma) = H(M) - (1-P3)m00 I has eigenvalues shifted by a constant that can become negative for small P3, so Ma may not be physically realizable; while the paper uses the filter only for visualization, the interpretation of the filtered observables as physical properties of the sample is thereby weakened.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a post-processing \"isotropic depolarization filter\" (IDF) that subtracts the perfect-depolarizer term (1-P3)m00 M3 from each measured Mueller matrix image M, yielding a filtered matrix Ma. The authors derive that, after renormalization, all non-intensity polarimetric observables (diattenuation, polarizance, indices of polarimetric purity, depolarization index, spherical purity) are rescaled by the factor 1/P3(x,y). They apply the filter to ex-vivo lamb heart and cattle brain sections and report that structures such as myocardial/subendocardial boundaries and white-matter fiber tracts become visible in the filtered images where they were invisible or unclear in the unfiltered images and in intensity images.","tokens_in":21416,"tokens_out":6197,"duration_ms":63186,"significance":"The filter is simple, parameter-free, and readily implementable, and the algebraic derivation from the characteristic decomposition is clear. The premise that isotropic depolarization can mask anisotropic depolarization-related contrast is physically plausible and consistent with the authors' earlier work on the IPP framework. However, the empirical demonstration is not yet convincing: the reported contrast and dispersion increases are exactly the deterministic 1/P3 rescaling derived in Eqs. (8) and (13), not independent evidence of performance; the filtered matrix Ma is not shown to be physically realizable; and there is no noise analysis, no independent ground truth (e.g., histology), and no public data. The method may be a useful visualization tool, but the central claim that it \"reveals hidden bioimaging information\" requires substantially stronger validation.","major_comments":[{"comment":"The reported increases in dispersion by factors of 3.92 (heart, Sec. 3.2) and 7.57 (brain, Sec. 3.3) are not empirical evidence of improved performance. These numbers are exactly 1/P3 for the respective samples, as the paper itself notes (\"the contrast is enhanced by an amount of 1/P3\"), so the enlargement of the point clouds in Figs. 2(d) and 3(d) follows by construction from the definition of the filter. A genuine demonstration of advantage requires metrics that are invariant to this trivial rescaling, such as contrast-to-noise ratio against a measured noise floor, receiver-operating-characteristic analysis for tissue discrimination, or comparison with independent structural data; none is provided.","section":"Secs. 3.2 and 3.3, with Eqs. (8) and (13)"},{"comment":"The filtered covariance matrix satisfies H(Ma) = H(M) - (1-P3)m00 I. Using the IPP relation (Eq. (12)), one finds that the smallest eigenvalue of H(Ma) is -3λ_min, which is negative for any sample with P3<1. Consequently Ma violates Cloude's criterion and is not a physically realizable Mueller matrix. The manuscript nowhere acknowledges this, instead interpreting P'_n and D' as physical observables of the sample. The authors should explicitly state that the filtered observables are rescaled versions of the original data rather than physical depolarization properties, or modify the filter to ensure physical realizability; otherwise the claim that the filter \"removes\" isotropic depolarization as a physical operation is not supported.","section":"Sec. 2.2.2, Eq. (9), and Supplement Eq. (S.2)"},{"comment":"The filter divides every non-intensity observable by P3(x,y), which is estimated from the same noisy Mueller matrix through the eigenvalues of H (Eq. (12)). For the heart sample P3 = 0.032, giving a gain of ~31, and for brain white matter P3 = 0.125, giving a gain of ~8. In low-P3 regions the eigenvalue spectrum of H is nearly degenerate, so relative errors in P3 can be large; the division then amplifies noise multiplicatively. The manuscript provides no repeated-measurement noise characterization, no flat-field control, and no histology or other independent ground truth, and the data are not publicly available. Without such controls, the \"revealed\" fiber tracts (Fig. 3(b)) and myocardial boundaries (Fig. 2(b)) could be spatial fluctuations of measurement noise rather than genuine anisotropic depolarization structure.","section":"Secs. 3.2-3.3 and Table S1"},{"comment":"The choice of observable shown for each sample is made post hoc: P1 is chosen for the heart because it was \"the most interesting metric\" and D for the brain because it \"provided the best results\". Since the authors presumably computed many observables (P1, P2, P3, P∆, Ps, D, P), selecting the best-looking one per sample introduces selection bias and inflates the apparent performance. Please present the results for all tested observables for both main samples, or pre-specify the observable selection rule; otherwise the comparison is not a fair test of the filter's general utility.","section":"Secs. 3.2 and 3.3"}],"minor_comments":[{"comment":"The text refers to \"the intensity image of the brain section in Fig. 3 (a)\" and \"the application of the filter on D (Fig. 3 (c))\", but in the caption (a) is the unfiltered diattenuation D and (c) is the intensity image; the cross-references should be corrected.","section":"Sec. 3.3 and Fig. 3 caption"},{"comment":"The inequality in Eq. (S.2) is written as \"λ4 ≤ λ3 ≤ λ2 ≤ λ1 ≤ 0\"; for physical covariance matrices the eigenvalues should be nonnegative, so the final inequality should be \"λ4 ≥ 0\" (or the ordering should be stated consistently with nonnegative eigenvalues).","section":"Supplement, Eq. (S.2)"},{"comment":"The claim that the filter effect on retardance \"can be considered negligible\" is presented as a hypothesis based on unpublished observations. Please provide supporting data or explicitly label this as an untested assumption.","section":"Sec. 2.2.1"},{"comment":"The notation \"D′T P′\" for the off-diagonal block of the filtered Mueller matrix is not defined; please define block notation for clarity.","section":"Eq. (7)"},{"comment":"There are minor language issues: \"These proves\" (Abstract), \"hidden\" used as a verb in Conclusions (\"the isotropic depolarization usually hiddens\"), and \"ad\" in the phrase before \"treatment\" in Sec. 3.1. These should be corrected.","section":"Abstract and Conclusions"}],"recommendation":"major_revision","confidential_remarks":"The novelty of the filter is essentially a pixelwise division by P3, which is presented as a physical removal of isotropic depolarization. The reported contrast factors are derived, not measured, and the physical realizability issue is unaddressed. The paper may be salvageable if the authors provide a careful noise analysis, independent ground truth for at least one sample, and a transparent treatment of all tested observables; without these, the empirical claims are not yet convincing enough for publication in a serious journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the filter reduces to dividing every non-intensity polarimetric observable by P3, and the paper says so explicitly. That is both the strength and the weakness: the idea is simple, parameter-free, and correctly derived, but the empirical evidence is thin and partly circular.\n\nWhat's new: the authors take the perfect-depolarizer subtraction from the characteristic decomposition (their own Ref. [35]), apply it to Mueller images, and test it on heart and brain sections. The derivation of P_n' = P_n/P3 and D' = D/P3 is transparent and correct. The conceptual motivation—low P3 compresses the dynamic range of all other observables, so rescaling can unmask spatial contrast—is reasonable and could be practically useful in polarimetric imaging.\n\nThe soft spots are substantial. First, the headline contrast increases (3.92x, 7.57x) are not independent performance metrics; they are exactly 1/P3, as Eq. (13) shows. Second, the 'best' observable is chosen per sample after inspecting the data (P1 for heart, D for brain), with no stated selection rule and no multiple-comparison correction. Third, there is no noise characterization, no repeated measurements, no flat-field control, and no validation against histology or any other ground truth. With heart-sample P3 around 0.032, the amplification factor is ~31, so the stress-test's scenario—spatial noise fluctuations being promoted to 'fiber tracts'—is a real possibility and the paper does nothing to rule it out. Fourth, Table S1 reports heart myocardium P1=0.1321, P2=0.0753, P3=0.0320, which violates the IPP ordering P1 <= P2 <= P3 stated in the same document. This suggests either unsorted eigenvalues or a data/list error; either way it needs fixing.\n\nI would send this to peer review, not desk-reject it. The math can be checked quickly, the method is cheap to reproduce, and the application area is active. But the revisions should be major: release the data, add a noise propagation or repeated-measurement analysis, correct the ordering inconsistency, and temper the 'outstanding performance' language. If the effect survives a proper noise analysis, this becomes a useful practical paper for the polarimetric imaging community.","headline":"A transparent, parameter-free rescaling of polarimetric observables by 1/P3; mathematically correct, but the tissue-demonstration claims need noise analysis and independent validation.","tokens_in":21980,"tokens_out":3631,"would_cite":false,"duration_ms":36860,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a parameter-free isotropic depolarization filter removes isotropic depolarization from measured Mueller matrices and amplifies the anisotropic component by 1/P3, revealing heart and brain structures that are…","keywords":["isotropic depolarization","anisotropic depolarization","Mueller matrix polarimetry","characteristic decomposition","indices of polarimetric purity","depolarization filtering","biomedical polarimetric imaging","white matter tract imaging"],"falsifier":"Take a structurally uniform depolarizing phantom (for example, a slab of scattering material with no internal organization), measure its Mueller matrix image with the same polarimeter, and apply the IDF; if the filtered image displays contrast comparable to the heart or brain results, then division by $P_3$ is amplifying measurement noise rather than tissue structure. A quantitative version: for each pixel compute the unfiltered observable's per-pixel uncertainty $\\sigma_Q$ from repeated frames; wherever $P_3(x,y)$ drops below about $3\\sigma_Q / Q(x,y)$, the filtered observable $Q/P_3$ is noise-dominated and should not be interpreted as structure.","tokens_in":20950,"feed_emoji":"🔬","tokens_out":10165,"duration_ms":89685,"temperature":0.7,"pith_summary":"Biological tissues often depolarize light so strongly that the anisotropic part of the depolarization - the part that carries information about fiber orientation and tissue structure - is masked by an isotropic background. The paper's central claim is that this mask can be removed with a simple, parameter-free filter: subtract the perfect-depolarizer term from the measured Mueller matrix and renormalize. After the filter, every standard polarimetric observable (diattenuation, polarizance, indices of polarimetric purity, depolarization index) is rescaled by the factor $1/P_3(x,y)$, where $P_3$ is the index of polarimetric purity tied to isotropic depolarization. In ex-vivo lamb heart sections the filter reveals myocardial and subendocardial boundaries and epicardial edges; in cattle brain sections it identifies individual white-matter fiber tracts that are invisible in both intensity images and unfiltered polarimetric images. If the claim holds, this is a zero-parameter post-processing step that can be added to any Mueller-matrix imaging workflow.","feed_headline":"Divide out isotropic depolarization to reveal hidden tissue","feed_subtitle":"A parameter-free Mueller filter rescales polarimetric images by 1/P3, revealing hidden fiber tracts.","key_machinery":"The central object is the isotropic depolarization filter (IDF), defined as $M_a = M - (1-P_3)(m_{00}\\hat{M}_3)$, where $\\hat{M}_3 = \\mathrm{diag}(1,0,0,0)$ is the perfect depolarizer and $1-P_3$ is its weight in the characteristic decomposition. The filter's action is to remove the perfect-depolarizer term and renormalize; the resulting normalized elements are the original ones divided by $P_3(x,y)$, so every non-intensity observable scales by $1/P_3$. It works because $P_3$ measures the anisotropic fraction of depolarization, and the inequality $P_1\\le P_2\\le P_3$ means a low $P_3$ compresses the dynamic range of $P_1$ and $P_2$; setting $P_3$ to 1 projects the data onto the top surface of the Purity Space where tissue classes separate.","core_discovery":"The core discovery is the isotropic depolarization filter (IDF). Writing any Mueller matrix through its characteristic decomposition as $$M = P_1(m_{00}\\hat{M}_{J0}) + (P_2-P_1)(m_{00}\\hat{M}_1) + (P_3-P_2)(m_{00}\\hat{M}_2) + (1-P_3)(m_{00}\\hat{M}_3),$$ the last term is the isotropic, perfect-depolarizer contribution. Setting $M_a = M - (1-P_3)(m_{00}\\hat{M}_3)$ and renormalizing gives $m'_{00}=P_3 m_{00}$ while all other elements are unchanged before normalization; consequently every non-intensity observable satisfies $Q'(x,y)=Q(x,y)/P_3(x,y)$. Because $P_3$ is the proportion of anisotropic depolarization, dividing by it magnifies the anisotropic channel. In the heart sample the filtered $P_1$ channel separates myocardium from subendocardium and shows epicardial boundaries; in the brain sample filtered diattenuation $D'$ resolves white-matter tracts by their orientation, with data-cloud dispersion in the Purity and CP spaces increasing by factors 3.92 and 7.57.","pith_inferences":["A corollary the paper leaves implicit is a noise floor: since the filter divides every observable by $P_3(x,y)$, its useful dynamic range is bounded by the accuracy of the measured Mueller matrix; users should mask or smooth pixels where $P_3$ is comparable to the per-pixel uncertainty of the unfiltered observable.","The same $1/P_3$ scaling suggests the IDF could be composed with other Mueller decompositions (Lu-Chipman, Arrow, symmetric); the paper only hypothesizes that retardance-based contrast gains are negligible, so a derivation or numerical test would settle that claim.","Because the filter is parameter-free and increases inter-class separation by a factor near $1/P_3$, it should transfer directly to automatic tissue classification pipelines as a preprocessing step, though the paper does not report classification experiments."],"forward_implications":["Any Mueller-matrix imaging pipeline can apply the IDF as a post-processing step with no extra measurements and no free parameters; the filtered observable is just the original divided by the per-pixel $P_3$.","In soft tissues where $P_3$ is small, the contrast gain is large: the heart sample shows myocardial and subendocardial borders and epicardial edges that are absent from the unfiltered $P_1$ image.","In brain tissue, filtered diattenuation $D'$ resolves individual white-matter tracts (superior longitudinal fasciculus, cingulum, callosal U-fibers, internal-capsule coalescence) that unfiltered $D$ cannot separate.","The filter increases inter-class dispersion in Purity and CP spaces by a factor near $1/P_3$ (3.92 for the heart regions, 7.57 for the brain regions), so it should improve tissue classification as well as visualization.","Because the filter is general, any sample with significant isotropic depolarization, not only biomedical tissues, can be treated the same way."],"supporting_citations":[{"why":"Establishes that P1 and P2 encode anisotropic depolarization while P3 encodes isotropic depolarization, and gives the split M = P3 m00 Ma + (1-P3)m00 Miso that the filter exploits.","marker":"[35]"},{"why":"Provides the characteristic decomposition of the Mueller matrix into four physically interpretable terms weighted by IPP combinations.","marker":"[36]"},{"why":"Defines the indices of polarimetric purity as eigenvalue-based observables of the covariance matrix, used to derive the filtered IPP.","marker":"[37]"},{"why":"Supplies the Mueller-matrix observables (diattenuation, polarizance, P_Delta, Ps) and covariance formalism used for the filter calculations.","marker":"[39]"},{"why":"Demonstrates that diattenuation imaging reveals white-matter fiber tracts, the baseline the brain results extend.","marker":"[33]"},{"why":"Establishes the indices of polarimetric purity as contrast observables for biological tissue imaging, the baseline for the heart result.","marker":"[14]"}],"fun_headline_variants":["Isotropic depolarization filter reveals hidden tissue structures","Depolarization filter that removes isotropic part reveals hidden tissue","Filter dividing by isotropic depolarization exposes tissue details","Revealing heart and brain detail by filtering out isotropic depolarization","Isotropic depolarization filter sharpens hidden tissue contrast"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the measured Mueller matrices are accurate enough that pixel-level values of $P_3$ - often below 0.15 and as low as about 0.03 in the supplementary data - reflect genuine anisotropic tissue structure rather than detector noise, calibration error, or depolarization-estimation error.","fun_headline_variants_meta":{"raw":{"variants":["Isotropic depolarization filter reveals hidden tissue structures","Depolarization filter that removes isotropic part reveals hidden tissue","Filter dividing by isotropic depolarization exposes tissue details","Revealing heart and brain detail by filtering out isotropic depolarization","Isotropic depolarization filter sharpens hidden tissue contrast"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001718,"raw_usage":{"total_tokens":6807,"prompt_tokens":966,"completion_tokens":5841,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":582,"completion_tokens_details":{"reasoning_tokens":5763}},"tokens_in":582,"tokens_out":5841,"duration_ms":46045,"temperature":1.0,"reasoning_tokens":5763,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T17:53:48.008462+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a structurally uniform depolarizing phantom (for example, a slab of scattering material with no internal organization), measure its Mueller matrix image with the same polarimeter, and apply the IDF; if the filtered image displays contrast comparable to the heart or brain results, then division by $P_3$ is amplifying measurement noise rather than tissue structure. A quantitative version: for each pixel compute the unfiltered observable's per-pixel uncertainty $\\sigma_Q$ from repeated frames; wherever $P_3(x,y)$ drops below about $3\\sigma_Q / Q(x,y)$, the filtered observable $Q/P_3$ is noise-dominated and should not be interpreted as structure.","supporting_citations":[{"cited_title":"Canabal-Carbia, I","cited_arxiv_id":null,"evidence_quote":"Establishes that P1 and P2 encode anisotropic depolarization while P3 encodes isotropic depolarization, and gives the split M = P3 m00 Ma + (1-P3)m00 Miso that the filter exploits."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the characteristic decomposition of the Mueller matrix into four physically interpretable terms weighted by IPP combinations."},{"cited_title":"Invariant indices of polarimetric purity. Generalized indices of purity for nxn covariance matrices","cited_arxiv_id":"0807.2171","evidence_quote":"Defines the indices of polarimetric purity as eigenvalue-based observables of the covariance matrix, used to derive the filtered IPP."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Mueller-matrix observables (diattenuation, polarizance, P_Delta, Ps) and covariance formalism used for the filter calculations."},{"cited_title":"Van Eeckhout, A","cited_arxiv_id":null,"evidence_quote":"Establishes the indices of polarimetric purity as contrast observables for biological tissue imaging, the baseline for the heart result."}],"review_version":1}